Rerum Causae

Rerum Causae
Volume VII, Issue I

August 2015
All rights reserved.

Copyright c Contributors to Rerum Causae 2015.
All views expressed herein are those of the authors and do not necessarily
reflect the views of the editors or the department. This journal claims no
special rights or privileges. All correspondence should be addressed to the
editors:
Rerum Causae,
c/o Department of Philosophy, Logic and Scientific Method,
London School of Economics,
Houghton Street,
London WC2A 2AE
Email: rerumcausae@lse.ac.uk
Cover design: Yi Li and Matilda Kivel
Acknowledgements
The Rerum Causae editorial team deeply appreciates the support

we have received. We are grateful for the generous funding received from both the Department for Philosophy, Logic and Scientific Method at LSE and the P&E programme at Universitt
Bayreuth. The tremendously important guidance from Professor
Luc Bovens and Johannes Himmelreich is very much acknowledged, as well as Ewan Rodgers administrative support. Further
a special thank you to Yi Li and Matilda Kivel for designing the
cover art.
Editors
Luc Batty is an MSc Philosophy of Science student at the

London School of Economics (2014-2015).
He holds a
BA in Philosophy from the University of Warwick (20102013).
His main interests are the philosophy of logic,
mathematics, and language.
You can contact him at
[luc.batty@gmail.com].
Chlo de Canson (editor-in-chief) has just graduated from
LSE with a BSc Philosophy, Logic and Scientific Method
(2012 - 2015). From October 2015, she will be reading
the MPhil Philosophy at the University of Cambridge. Her
main areas of interest are in philosophy of science, philosophy of probability and logic. You can contact her at
[chloe.decanson@gmail.com].
Brian Hendrick is an MSc student in Philosophy and Public Policy at the London School of Economics (2014-2015).
He holds a BA (Hons.) degree with a double major in Political Science and Philosophy from the University of Guelph in
Ontario, Canada. His research interest include digital policy,
cyber affairs, and philosophy of technology. Brian will begin
working for the Civil Service Fast Stream (Digital and Technology Scheme)
next fall. He can be contacted at [brianhendrick@outlook.com].
Lewis Humphreys is a BSc Philosophy, Logic and Scientific

Method student at the London School of Economics (20132016). His main areas of research interest are science and
policy, bioethics and general philosophy of science. Outside
of academia, Lewis works in a pub where often he tests his
philosophical arguments on unsuspecting punters. He can be contacted at
[lrghumphreys@gmail.com].
Yi Li studied MSc Philosophy and Public Policy at the London School of Economics (2014-2015). Her research interests include personal identity, metaethics, and ethical issues
in health and science policy. She hopes to work in health policy, and to one day gain the respect of a cat. You can contact
her at [isthisyi@gmail.com].
Gourav Krishna Nandi is an MSc student in Philosophy of
Science at the London School of Economics (2014-2015).
He holds two bachelor degrees from Montana State University Bozeman in history of science and philosophy, with a
minor in mathematics. His research interests include history
and philosophy of biology, and history of medicine in 19th
century England. Gourav will be pursuing an MPhil course in the history and
philosophy of science and medicine at the University of Cambridge next fall.
He can be contacted at [nandi.gouravkrishna@gmail.com].
Eunice Tse is a BSc Philosophy, Logic and Scientific
Method student at the London School of Economics and Political Science (2013-2017). Her main fields of interest are
political philosophy, logic and existentialism. She hopes
to enjoy her forthcoming year abroad, after which she will
hopefully know what she actually wants to do. She can be
contacted at [y.tse3@lse.ac.uk].
Contents
Somayeh Tohidi
Putnam and the Reality of Time
Daniel Matthias Mayerhoffer

16
Analytical Approximation of the Results of Schellings Checkerboard Model
Yi Li
30
Is the Use of QALYs as the Unit of Healthcare Justice Unfair to Disabled
People?
Franziska Proprawe
Is the Concept of Rights Independent of the Reasons for Them?
50
Julia-Maria Franz
72
How Ought We Treat Family in Political Liberalism? Implications of the
Separation of Parent-Child Relationship and Marriage
Chloe de Canson
82
The Problem of Old Evidence in Bayesian Confirmation Theory: A Solution
Based on the Difference Between Update and Justification
Geoff Keeling
Evolutionary Biology, Moral Intuitions and the Is-Ought Problem
106
Shambhavi Shankar
A Defence of Smiths Solution to The Moral Problem
118
Conference Programme
129
Preface
The greatest challenge to any thinker is stating the problem in a way that
will allow a solution.
Bertrand Russell
This volume contains the papers that were presented at the third LSE-Bayreuth
student philosophy conference, which took place at LSE on 7th and 8th May
2015. The conference was the result of a joint initiative between the Department of Philosophy, Logic and Scientific Method at the London School
of Economics, and the Philosophy and Economics programme at Universitt
Bayreuth.
The conference brought together eight students, both undergraduates and
postgraduates, as well as two keynote speakers and eight commentators from
both departments. The first keynote speaker, Prof. Luc Bovens from LSE,
gave a talk about the ethics of nudge, and the second, Prof. Olivier Roy, talked
about the logical structure of Scanlons contractualism. The student presentations were on equally varied topics. On the first day, we heard Somayeh
Tohidi discussing Putnams view on the reality of time, Daniel Mayerhoffer
presenting some results on Schellings checkerboard model of racial segregation, Yi Li questioning whether the unit currently used to distribute healthcare
is fair to disabled people, and Franziska Poprawe debating whether the concept of rights is independent of the reasons for them. The second day began
with Julia-Maria Franz exploring how family ought to be treated in political
liberalism, after which Chlo de Canson presented a potential solution to the
problem of old evidence in Bayesian confirmation theory. This was followed
by Geoff Keeling presenting an is-ought problem in the context of moral intuitions and evolutionary biology, and finally by Shambhavi Shankar defending
Smiths solution to the Moral Problem against Brinks externalist challenge.
Both the commentaries and the Q&A sessions that followed the presentations provided an opportunity for students to further defend their arguments in
the light of the fruitful comments and productive objections that were raised.
The richness of the debates that took place can only show, as per Bertrand
Russell, that the presenters introduced problems of such philosophical depth
and interest that they sparked the audiences desire to look for a solution.
By publishing this volume, we hope, through a presentation of the outstanding work by students of LSE and UBT, to give our readers a taste of
the different philosophical traditions at both universities. We also hope to encourage all LSE and UBT philosophy students to take part in this wonderful
experience next year.
The Editors
Rerum Causae, 7 (2015), 1-14

Somayeh Tohidi
Abstract
In his famous article Time and Physical Geometry, Putnam claims
that the presentist view of time is inconsistent with Special Theory of
Relativity (STR). He provides two main formulations for his argument.
In this paper, I first reveal the logical flaw in his first formulation. Second, I will show how it can be amended. Third, I show how his second
formulation is logically sound, given my analysis of his first formulation. Finally, and based on these logical evaluations, I show that Sklars
criticism of Putnams argument is not well founded.
Introduction
Putnams article1 is well-known for using special theory of relativity (STR)
to argue against presentism. In his article he argues that STR has falsified
presentist view of time, i.e. the metaphysical view that all and only things
which are in my present are real. In his argument he introduces a criterion for
reality in terms of a physical relation R, i.e. he says all and only things which
are in relation R to me are real. But he does not get into the detail of what R
is. He also assumes that if one wants to decide about reality of objects based
I would like to thank Dr. Matthew Parker, Dr. Bryan Roberts and Jonathan Burton
for their helpful comments.

Putnam, Time and Physical Geometry.
1
Somayeh Tohidi
on their being in relation R to an observer, then all observers are on a par and
no observer has a privileged status. So the fact that the proposed criterion of
reality (CR) has been stated in terms of relation R to me, does not confer any
privileged status to me, and this criterion can be stated in terms of relation
R to any observer. This principle is known as the principle of no privileged
observer (NPO). Putnam then infers from this principle that relation R in the
criterion of reality is transitive. This inference is one of the most-criticised
steps of Putnams argument, since he does not provide any justification for it.
In this paper, I will give a logical analysis of Putnams inference and
prove that it is not logically sound. I will also suggest a way to amend it,
i.e. I will propose a weaker conclusion which can be inferred from NPO
and suffices for Putnams argument. I will also show that Sklars criticism of
Putnams argument, which is based on criticizing this step of his argument,
is not well founded.Putnam provides two main formulations of his argument
against presentism. These formulations differ in terms of their premises and
logical structure.2 In the first section, I will give a precise account of the
first formulation and identify the dubious step. In the second section, I will
logically analyse this step and falsify it. Then I will propose an alternative
step to replace the falsified one. In the third section, I will show Putnams
second formulation of his argument is logically sound. In the final section,
I will assess one of Sklars criticisms towards Putnams argument and show
that it is not well founded.
I. The First Formulation
In this section, Putnams first formulation of his argument against presentism
is rigorously analysed. In this formulation he argues that if all things that
2
In fact he gives three formulations. But one of them (the one on the 3rd page) is
more like a warm-up and apparently is just intended to pave the way for the other
two. Thats because just after finishing it, Putnam point to a flaw in it and explains
why it does not work. The first main formulation is on the 4th page of the paper
(Ibid, p.243) and the second one is on the pages 7-8 (Ibid, pp.246-7).
2

exist now are real then there is an event in future which is real. Here is his
argument:
1. All things that are simultaneous to me-now are real. (Assumption)
2. There is a relation R such that all and only things that stand in relation
R to me-now are real. (Criterion of Reality) (Assumption)
3. For any observer O, if it is the case that all and only things that stand in
relation R to me-now are real and observer O is real then it is also the
case that all and only things that stand in relation R to observer O are
real. (No Privileged Observer) (Assumption)
4. (1) and (2) imply that all things that are simultaneous to me-now are
in relation R to me-now.
5. (2) and (3) imply that R is transitive.
6. (4) can be considered as a law of nature and according to STR laws of
nature are the same for all inertial observers. Therefore if you-now are
an inertial observer then all things that are simultaneous to you-now
are in relation R to you-now.
7. You-now are simultaneous to me-now and is moving relative to it, with
a velocity that is considerable with respect to speed of light. (Assumption)
8. (7) and STR imply that there is an event x which is in future of me-now
but simultaneous to you-now. (Figure 1)
9. (8) and (6) imply that event x is in relation R to you-now.
10. (7) and (4) imply that you-now are in relation R to me-now.
11. (5), (9), (10) imply that event x is in relation R to me-now.
12. (11) and (2) imply that event x is real which means there is an event in
future of me-now which is real.
3
Somayeh Tohidi
Figure 13
Lets go through the assumptions of this argument to see where they come
from. Line (1) is the antecedent of the conditional proposition this argument
is trying to prove, i.e. if all things that exist now are real then there is
an event in future which is real. So, it is reasonable to include it as one of
the premises and then try to reach the consequent of that proposition. Line
(2) is an important assumption. Putnams notion of reality is absolute and
not relative, i.e. he never talks about reality to an observer.4 However, he
introduces a criterion for this absolute notion of reality in terms of a physical
relation R, which is line (2) of the above argument:
Criterion of Reality (CR): All and only things that stand in relation R to
me-now are real.
3
4
Norton, Einstein for Everyone

This point is very important when we come to investigate Sklars criticism of Putnam.
4

In line (3) Putnam is referring to his third assumption about realness. He
believes there is no privileged observer that can be considered as the sole
reference for determining whether things are real. Every real observer is on
a par with any other in this regard. That is, if the realness of things can be
determined by checking whether they have a certain relation to me-now, then
their realness can also be determined by checking whether they have that
relation to any other real observer. Line (7) is the second among the three
assumptions Putnam makes about realness: (ii) At least one other observer
is real, and it is possible for this other observer to be in motion relative to
me.5
Now that we know all the assumptions, lets go through the steps of the argument. Line (4) of the argument is logically straightforward, but it contains
the following point, which is very important for our discussion in the fourth
section of this paper: In this line, relation R has not been taken to be equivalent to simultaneity. Simultaneity is just a subset of relation R, according to
line (4). So for Putnam, relation R is a physical relation but not simultaneity.6 Line (6) looks like a further assumption but Putnam introduces it as an
implication of STR. According to this line of the argument, the sentence all
things that are simultaneous to me-now are in relation R to me-now can be
considered as a law of nature, i.e. it is true for every inertial observer. This
may be justified by a simple reductio ad absurdum. Lets assume the above
sentence (line (4) of the argument) is not true for every inertial observer and
only true for me-now. Then it means relation R, which is a physical relation,
has specified a simultaneity, i.e. my simultaneity, which is different from
others and this is against STR, according to which there is no absolute simultaneity. Line (8) is one of the key points of Putnams argument. As Figure
1 illustrates, there are infinitely many events which are on your simultaneity
5
6
Putnam, Time and Physical Geometry 241

In one occasion in his paper, Putnam takes relation R to be the relation of simultaneity and argues based on this assumption. But at the end of his argument, he says
this argument does not work because simultaneity to an observer in his coordinate
system is not transitive.
5
Somayeh Tohidi
hyper-surface (simultaneity line in the above figure) but above my simultaneity hyper-surface, and this is exactly what line (7) claims. Looking at other
steps of the argument, one will find them logically straightforward, except for
line (5). It is not clear how NPO and CR imply transitivity of R. The curious
thing is that Putnam does not give any justification for either of these claims.
The missing link between NPO and transitivity of R is well known in the
literature about Putnams paper. For instance Sklar writes:
Why one would think that such a doctrine of No Privileged Observers would lead one immediately to affirm the transitivity of
reality for, given that one has already relativised such previously nonrelative notions as that of simultaneity, is beyond me.7
In fact Putnam makes two claims about the relation between NPO and transitivity of R:
1. If NPO holds then R in CR is transitive. He says: what the principle that There Are No Privileged Observer requires is simply that the
relation R be transitive.8
2. If the relation R in CR is transitive then NPO holds. He writes: the
principle III is satisfied because the relation of simultaneity is transitive.9
Here, principle III refers to Putnams third assumption about realness, which
he also calls NPO. To check whether either of these claims holds, it will be
more convenient to talk in a formal language.
II. Logical Analysis of Putnams Claims
Before starting to formally examine Putnams claims, one point is necessary.
As I said before, Putnam explicitly refers to his third assumption about real7
8
9
Sklar, Philosophy and Spacetime Physics 291.

Putnam, 242.
Ibid., 241.
6

ness as the principle of No Privileged Observer. Here is his third assumption
about realness:
(iii) If it is the case that all and only the things10 that stand in a certain
relation R to me-now are real, and you-now are also real, then it is also
the case that all and only the things that stand in the relation R to you-now
are real.11
By closely examining the third assumption, we will understand that it is not
general enough and can only be considered as one consequence of NPO. Assumption (iii) is about you-now (one specific observer) and is saying that if
realness of things can be determined by their relationship to me-now then
their realness can also be determined by their relationship to you-now. Nevertheless NPO in its general form talks about any real observer, i.e. it says
if realness of things can be determined by their relationship to me-now then
their realness can also be determined by their relationship to any other real
observer.
In what follows, I will translate into formal language the general form of
NPO, and not assumption (iii), which is just one instance of the general form.
x, y: variables
i: I-now
Lx: x is real
xRi: x stands in relation R to me-now
Criterion of Reality (CR): 8x(Lx $ xRi)
NPO: 8y((8x(Lx $ xRi) ^ (Ly)) ! 8x(Lx $ xRy))
As we can see, NPO and CR are sentences in first order logic. These sentences
have different values under different interpretations. That means, even if we
fix the domain of discussion (D), different combinations of the sets R and
10 In
this context the term things refers to any event or observer in space-time.
241.
11 Ibid.,
Somayeh Tohidi
L, lead to different truth values for NPO and CR. It seems the claim Putnam
proposes is that:
1. If NPO and CR are true under an interpretation then that interpretation
contains a transitive R.
2. If CR is true under an interpretation and that interpretation contains a
transitive R, then NPO will also be true under that interpretation.
In what follows, I use semantic analysis of NPO to argue that both of these
claims are false. To falsify the first claim, one interpretation that contains a
non-transitive R, but renders NPO and CR true would suffice. Consider the
following interpretation:
D = {i, u, e}
R = {(i.i), (u, i), (i, e), (u, u), (i, u)}

L = {i, u}
In order to simplify the logical calculations, in this interpretation, I have assumed that there are only three objects in the domain: I-now and two other
things. Under this interpretation:
v(8x(xRi $ Lx)) = 1
because for every x-variant v0 of v, v0 (xRi) = v0 (Lx)
v(8y[(8x(Lx $ xRi) ^ (Ly)) ! 8x(Lx $ xRy)]) = 1
because for y = i, u: v(Ly) = 1 and v(8x(Lx $ xRy)) = 1
and for y = e : v(Ly) = 0 and v(8x(Lx $ xRy)) = 0.
So the values of NPO and CR under this interpretation are 1. However the set
R in this interpretation is not transitive, because (u, i) 2 R and (i, e) 2 R, but
(u, e) 2
/ R.
One may object the above falsification by saying that if we take the do-
main of interpretation (D) equal to the set of real things (L) then NPO and CR
imply transitivity of R. In other words R is transitive among real things. This
8

is true but this conclusion, i.e. transitivity of R among real things is not sufficient for Putnams argument. Because when in step (11), we infer that event
x is in relation R to me-now, we dont yet know it is real. Lets examine the
second claim. To falsify this claim, it suffices to introduce an interpretation
v, which contains a transitive R that makes CR true and NPO false:
D = {i, u, e}
R = {(i, i), (u, i), (e, i)}

L = {i, u, e}
v(8x(xRi $ Lx)) = 1,
because for every x-variant v0 of v, v0 (xRi) = v0 (Lx)
v(8y[(8x(Lx $ xRi) ^ (Ly)) ! 8x(Lx $ xRy)]) = 0,
because for y = e, u : v(Ly) = 1 but v(8x(Lx $ xRy)) = 0.
Therefore under this interpretation, the value of NPO is 0 and the value of
CR is 1. However, the relation R in this interpretation is transitive. So it
seems that Putnams claims about the relation of NPO and transitivity of R
are provably false. However that is not the end of the story. Putnams main
claim, which is vital for his argument against presentism, is the first one, i.e.
if NPO and CR hold then R is transitive. In fact Putnam can derive a weaker
conclusion from NPO and CR which would suffice for his argument. He can
claim:
For any two observers u and e, if e stands in relation R to u, and u
stands in relation R to me-now then e is real.
In a more formal language:
For all e and u, if (e, u) 2 R and (u, i) 2 R then e 2 L
It is important to note that the above claim is not tantamount to transitivity of
R. Thats because we have fixed one of the observers, i.e. me-now. Here is
the proof of this claim:
Somayeh Tohidi
Take the domain of interpretation be D = {i, u, e}. Since v(CR) = 1
and (u, i) 2 R, so u 2 L. That means for y = u, the value of left hand
side (LHS) of NPO is 1. On the other hand we know v(NPO) = 1. That

means for every y the value of LHS and RHS is the same. Therefore
for y = u, the value of RHS is 1, i.e. 8x(Lx $ xRu). On the other hand
we have assumed (e, u) 2 R. So e 2 L.
The fact is that the above statement suffices to rescue Putnams argument.
The only thing we need to do is to replace line (5) of the argument by this
weaker conclusion and everything else will be fine (except for some minor
modifications including merging lines (11) and (12)).
III. The Second Formulation

Everything I discussed so far was about Putnams first formulation of his
argument. In his second formulation Putnam adds an assumption to his argument and draws a stronger conclusion. He argues that if all things that exist
now are real then everything in future and past is also real. He does so
by adding the assumption that for every event x in space-time there is an observer o such that x is in relation R to o and o is in relation R to me-now. By
adding this assumption to the argument mentioned before, he can conclude
that every event in space-time is real.
However Putnams second formulation differs from the first one, not only
in terms of the above assumption but also in one of the steps. In fact he
replaces line (5) of the first formulation by the following:
Principle III then requires that I also count every thing and event
which bears the transitive closure of R to me (i.e., which bears
R to me, or which bears R to something that bears R to me, or
which bears R to something that bears R to something that bears
R to me, or....) as real.12
12 Ibid.,
246.
10

This inference is exactly the weaker conclusion I mentioned above. In this
quotation Putnam infers from NPO, the reality of those objects which are in
the transitive closure of R to me, and not transitivity of R in general.13 So this
second formulation does not have the problem of the first formulation and is
logically sound.
IV. Sklars Criticism

In this section I will argue that Sklars account of Putnams argument is not
precise and that is why some of his criticisms towards Putnam are not well
founded. Here is Sklars account of Putnams argument:
Consider an observer at a place-time. According to the doctrine
in question, events in his future (say) are not determinately real.
But according to relativity there is going to be another observer,
coincident with the first, and hence certainly real to him, since
immediately present to him. Now many of the events future to
the first observer will be present to the second, so long as the
two observers are in relative motion. Indeed, for any future event
(relative to the first observer and spacelike separated from him)
there will be a second observer such that that event is present to
the second observer when the second observer is coincident with
the first. So the future event will be real, relative to the second
observer. But surely being real to is a transitive notion. If the
event is real to the second observer who is real to the first, it must
be real to the first observer, contradicting our original claim that
events future to an observer lack reality for him.14
13 He does not mention any reason for making this change.
He uses this second formu-
lation only to infer a stronger conclusion about reality of everything by introducing

a further assumption. So I presume he was not aware of the fact that inferring this
weaker conclusion is logically sound while inferring transitivity of R is not.
290.
14 Sklar,
11
Somayeh Tohidi
First, in this account Sklar has taken reality to be a relation. However, as
we have seen, the notion of reality for Putnam is not a relation but a predicate.
Putnam introduces a criterion in terms of a physical relation (R) for reality.
And in the rest of his argument, he only deals with this physical relation of R
and not the relation of reality for.
Second, in this account, Sklar has assumed the transitivity of the relation of reality for, while Putnam does not assume transitivity of R but infers it from NPO. The curious point is that Sklar is aware of this fact and
says, elsewhere in his paper as I quoted before,15 that he does not understand
how Putnam infers transitivity of R from NPO. So, maybe since he thinks
this inference is unjustifiable, he omits this step altogether in his account of
Putnams argument. He then targets this assumption, i.e. transitivity of reality for and says it is not a plausible assumption, considering the fact that
the relation of simultaneity for an observer is not transitive in Minkowskis
space-time.
Even if we put aside all the mentioned shortcomings of Sklars account
of Putnam, and assume he is talking about non-transitivity of the physical relation of R, his criticism is still not effective. This is because non-transitivity
of the relation simultaneity for an observer does not have anything to do
with non-transitivity of R. Putnam, in his argument has not taken R to be
equivalent to the relation of simultaneity. For him, the latter is a subset of
the former. So, one cannot infer non-transitivity of R from non-transitivity of
simultaneity.
Conclusion
It has been shown in this paper that a precise investigation of Putnams arguments would reveal that his first formulation is logically flawed, but his
15 Why one would think that such a doctrine of No Privileged Observers would lead
one immediately to affirm the transitivity of reality for, given that one has already
relativised such previously nonrelative notions as that of simultaneity, is beyond
me. (Sklar, 291.)
12

second formulation is sound. The logical flaw is about the link between the
principle of no privileged observer (NPO) and transitivity of relation R. It was
proven that NPO does not imply transitivity of relation R. However, it implies
a weaker conclusion and that is everything, which bears to me the transitive
closure of R, is real. This weaker conclusion suffices for Putnams argument.
Having these in mind, I have shown that Sklars account of Putnams argument is not precise and therefore his criticism of Putnam is not well founded.
That is, because Sklars criticism is based on the assumption that Putnam
takes R to be equivalent to the relation of simultaneity for an observer, while
he does not.
13
Somayeh Tohidi
References
[1] Putnam, H. Time and Physical Geometry, The Journal of Philosophy,
Vol. 64, No.8 (1967): 240-247.
[2] Sklar, L. Philosophy and Spactime Physics, London, University of California Press, 1985.
[3] Norton, J. Einstein for Everyone, Nullarbor Press, 2007.
Somayeh Tohidi is a MSc student in Philosophy of Science at the London

School of Economics and Political Science (2013-2015). She did her BSc
in Chemical Engineering at Sharif University of Technology back in Tehran.
Her interests cover general issues in philosophy of science, logic and set theory, and philosophy of mind. She hopes to continue research in the future.
You can contact her at [somayehtowhidi@gmail.com]
14
Rerum Causae, 7 (2015), 16-28
Analytical Approximation of the Results of

Schellings Checkerboard Model
Abstract
Thomas Schelling shows in his checkerboard model that even mildly
segregationist preferences of the single agents are sufficient to bring
about high rates of segregation. This paper analytically investigates
the model. Taking advantage of the knowledge about the agents preferences, a notion of available neighbourhood compositions (those fulfilling an agents preferences) is developed. The unweighted average
segregation of those available neighbourhoods allows a reasonable approximation of the simulation results, which overrates the actual final
rate of segregation instead of underrating it as done intuitively. Additionally, the concept of neighbourhood compositions possibly serves as
explanations for the behaviour of the simulation.
Introduction
The importance of Schellings spatial proximity model or checkerboard model
for the growth of social modelling as best practice example is uncontrover
The author thanks the Elite Network of Bavaria for financial support for programming the simulating model used to obtain the benchmark results for this paper.
16

sial.1 It undoubtedly provides a powerful tool to tackle the problems in its
fields of application and changed the way in which those problems were perceived. As Zhang puts it: The spatial proximity model has every ingredient
of a great theoretical work: it addresses an important real world issue; it is
simple to understand; it produces unexpected results; and it has deep implications for various social sciences.2
Schelling envisions an area either a line or a grid of cells,3 populated by
a fixed number of agents of two kinds. When applied to residential segregation, those kinds are black/white but there are several other interpretations
such as boys/girls or poor/rich. Agents have preferences over the kind of
their neighbours and move starting from a patterned or more often random
initial distribution to fulfil these preferences. At the time Schelling proposed
his models, most authors thought that the empirically observed residential
segregation was brought about by highly segregationist preferences or very
1
It is wrong to give credit for this model to (solely) Schelling, because actually
James M. Sakoda was the first person to develop a CA [cellular automata, A/N]
based model in the social sciences. The model did not see the light of the day as
part of the CA literature at the time or make any reference to previous CA research.
It was baptised a checkerboard model. Sakoda published the article The Checkerboard Model of Social Interaction in 1971 but the basic design of the model was already present in his unpublished dissertation of 1949. The central goal of his model
was to understand group formation. The model grew out of an attempt to portray
the interaction in a relocation center during World War II (Sakoda 1971, p. 120),
where members of the Japanese minority in the US were evacuated after Japans
attack on Pearl Harbour. (Hegselmann and Flache, Understanding Complex Social Dynamics: A Plea For Cellular Automata Based Modelling). While offering a
broader approach to modelling group formation processes, Sakodas model is able
to explain residential segregation as Schelling does (Sakoda, The checkerboard
model of social interaction). However, this paper will refer to Schelling, since his
2
3
model is the one famously worked with in the literature.

Zhang, Tipping and Residential Segregation: A Unified Schelling Model.
Note here: Pancs and Vriend have shown that the one-dimensional and the twodimensional model are not two versions of the same model but two distinct models.
(Pancs and Vriend, Schellings Spatial Proximity Model of Segregation Revisited.
17
Analytical Approximation of Schellings Checkerboard Model

restrictive social institutions. Schelling finds that, contra this widespread position, even moderate preferences for same-kind neighbours at the individual
level are likely amplified into high segregation at the macro level.4 Therefore, macrobehaviour in a society does not represent its individual members
micromotives.5
This paper investigates agents preferences analytically to work out a concept of neighbourhood compositions, which agents with a certain preference
are happy within. Using that concept, it is possible to analytically approximate the simulation results, i.e. only using data given directly in the model
description. Such a prediction questions the emergence of Schellings simulation results and their importance for showing that segregation comes about.
However, the concept developed in this paper may help to uncover the mechanisms at work in the checkerboard model, too. Thus, it enables to use simulation to understand why and how segregation comes about.
I. Description of the Checkerboard Model Underlying this Paper
The size of the two-dimensional model world is 21 by 21 cells. It has the
shape of a torus (despite the finite size of the world there are no borders
and corners), thus each cell has eight direct neighbouring cells. Seventy-five
percent of the cells are occupied, so the population is 330, the two kinds
having equal numbers. The definition of the neighbourhood of an agent is
taken from Moore, i.e. each agent has a maximum of eight neighbours. When
the satisfaction of an agents preferences is checked, she counts herself as a
neighbour of her own kind; empty cells in the neighbourhood are not counted
at all.
The agents preferences, taken from Schelling, are represented by the following utility function:
4
5
8
<0
ui =
:1
for
for
xi
yi
xi
yi
Schelling, Models of Segregation.

Schelling, Micromotives and Macrobehavior.
18
<q
q
1
0.8
ui
0.6
0.4
0.2
0
0
0.2
0.4
0.6
0.8
xi
yi
Figure 1: Utility function of the agents with q = 0.5
Here, y is the total number of agent is neighbours and x is the number

of agent is neighbours being of the same kind as i. If the ratio is below a
certain threshold q , the agent is discontent. There is no additional (random or
pre-set) factor influencing the calculation whether an agent is content or not.
Figure 1 depicts the utility function.
At the start of the simulation, the agents are distributed randomly. Agents
are randomly picked for a possible movement, but only discontent agents
actually move. They do so to the closest free cell satisfying their preferences.
Distance between two cells is the smallest number of cells that one has to pass
(horizontally and vertically) in order to get from one cell to the other. If there
are two or more cells satisfying an agents preferences in exactly the same
distance, there is a random selection between them, if there is none in the
whole world, the agent stays where she is. By moving, an agent may make
previously unhappy agents in her old or new neighbourhood happy and vice
versa. This mutual influence of agents is not determinable analytically.
19

II. Analytical Remarks Concerning Neighbourhoods
It is possible to classify the neighbourhoods by considering only those features relevant for the agents decision whether they are happy. This means,
neighbourhoods differ in the number of total and own-kind neighbours living within (since those are necessary to calculate the rate of own-kind neighbours). Such a classification yields 45 different neighbourhood compositions,
all bringing about a certain rate of own-colour neighbours, shown in Table 1.6
1
2
3
4
5
1.00
0.50
0.33
0.25
0.20
0.17
0.14
0.13
0.11
1.00
0.67
0.50
0.40
0.33
0.29
0.25
0.22
1.00
0.75
0.60
0.50
0.43
0.38
0.33
1.00
0.80
0.67
0.57
0.50
0.44
1.00
0.83
0.71
0.63
0.56
1.00
0.86
0.75
0.67
1.00
0.88
0.78
1.00
0.89
6
7
8
9
1.00
Table 1: Neighborhood segregation (number of own-kind, vertical)

depending on the number of agents in the neighbourhood (horizontal)
This rate of own-colour neighbours will be called neighbourhood segrega6
Schelling himself is aware of an agents possible neighbourhoods but does not expand on their potential importance to explain the phenomena observed by him.
A possible reason is that he avoids relying solely on computational evaluation
(Schelling, Dynamic Models Of Segregation), even though being willing to Letting a Computer Help with the Work. (Schelling, On Letting the Computer
Help with the Work - Teaching Material), as shown by Hegselmann (Hegselmann,
Thomas C. Schelling and the Computer: Some Notes on Schellings Essay On
Letting a Computer Help with the Work.)
20

tion. By comparing q to the neighbourhood segregations, one can determine
how many neighbourhood compositions are available to agents with a certain
threshold. For q -values as in Table 2, this reveals a continuous almost linear decrease of the number of available neighbourhood compositions with
q becoming more demanding, which is not surprising.
Neighbourhoods
0.1
45
0.2
41
0.3
36
0.4
32
0.5
29
0.6
23
0.7
18
0.8
14
0.9
Table 2: Segregationist preference and number of available neighborhoods
However, the number of available neighbourhood compositions alone does

not allow establishing how hard it is for an agent to find a neighbourhood satisfying her preferences, because certainly not all neighbourhoods are equally
likely to occur and the likeliness of occurrence at a certain point in a simulation run is highly dependent on the previous history of that run: under
random distribution, it is trivial to calculate how often a certain neighbourhood occurs in average (e.g. due to the population specifications a neighbourhood will most likely have 7 agents in it, 3 or 4 of them having the colour
of the agent in question). As purposive movements of agents come into play,
such calculations are not possible, since an agents movement influences the
availability of neighbourhoods for all other agents. This mutual influence of
agents movements on the available neighbourhoods (and thus the satisfaction
21

of their preferences) seems to cause the emergent dynamic of the simulation.
Availability of neighbourhood compositions is determinable by purely
analytical means because it follows directly from properties of the agents,
stated explicitly in the model description. The emergent effect in a checkerboard model is a process of segregation resulting in a state not expected when
analysing the available neighbourhood compositions of an individual agent.
Therefore, on the one hand we should not expect the simulation results to be
predictable by an analysis of the available neighbourhood compositions. On
the other hand, available neighbourhood compositions are a possible key to
understanding why and how segregation comes about.
III. Simulation Results
The simulations show that the checkerboard model with segregationists yields
the results described by Schelling and proven robust by several authors, as
Muldoon et al7 or Zhang8 . Nevertheless, the question is, in which sense those
results are emergent. The final rate of segregation at a macro level is indeed much higher than the agents individual threshold values for own-kind
neighbours. However, as explained above, an analysis of agents preference
is possible by looking at their available neighbourhood compositions, too.
Hence, the simulation results at a macro level should as well differ from
any simple aggregation of available neighbourhood compositions, which does
not take into account the interdependences of agents movements mentioned
above. Clearly, it is such an aggregation to calculate the unweighted average segregation of available neighbourhood compositions for a given q : It
is achieved by listing all available neighbourhoods for an agent with a preference of q , calculating the ratio of own kind agents in each neighbourhood
and getting the average of all ratios. This average segregation of available
neighbourhoods will be called Q because it is analytically calculated from
the threshold value q .
7
8
Muldoon, Smith and Weisberg, Segregation That No One Seeks.

Zhang, Residential Segregation in an All-Integrationaist World.
22

q
Segregation
Difference
0.1
0.570
0.600
-0.030
0.2
0.577
0.645
-0.069
0.3
0.608
0.701
-0.094
0.4
0.677
0.756
-0.069
0.5
0.744
0.779
-0.0354
0.6
0.883
0.847
0.036
0.7
0.954
0.903
0.051
0.8
0.700
0.947
-0.247
0.9
0.582
1.000
-0.4 18
Table 3: Comparison of analytical and simulative results
Table 3 contrasts Q with the simulation results for a certain q . It shows

that Q allows a reasonable prediction of the simulation results, as long as
there is a strict equilibrium tendency.9 Furthermore, for mildly segregationist
preferences, the results are overestimated. The simulation results are considered emergent because they are underestimated when using q to predict
segregation. Thus, analysing available neighbourhoods to predict the simulation results more accurately questions their emergence.10
Nevertheless, there still is a surprising result, namely the difference between the average segregation of available neighbourhoods and average rate
of segregation resulting from the simulation. This emergent feature has to
be explained by further investigating the simulation and working out mechanisms. It was shown that relying on Q instead of q yields reasonable predictions about the final rate of segregation but jury is still out, whether relying
9
Note here, that due to the low variance of final segregation, not only average values
but the result of a single run, too, are predictable.
findings are not solely caused by the neighbourhood definition underlying
10 These
this paper, i.e. similar (and partly even stronger) results are generated when agents
do not count themselves as their own neighbours, as the appended data shows.
23

on Q to make statements about the agents behaviour and thus the simulation
results is reasonable itself.
IV. Reasons for Making Use of the Average Neighbourhood Segregation
to Evaluate the Model and Predict its Results
Speaking of some situation m as preferred by an agent i often perceives that
the agent maximises her utility, if and only if she is places herself in m :
Speaking of some situation m as preferred by an agent i often perceives that
the agent maximises her utility, if and only if she places herself in m :
8m0 6= m : ui (m0 ) < ui (m )
Furthermore, if situations only differ with respect to the amount of a good X,
the agent maximises her utility by getting the amount of X that she gets in
m :
8m0 6= m : x(m0 )
x(m )
Thus, someone first looking at the simulation of the checkerboard model for
agents with mildly segregationists preferences is surprised about the highly
segregated result not because those results could not have been foreseen but
because she implicitly dropped the at least when interpreting the description
a preference for at least q own kind neighbours.
However, q is the neighbourhood segregation, which an agent is willing
to exceed in order to maximise her utility. In the checkerboard model, good
X is the neighbourhood segregation: xyii = x. This means, the amount of X in
the preferred situation x(m ) is exactly the threshold value q :
x(m ) = q
From the agents utility function, it follows that while neighbourhoods
bringing about a lower rate of neighbourhood segregation than q give the
agent less utility, all neighbourhoods bringing about a rate of neighbourhood
segregation of q or higher than q give the agent the same utility:
24
8m0 withx(m0 ) < q : ui (m0 ) < ui (m )

But: 8m0 withx(m0 ) > q : x(m0 ) i x(m )
Thus: 8m0 with
xi
> q : ui (m0 ) = ui (m )
yi
This means, q is no suitable choice to start an evaluation of the simulation process and results with. Available neighbourhoods in general better
capture, when an agents preference is fulfilled. When analytically examining
the model, one knows, which neighbourhoods are available to segregationists
and which rate of segregation they would bring about when taken. However,
one does neither know, how often these neighbourhoods will occur nor how
often they will be taken in an equilibrium stable state. In that situation, it is a
common strategy to assume an equal or normal distribution of actually taken
neighbourhoods over the set of available neighbourhoods. One may then approximate the mean neighbourhood segregation of the taken neighbourhoods
(which equals the final rate of segregation on the macro level in the simulation) by calculating the mean neighbourhood segregation of available neighbourhoods. That means, it is neither a non-reasonable nor an uncommon idea
to use Q when describing or predicting agents behaviour , but several refinements and enrichments e.g. by taking into account parameters of dispersion
are possible. By way of illustration, one may calculate a Q0 < Q by firstly applying a normal distribution of the agents over all neighbourhoods and after
that normally redistributing all agents initially placed on unavailable neighbourhoods over all available neighbourhoods. Since Q0 captures the notion
of an expected movement behaviour of (initially) unhappy agents not taking into account interactions, any differences between Q0 and the simulation
output are likely caused by agents movements influencing other agents. Alternatively, it is a calculation of lower bounds for the final rate of segregation
given a certain q using methods of linear optimisation is worth considering.
In general, Q seems to be one key to capture the behaviour of the model.
25

Conclusion
In this essay, I argued that it is possible to predict the simulation output of
the checkerboard model with segregationists only using data directly derived
from the model description and thus independent from simulating the model.
To do so, one simply has to take advantage of the knowledge about the agents
preferences and the neighbourhood definition used in the model to grasp the
notion of available neighbourhood compositions. Available neighbourhood
compositions are those, with a rate of own kind neighbours of at least the
agents threshold value. With that information it is possible to calculate the
average segregation of available neighbourhoods and use it as reasonable prediction of the simulation results concerning segregation, as long as there is a
strict equilibrium tendency. Moreover, using the average segregation of available neighbourhoods instead of the threshold value as a starting point, the
simulation results in a rate of segregation lower than expected for mildly and
moderately segregationist preferences. Thus, high segregation caused by low
threshold values does not seem to be a surprising effect. Additionally, the
consideration of available neighbourhood compositions points towards the
necessity of explaining the mechanisms at work during a segregating process
in the model and may provide a tool to do so, especially when supported by
calculation of lower bounds.
Appendix
The appendix containing all the simulation outputs, further information about
the evaluation and the model (with its ODD-description) used to obtain the
benchmark results for this paper is available to the reader electronically:
https://www.dropbox.com/sh/6hmrae4dqwlh4uw/AABi9VmU6sceDdRRheiy
Mb85a?dl=0
26

References
[1] Hegselmann, R. Thomas C. Schelling and the Computer: Some Notes
on Schellings Essay "On Letting a Computer Help with the Work".
Journal of Artificial Societies and Social Simulation, 2012.
[2] Hegselmann, R. and A. Flache. Understanding Complex Social Dynamics: A Plea For Cellular Automata Based Modelling. Journal of
Artificial Societies and Social Simulation, 1998.
[3] Muldoon, R. et al. Segregation That No One Seeks. Philosophy of Science, 2012.
[4] Pancs, R. and N.J. Vriend. Schellings Spatial Proximity Model of Segregation Revisited. London: Department of Economics at Queen Mary,
University of London, 1978.
[5] Sakoda, J.M. The Checkerboard Model of Social Interaction. Journal
of Mathematical Sociology (1971): 119-132.
[6] Schelling, T.C. Models of Segregation. American Economic Review
(1969): 488-493.
[7] Schelling, T.C. Dynamic Models Of Segregation. Journal of Mathematical Sociology (1971): 143-186.
[8] Schelling, T.C. On Letting the Computer Help with the Work - Teaching
Material, 1972.
[9] Schelling, T.C. Micromotives and Macrobehavior. New York: Norton,
1978.
[10] Zhang, J. Residential Segregation in an All-integrationaist World.
Journal of Economic Behavior and Organization (2004): 533-550.
[11] Zhang, J. Tipping and Residential Segregation: A Unified Schelling
Model. Journal of Regional Science (2011): 196-193.
27

Daniel Mayerhoffer was a BA student in Philosophy & Economics at the
University of Bayreuth (2011-2015). He just graduated, and is going on to
continue his studies in Computational Social Modelling. In this field, he
hopes to continue research in the future. You can contact him at [daniel.mayer
hoffer@gmx.de].
28
Rerum Causae, 7 (2015), 30-48
Is the Use of QALYs as the Unit of

Healthcare Justice Unfair to Disabled
People?
Yi Li
Abstract
The quality-adjusted life years (QALY) is a measure of health-related
quality of life commonly used in healthcare priority-setting. Years of
life without a disability generates more QALYs than years of life with
a disability, so any method of priority-setting that takes into account the
number of QALYs generated would have a reason prioritise the interests of non-disabled people. I argue that this potential unfairness arises
because several methodological features of health state valuation make
it so that the QALY does not adequately reflect health-related quality of
life with a disability.
The Potential for Injustice

Many contemporary healthcare systems, including the United Kingdoms National Health Service (NHS), make decisions about distribution of scarce resources by determining the number of quality-adjusted life years (QALYs)
My sincere thanks to Professor Alex Voorhoeve, who reviewed this draft, and to
Dr. Jan-Willem van der Rijt for his thought-provoking response at the 3rd LSEBayreuth Student Philosophy Conference.
30
Yi Li
each medical intervention would produce per monetary unit. Various forms
of health state valuation are used to determine how many QALYs a year of
life in a given health state generates. According to current methods of health
state valuation, a year of a disabled persons life produces fewer QALYs than
a year of a non-disabled persons life. Hence, in cases where a life-saving
or life-extending treatment can be given to either a disabled person or a nondisabled person and the treatment would be equally effective in both, a policy
that calls for maximising QALYs demands that we give to the treatment to the
non-disabled person a consequence that some philosophers have argued is
unjust.
John Harris has argued that it is unfair to favour non-disabled people over
disabled people in the allocation of life-saving or life-extending treatments. In
what Harris calls the double jeopardy objection,1 he asserts that it is unfair
for a person who is already disadvantaged (by being disabled) to be singled
out for further disadvantage (by being less likely to receive life-saving or lifeextending treatment) because of that first disadvantage. I will argue that the
double jeopardy objection fails gives a full explanation of why it is unfair to
commit double jeopardy in the case of disability, and that even if the double
jeopardy objection goes through, its scope will be quite limited. Instead, I will
turn my attention to a separate set of concerns about the use of QALYs as the
unit of healthcare justice that existing health state valuation procedures fail
to produce weights that reflect lived quality of life with a disability.
The double jeopardy objection can be interpreted in two ways: We might
take it to be arguing (1) that policymakers have a prima facie duty not to exacerbate existing disadvantages, or (2) that disability is not in itself a relevant
criteria for resource allocation decisions, and so it would be unjust to take it
into account. With regard to the first interpretation, it seems plausible that
we have ethical reasons to avoid causing further disadvantage to people who
are already disadvantaged. But we are not obligated, all things considered, to
avoid committing double jeopardy in every case. Even in healthcare, we can
find instances where committing double jeopardy seems justified. For exam1
Harris, QALYfying the value of life.

31
Is the Use of QALYs Unfair to Disabled People?

ple, if a cancer patient needs a organ transplant, the fact that she has received
a terminal diagnosis (a disadvantage) does, in many cases mean, that she will
be moved down on the transplant list (a second disadvantage). It seems at
least plausible that it is just to favour patients who do not have terminal diagnoses over those who do in allocating scarce resources. If this is so, then, it is
sometimes just to use one disadvantage as a criteria with which to distribute
another disadvantage.
In order to explain why it is unfair to commit double jeopardy in the case
of terminal illness but not in the case of disability, we must turn to the second
interpretation of the double jeopardy objection that it is unfair to distribute
disadvantage on the basis of disability because disability (unlike terminal illness) should not be a relevant factor in healthcare resource allocation decisions. There are relevant differences between the case of terminal cancer and
the cases of disability with which we are concerned. While the terminal cancer patient will die soon with or without the treatment, the disabled patient
is predicted to live roughly as long as non-disabled people typically do. The
condition of a terminal patient may also reduce the effectiveness of the treatment, while this would not be the case for many disabilities. Thus, if it is
these aspects of terminal cancer that make it relevant in healthcare allocation
decisions, then disability is not in it of itself relevant.
Yet, those who favour the use of QALYs to distribute healthcare resources
would assert that disability is a relevant factor (in addition to lifespan, effectiveness of treatment, and so on) because the goal of healthcare is to maximise
quality of life and disabled people experience, on average, a lower quality of
life than non-disabled people.2 After all, in contexts where a healthcare system can choose to distribute resources towards treating a disability or not
doing so, and in contexts where a healthcare system can choose between a
treatment that restores the patient to full health or leaves her in a disabled
health state, disability is a relevant factor. It seems that the the burden of
proof lies with proponents of the double jeopardy objection to explain why
2
For example: Singer et al., Double jeopardy and the use of QALYs in health care
allocation.
32
Yi Li
disability should be relevant in the aforementioned contexts, but not in the
context of life-saving or life-extending treatments. The question of whether
or not disability is relevant in this context is precisely what is at stake when
we ask if it is unfair to use QALYs as the unit of healthcare justice, and the
double jeopardy objection does not get us any closer to answering that question.
Even if the double jeopardy objection goes through, it would not threaten
the use of QALYs as the unit of healthcare justice because not all forms of priority setting that uses the QALY commit double jeopardy. Utilitarian forms
of priority setting (called cost-effectiveness analysis, or CEA) are concerned
solely with maximising the number of QALYs added. In contrast, prioritarian alternatives, like the one proposed by Trygve Ottersen,3 also give special
concern to the worst off by apply equity weights to gains to the worse off.
The question of who counts as the worse off differs from account to account,
but it is very likely to include the disabled and chronically ill. For example, Ottersens account proposes giving priority to the interests of those will
enjoy the fewest QALYs over the course of their lifetime, a group that will
certainly include most disabled people.4 Equity weighting can eliminate or
mitigate the effects of double jeopardy, depending on the magnitude of the
weights used. Even if CEA is vulnerable to the double jeopardy objection,
prioritarian forms of priority setting may not be.
I think the use of QALYs as the unit of healthcare justice face a more
pressing concern the possibility that the QALY is an inadequate measure
of quality of life with a disability. It systematically undervalues disabled
health states, by (1) conflating individual preferences and social preferences,
(2) by directing participants to adopt a perspective that does not reflect their
typical outlook, and (3) by failing to exclude social factors that disadvantage
disabled people. Because my objection targets the QALY itself, it applies not
just to CEA, but also to prioritarian approaches, as well as any other system
of healthcare resource allocation that uses the QALY.
3
4
Ottersen, Lifetime QALY prioritarianism in priority setting.

Ibid.
33

For simplicity, I will limit my discussion to a single method of health state
valuation: the time trade-off method (TTO). Moreover, in my references to
disability, I will assume that the disability in question (1) does not affect the
persons longevity, (2) does not hinder the effectiveness of the treatment, and
(3) does not render the persons life no longer worth living. While I realise
that not all cases of disability match these assumptions, I will exclude these
considerations in order to examine whether disability in it of itself should
make a difference in the allocation of healthcare resources.
I. Individual and Social Preferences
Measures of well-being can be subjective (e.g. a good life is one where ones
preferences are satisfied) or objective (e.g. a good life is one that has certain
features, though they need not be the same features for everyone).5 In order
for a measure of wellbeing to be useful in resource allocation, it must embody
features of both subjective and objective measures. A subjective measure of
well-being will be good at reflecting what people want for themselves (individual preferences) and will maintain neutrality on questions of what constitutes wellbeing, but it will not successfully reflect what people want for others
(social preferences). When it comes to deciding how resources are allocated
in a publicly funded healthcare system like NHS, social preferences should
play an important role. For example, even some individuals very much want
cosmetic surgery for themselves, it still seems justifiable for cosmetic surgery
to not be covered by a public healthcare system if the public at large does not
believe it is the responsibility of the government to fund this service. Conversely, an objective measure of well-being accounts for importance to the
public, but it risks being insufficiently neutral. For example, an objective theory might consider it important for children on the autism spectrum undergo
therapy that ensures they behave in as neurotypical a way as possible, even
though the children themselves and their families do not view the condition
as a dysfunction that must be ameliorated.
5
Parfit, What makes someones life go best.

34
Yi Li
The QALY attempts to find a middle ground, by averaging individual
preferences. However, I argue that this approach is seriously flawed because,
in doing so, many health state valuation procedures use data about individual
preferences to draw inferences about social preferences. For example, the
time trade-off method (TTO) asks participants to imagine that they are in a
given disabled health state d, and to decide how many years off the end of
their life they would give to be in full health instead of in health state d. Their
responses are used to determine how many QALYs a year in health state d
generates (between 0 and 1, with 0 being death and 1 being full health). So
if participants would, on average, give up half their remaining lifespan to live
without disability d, then d is valued at 0.5 and generates 0.5 QALY per year
of life. The more years participants are willing to give up to be rid of D, the
fewer QALYs a year in health state d generates. Though the weight given to
d was calculated using data about individual preferences, that same weight
is used to make decisions about healthcare resource allocation that should
reflect social preferences.
The TTO asks participants to make intrapersonal trade-offs, where one
can gain a benefit (in this case, being non-disabled) at the cost of giving up
some other benefit (years off the end of ones life). In contrast, problems of
resource allocation in healthcare are cases of interpersonal trade-offs where
we can provide a benefit to one group at the cost of failing to provide a benefit to a second group. The TTO conflates individual preferences with social
preferences by attempting to use data about individuals willingness to make
an intrapersonal trade-off to draw inferences about their willingness to accept interpersonal trade-offs. But, as Frances Kamm points out, the willingness of individuals to make an intrapersonal trade-off is no indication of their
willingness to make the corresponding interpersonal trade-off.6 Consider the
following set of cases:
Years Imagine that if by giving up years off the end of your life, you could
become ten times wealthier than you currently are. How many years
6
Kamm, Bioethical Prescriptions 424-6.

35

would you give? Would you give any?
Life-saving treatment You can give a life-saving treatment to either Tonya
or Maria. They are equivalent in all respects except that Tonya is ten
times as wealthy as Maria. To whom would you give the treatment?
I would think that, in response to Years, many people would be willing
to give up at least a few years to gain an attribute they genuinely value, be it
wealth, intelligence, or something else. In contrast, I would suspect that few
would respond that we should favour Tonyas interests over Marias in Lifesaving treatment. We can prefer being wealthy to not being wealthy, without
preferring wealthy people over middle-class people. This is a case where we
are willing to make an intrapersonal trade-off, but where we are not willing to
make the corresponding interpersonal trade-off. Yet, the time trade-off methods asks questions of the type in Years to draw conclusions about what we
should do in situations like Life-saving treatment.When we replace gains in
health with gains in wealth, intelligence, or some other valued attribute, we
can recognise that in doing so, the TTO problematically conflates individual
preferences with social preferences. In the same way that we may give a positive number in Years but refuse to favour Tonya in Life-saving treatment,
we might be willing to give up years of our life to be rid of a certain disability
on the TTO without being willing to favour the interests of people who do not
have that disability over the interests of people who do.
Empirical evidence supports the idea that social preferences differ in significant ways from individual preferences. Studies demonstrate when individuals are asked directly about how healthcares resources should be distributed,
their responses diverge from conclusions drawn from responses to health state
valuation procedures.7 For example, the public tends to favour giving greater
shares of resources to help those who are more severely ill, even when those
resources would help them less (that is, produce fewer added QALYs) than
they would help those who are less severely ill.8
7
8
Reviewed in Menzel, Allocation of Scarce Resources.

Nord, Health politicians do not wish to maximise health benefits.
36
Yi Li
Kamm explains this phenomenon by pointing out that the moral relevance
of a given benefit depends on the context. For example, if we could either
give a flu vaccine to one person or give the vaccine to another person and
cure her headache in the process, it seems the additional benefit of alleviating
a headache is significant enough to motivate us to give the vaccine to the
second person. At the same time, if we can save one persons life or save
another persons life and cure her headache in the process, it seems that we
can be reasonably ambivalent about who to save. In the context of receiving a
flu vaccine, the added benefit of curing a headache is morally relevant, but in
the context of receiving a life-saving treatment, that same benefit is no longer
morally relevant. Kamm suggests that in the same way, the added benefit of
being non-disabled may be morally relevant in the context of the TTO, where
only years at the end of ones life are at stake, but morally irrelevant in the
context of receiving a life-saving treatment. Kamm calls this the Principle of
Irrelevant Goods.9
This argument provides what the double jeopardy objection lacks an
explanation as to why disability is not relevant in the context of life-saving
and life-extending treatments. In such contexts, disability is irrelevant because the benefit that each patient stands to receive is so great that it swamps
the benefit of being non-disabled. A disabled person might want badly enough
to be in full health that she is prepared to give up some years off the end of her
life in order to do so, but still value her life just as much as a non-disabled person person values theirs just as a middle class person might give up many
years of their lives to be wealthy, but still value her life just as much as very
wealthy people value theirs. Indeed, empirical evidence demonstrates that
disabled people even those with very severe disabilities tend to report
that they are very or somewhat satisfied with their lives.10 As Kamm emphasises, the idea that if one can have only x, one cares about it as much as one
9
Kamm, Bioethical Prescriptions 424. Kamm has complicated her view on this issue
in more recent work, but I will not discuss her work in very much detail here.
Harris and Associates, The ICD Survey of Disabled Americans; Albrecht and
10 Louis
Devlieger, The disability paradox.

37

would care about y if one had it does not mean that one cares to have x as
much as one cares to have y.11 If this so, then even if disabled participants report that they would give up a significant number of years to be non-disabled
on the TTO, the resulting value should not be used to draw inferences about
how life-saving or life-extending resources should be allocated.
This argument is corroborated by empirical findings that suggest that
though participants are willing to make intrapersonal trade-offs when it comes
to disability, the public is averse to making interpersonal trade-offs that favour
non-disabled people over disabled people. A study in Norway found that participants were unwilling to favour non-disabled people over paraplegic people
in distributing a life-saving or life-extending treatment.12 Participants may be
particularly averse to making interpersonal trade-offs when it comes to disability because they recognise that being disabled is a socially salient trait and
are sensitive to the danger of discriminating against disabled and chronically
ill people.
The time trade-off method conflates individual preferences and social
preferences by attempting to use data about peoples willingness to make
intrapersonal trade-offs to draw inferences about their willingness to make
interpersonal trade-offs. This ignores the potential that a benefit might be
very important in one context but irrelevant in another. We have good reason to believe that the benefit of being non-disabled, like the benefit of being
wealthy, is one of these cases: That most people are willing to sacrifice years
of their life to be non-disabled is not necessarily an indication that we as a
society are or should be willing to prioritise the lives of non-disabled people
over the lives of disabled people.
11 Kamm,
Deciding Whom to Help 235.
12 Nord, The relevance of health state after treatment in prioritising between different
patients.
38
Yi Li
II. Perspectives and Interval Scales
The discrepancy in our intuitions between Ten years and Life-saving treatment can be explained in another way as well. In Life-saving treatment, we
are aware that both people in the thought experiment need treatment and will
die if they do not receive it. The immediate prospect of death may affect our
evaluation of the case. The spectre of death in Life-saving treatment makes
extravagant wealth seems much less important than it did in Years. In this
section, I argue that our perspective affects of our valuation of various health
states, and that the time trade-off method imposes an artificial perspective that
does not reflect the way disabled people ordinarily view their health states.
Suppose that Aiko is in full health and believes that quadriplegia is only
slightly better than death. She claims that if she became quadriplegic tomorrow, she would prefer to be euthanised because being a quadriplegia is
preferable to death by such a marginal degree that continuing to live as a
quadriplegic person would not merit the cost, effort, burden to family, etc
of doing so. Yet, if Aiko had a terminal illness and was offered a treatment
that would save her life but leave her quadriplegic, she may be very happy to
accept the treatment. That is, her view of quadriplegia in face of imminent
death may differ significantly from her view of quadriplegia when she is in
full health. In other words, Aikos perspective affects her evaluation of various health states. Her evaluation of quadriplegia from the perspective of full
health (1.0) is dramatically different than her evaluation of quadriplegia from
the perspective of imminent death (0). The story I have told about Aiko is an
intuitive one the world is rife with anecdotes about how ones appreciation
of some thing changed dramatically once it was in immediate danger of being
taken away.
The value of any given health state is supposed to fall on an interval
scale the difference between 0.1 and 0.2 is meant to be equivalent to the
difference between 0.9 and 1.0. Yet, in practice, the perspective from which
we think about various health states affects the way we perceive these intervals. The difference between health states that are closer to our perspective
39

are exaggerated while the difference between health states that are further
away from us seem smaller. That is, from the perspective of someone at
a health state of 1.0 or close to it, the difference between quadriplegia and
death may seem small, but from the perspective of someone at a health state
of 0, the difference between death and quadriplegia seems very significant
indeed.
Note that our perspective does not necessarily have to be the health state
we actually occupy.13 People with an acute illness will maintain the perspective of full health, because they are aware that their reduced state of health
is not permanent. For example, if Aiko is typically in full health, then having even a very bad case of pneumonia for a week is unlikely to change her
perspective about the badness of quadriplegia in the way having a terminal illness might. Conversely, when Aiko is terminally ill, she considers quadriplegia from the perspective of death, because she is imagining a future where is
either dead or quadriplegic.
I argue that TTO undervalues disabled health states by asking participants
to take a perspective they do not usually occupy. Though the TTO method
ostensibly asks participants to occupy the position of someone with the disability in question, it also presents full health as an open possibility. With
some exceptions (e.g. deaf people eligible for cochlear implants), full health,
or anything close to it, is not an open possibility for disabled or chronically ill
people and if that option became available, their valuation of their own health
state would likely change. Empirical research commonly finds that the array
of choices available changes the extent to which a person is satisfied with
their own situation.14 . In the same way, it is plausible that a persons valuation of a health state depends on what options are available.15 For example,
13 In
this way, this phenomenon is not the same as adaptation. Adaptation requires a
change in perspective through being at a reduced health state for some significant
period of time, but it is necessarily (though helpful) to be at a given health state at
all in order to take that perspective.
The Paradox of Choice.
15 This line of argument is akin to Kamms Sufficiently Good Only Option (SGOO)
14 Schwartz,
argument, which argues that when an option is sufficiently good and is the only
40
Yi Li
a deaf person may view her current situation much less positively once she
realises that she is in fact eligible for a cochlear implant, just as Maria might
feel much less satisfied with her life once she became aware that her college
classmate, Tonya, is now extremely wealthy.
It is a hotly contested question whether health state valuation should utilise
people with the disability being investigated or those without it. I believe that
the problem I have raised about perspective applies irrespective of whether or
not the participants have the disability in question. Regardless of whether or
not the participants have the disability, they are asked to envision a world in
which that disability is curable and that is not the perspective from which
disabled people typically regard their lives. Because TTO asks participants to
make valuations about disabled health states when full health is an option, it
asks participants to step into a perspective that most disabled people do not
typically occupy. And since our evaluations of health states depends heavily
on our perspective, this may mean that TTO produces distorted valuations
of disabled health states that do not reflect the lived experience of disabled
people.
III. Social Disadvantage
Finally, I argue that health state valuation, whether through TTO or any other
method, is almost certain to confound the social ill effects of disability with
its health-related ill effects.
QALYs are supposed to be measures of health-related quality of life
and only health-related quality of life.16 So long as people are in the same
available option, people will want it just as much as they want a better option. However, while the SGOO argument is concerned with the question of whether disabled
people value their lives as much as non-disabled people value theirs, my argument
points to a more general problem about how health state valuation procedures may
misrepresent disabled peoples lived quality of life. My argument may apply even
if the disability in question does not allow for a sufficiently good quality of life or
even if it is not the only available option (if other options are not easily accessible).
and Hirose, The value of health.
16 Bognar
41

health state, years of their lives are to generate the same number of QALYs,
even if social factors like wealth, race, or gender may have a dramatic effect
on the quality of their life. QALYs are not meant to take these social factors into account. There are good reasons for this. QALYs are supposed to
pick out the health states that require medical treatment that is, that require
changes be made to the person with that health state. In cases of social injustice, we think that changes should be made to the social conditions that create
the injustice, not to the person who suffers it. In fact, it would seem perverse,
for example, to suggest that black Americans use plastic surgery to become
white or that women in patriarchal societies undergo gender reassignment in
order avoid injustice. When black people and women face injustice, the problem is not with their being black or women, but with the social conditions that
make their lives worse because of it. It is these social conditions that must
be ameliorated and a focus on the individuals rather than the society distracts
from the real nature of the problem.
To some extent, the same is true of disability. Insofar as the badness of
a disability is the result of social conditions, those ill-effects should not be
taken into account by valuations of health-related quality of life. Though disability is different than class, race, or gender in that it typically involves a
reduction in health-related quality of life, it also involves the social disadvantages that make the lives of disabled people worse than they otherwise would
be. Here, I borrow the distinction between impairment and disability from
the social model of disability, where impairment refers to the health state that
marks one out as disabled, and disability refers to the social conditions that
makes worse the experience of having that health state.17 For example, a person diagnosed with clinical depression is impaired by the sense of low self
worth and suicidal ideation she experiences, but she is disabled by the social
isolation she faces because of stigma against people with mental health problems. The Office for Disability Issues report that almost a third of disabled
people reported difficulties in accessing goods and services.18 In the case of
17 Morris,
18 Office
Impairment and Disability.

for Disability Issues and Department for Work and Pensions, Disabilty facts
42
Yi Li
certain disabilities, like autism or deafness, the social disadvantage may be
most or even all of what make the condition bad. For some autistic people or
deaf people, the goal is not to be in full health, but to live in a world where the
social disadvantages of being autistic or deaf are ameliorated. There is little
doubt that in the case of almost every disability, there are ways in which a
change social conditions could dramatically improve the lives of people with
that disability, even if the severity of their impairment remains exactly the
same. Taking these considerations into account in health state valuation is
problematic not just because it abandons the aim for the QALY to be a measure of health-related quality of life, but also because it obscures the impetus
to change the social conditions that make disabled peoples lives worse than
they have to be.
This is not a problem for valuations of acute health conditions. Though
the same social facts that make life unnecessarily difficult for people who
experience fatigue because of a chronic condition would also make life unnecessarily difficult for someone who experiences fatigue as a symptom of
pneumonia, people who have pneumonia are not disadvantaged in the way
that disabled people are. Ones having pneumonia does not come define them
in any significant way in the way that one having a chronic illness does. Disabled people face social disadvantage precisely because their conditions persist for long enough or because the nature of their condition is such that it
becomes an inextricable part of the way others see them and the way they
seem themselves.
I have argued that the social disadvantages faced by disabled people should
not be taken into account in health state valuation. Yet, qualitative studies indicate that participants often take into account the social factors that affect
the experience of living with a disability. For example, some participants in
Baker and Robinson (2003) were concerned that in their disabled state, they
would be a burden or a source of emotional anguish to their families or that
there would be no one to care for them.19 These are not factors that necesand figures.
and Robinson, Response to standard gambles.
19 Baker
43

sarily come along with having a disability. If social care were more readily
available and/or if being a carer or having a disability were not so stigmatised,
these participants may have felt differently.
More subtly, participants were extremely concerned about their prospects
for independence20 what they would be able to do for themselves and what
they would need others to help with. The answers suggested that the participants felt their lives would be of lesser value if they were made dependent
on other people because of their disability, and that to be dependent on others
was to be a burden to them or to impose on them. Such considerations
might be viewed as a reflection of ableist attitudes in society. Though all people require help from others to survive most of us need help from farmers
to provide us with food, for example only the forms of help that disabled
people require are stigmatised.21 We might imagine that in a perfectly just
world, care would not only be available for those who require it, but that
people would not feel badly about requiring it.
It is not surprising that participants in health state valuation would take
into account social disadvantage. The social factors that disadvantage disabled people are so deeply entrenched in our mental picture of what it means
to have that disability that it may be difficult for people to imagine what life
with a disability would be like if these conditions were entirely rectified. Even
if participants were explicitly asked to disregard the social factors that disadvantage people with disabilities, it may be very difficult or impossible for
them to do so. These social factors may act as a confound that causes these
procedures to undervalue disabled health states.
A Divergence in Goals
I have argued that the time trade-off method may underestimate the quality
of life experienced by disabled people because it (1) conflates individual and
social preferences, (2) imposes an artificial perspective that may not reflect
20 Ibid.
21 Silvers,
Reconciling Equality to Difference.

44
Yi Li
the way disabled people typically regard their lives, and (3) and allows participants to take into account social factors, which confound valuations of
health-related quality of life.
Taken together, I believe these issues give us reason to believe that procedures like the time trade-off method systematically undervalue disabled
health states. Undervaluing disabled health states is problematic not just because it introduces inaccuracy into the process of healthcare priority-setting,
but because they do so in a way that systematically disadvantages some of the
worst off. Many governments accept that they have a particular duty to protect the interests of several groups that have been historically disadvantaged,
including disabled people. For example, in the UK, the Equality Act 2010
places a legal duty on policymakers to consider how each new policy will
impact these protected groups.22 Worse, it is not clear how the time trade-off
method and other methods of health state valuation can be improved to avoid
these flaws: Offering the possibility of being non-disabled is a necessary part
of the TTO, and it will always be difficult to untangle the medical and social
factors that are responsible for the disutility of having a disability.
I think these issues should lead us to seriously consider the possibility that
chronic and disabled health states should be evaluated in a categorically different way than acute health states. As Anita Silvers points out,23 with acute
conditions, the goal of healthcare is to cure or ameliorate the condition to the
fullest extent possible. Methods of health state valuation like the TTO reflect
this goal. But disabled and chronically ill people, many of whom cannot be
cured or do not want to be cured, do not share this goal: Their good is not the
common good, for the common good is to be cured, and that is a good from
which, in virtue of their chronic condition, they are definitely removed.24
For many disabled people, the improvement in the quality and length of their
lives (quantified as the number of QALYs) is not the only or even primary
goal of healthcare. For them, healthcare is a constant presence in their lives,
22 Equality
Act 2010
Judgment and Justice.
24 Silvers, Judgment and Justice 369.
23 Silvers,
45

and good healthcare, ideally, would be just one of a set of complex factors
enables them to have the fullest, most productive lives possible. For example,
less intrusive healthcare options may be preferable for disabled people, even
if they generate fewer added QALYs. Further discussion is merited about
whether priority-setting should or can be revised to better reflect the role that
healthcare plays in the lives of disabled people.
46
Yi Li
References
[1] Albrecht, G.L. and P.J. Devlieger, The disability paradox: high quality
of life against all odds. Social Science and Medicine, 1999. 48: 977988.
[2] Baker R. and A. Robinson, Responses to standard gambles: are preferences "well-constructed"?. Health Economics 2004; 13: 37-48.
[3] Bognar, G. and I. Hirose, The value of health, In: The Ethics of Health
Care Rationing: An Introduction London: Routledge; 2004: 29-52.
[4] Equality
Act
2010.
Available
from:
http://www.legislation.gov.uk/ukpga/2010/15/contents/,
(Accessed
15 July 2015).
[5] Harris, J., QALYfying the value of life. Journal of Medical Ethics
1987; 13: 117-123.
[6] Kamm, F.M., Bioethical Prescriptions: To Create, End, Choose, and
Improve Lives Oxford: Oxford University Press; 2013.
[7] Kamm, F.M., Deciding whom to help, health-adjusted life years and
disabilities, In: Anand S., F. Peter, and A. Sen (eds.) Public Health,
Ethics, and Equity Oxford: Oxford University Press; 2004: 225-242.
[8] Louis Harris & Associates Inc. The ICD Survey of Disabled Americans:
Bringing Disabled Americans into the Mainstream 1986, Washington,
DC.
[9] Menzel, P., Allocation of Scarce Resources, In: Rhodes R., L.P. Francis, and A. Silvers (eds.)The Blackwell Guide to Medical Ethics Oxford:
Blackwell Publishing; 2007: 305-322.
[10] Morris, J., Impairment and Disability: Construction an Ethics of Care
That Promotes Human Rights. Hypatia 2001; 16(4): 1-16.
47

[11] Office for Disability Issues and Department of Work and
Pensions,
Disability
facts
and
figures
Available
from:
https://www.gov.uk/government/statistics/disability-facts-and-figures,
(Accessed 23 July 2015).
[12] Ottersen, T., Lifetime QALY prioritarianism in priority setting. Journal of Medical Ethics 2013; 39: 175-180.
[13] Parfit, D., What makes someones life go best. Reasons and Persons
Oxford: Oxford University Press; 1984: 493-502.
[14] Silvers, A., Judgment and Justice: Evaluating health care for chronically ill and disabled people, In: Rhodes R., L.P. Francis, and A. Silvers (eds.) The Blackwell Guide to Medical Ethics Oxford: Blackwell
Publishing; 2007: 354-372.
[15] Silvers, A., Reconciling Equality to Difference: Caring (F)or Justice
for People with Disabilities. Hypatia 1995; 10(1): 30-55.
[16] Singer, P., J. McKie, H. Kuhse, and J. Richardson, Double jeopardy and
the use of QALYs in health care allocation. Journal of Medical Ethics
1995; 21: 144-150.
[17] Schwartz, B., The Paradox of Choice: Why less is more New York:
Ecco; 2004.
Yi Li studied MSc Philosophy and Public Policy at the London School of

Economics (2014-2015). Her research interests include personal identity,
metaethics, and ethical issues in health and science policy. She hopes to work
in health policy, and to one day gain the respect of a cat. You can contact her
at [isthisyi@gmail.com].
48
Rerum Causae, 7 (2015), 50-70
Is the Concept of Rights Independent of the

Reasons for Them?
Franziska Poprawe
Abstract
Wenar (2005) rejects both the will and the interest theory, as the correct theory of the function of rights, partly on the grounds that they are
constructed to favour particular positions in normative theorising, and
thus fail to do justice to the desirable demarcation between normative
stipulation and descriptive conceptual analysis. As an alternative, he
proposes the so called Several Function Theory. I argue that this theory, though superior to the will and the interest theory in some respects,
does not provide us with the purely conceptual, non- normative account
of rights that it purports to be. I argue that, though it does not en f orce a
particular normative theory in the way the will and the interest theories
do, it nevertheless relies on normative theorising. My analysis suggests
the speculation that a concept of rights is necessarily bound to normative
commitments.
Introduction
There is substantial disagreement about rights. A significant part of moral
and political philosophy is devoted to the idea that people have rights. Dis
I would like to thank Leif Wenar for the helpful comments and inspiring discussion
on an earlier version of this paper.
50
Franziska Poprawe
agreement in this literature on rights concerns two major questions:
(I) What are rights?
(II) Who ought to have which rights?
On the one hand, rights theorists address the conceptual question (I) of what
rights are. They seek to define the concept of rights. On the other hand, the
normative question (II) of what rights people have or should have is heavily
discussed in diverse contexts.
This article concerns the question of what rights are. It concerns the debate on the concept of rights. In this debate, Wesley Hohfelds four-instancesframework is broadly accepted as the logical form of rights. It classifies rights
into claims, privileges, immunities and powers. But beyond this agreement
on the logical structure of rights, what makes something a right is highly
disputed. Two theories have dominated the debate on the concept of rights
for decades. Both point towards the function that rights have for the rightholder to define what makes something a right. Will theorists hold that rights
make the right-holder a small scale sovereign1 They claim that rights are
entitlements that give the right holder control over duties of others. In contrast, Interest theorists argue that the only function of a right is to advance the
right-holders interests.2
Leif Wenar rejects both the will and the interest theory on various grounds.
Importantly, he reveals that the two theories are constructed to support the interpretations of rights entailed by moral theories, Kantianism and welfarism,
respectively. The debate between will and interest theorists over the concept
of rights, he argues, has become a proxy for the debate between Kantianism
and welfarism.3 Thus, the will and the interest theory both fail to do justice
to the desirable demarcation between normative stipulation and descriptive
conceptual analysis. As an alternative, Wenar proposes the Several Function
Theory according to which all those Hohfeldian incidents are rights which
1
2
3
Hart, Essays on Bentham: Studies in Jurisprudence and Political Theory.

Wenar, Rights.
Wenar, The Nature of Rights 224.
51

perform one of six specified functions. He argues that this theory fulfils the
standards we demand of a conceptual analysis of rights that does touch on
any normative dispute. The Several Function Theory, Wenar argues, solves
the conflict between the will and the interest theories and is superior to them
on the grounds that it captures best our ordinary understanding of what rights
are.
In this article, I argue that the Several Function Theory proposed by Wenar, though indeed superior to the will and interest theories in some respects,
does not provide us a purely conceptual analysis of rights. It does not give
us a satisfactory account of what rights are that is entirely detachable of our
normative reasoning. I argue that, though it does not en f orce a particular normative theory in the way the will and the interest theories do, it nevertheless
relies on normative presuppositions.
To do so, I first outline Wenars argument for his Several Function Theory
in section one. In section two, I investigate how this theory functions as a
framework to explicate rights assertions by applying it to an example case.
I conclude that the Several Function Theory does not allow us to determine
what classifies as a right without presupposing and at least implying some
commitments in normative theory. I elaborate on implications of this result
in section three. Further, I explore the possibility of generalizing my conclusion of the analysis concerning Wenars Several Function Theory to the
broader thesis that the concept of rights is necessarily normatively charged.
This in turn implies that there is no one unifying concept of rights. It implies
that we cannot answer question I) without presupposing, or at least implying,
an answer to question II). The two debates on rights are not separable, but
intertwined.
One preliminary comment: Wenar keeps his analysis general in scope, it
is meant to apply to moral, legal and customary rights of conduct. I restrict
my analysis to moral rights. The rights assertions I shall refer to concern
moral rights, normative claims are to be understood as moral judgements and
52
Franziska Poprawe
normative theories as moral theories.4
I. Leif Weinars Concept of Rights
In his article The Nature of Rights, Wenar5 proposes a new conceptual analysis of what rights are, that he believes solves the conflict between the will
and interest theory. His argument is based on the thesis that the concept of
rights consists of two parts: the first parts concerns what kind of things rights
are,6 that is, what we mean by rights assertions. The second part specifies
what rights do for the right-holder, that is, it specifies their function.
With regard to the first part of the concept of rights, Wenar7 argues that
all rights are a combination of one or more of the four Hohfeldian incidents:
privilege, claim, power and immunity. He presents Hohfelds logical structure
of rights in a modified, enriched way: A privilege is a right of the logical
form, A has a right to f and implies that A has no duty not to f and A
has no duty to f . It can either be a single privilege, granting an exemption
from a general duty, or a paired privilege in the sense of a permission, giving
the right holder discretion over some action. Wenar gives the example of a
sheriffs right to break down a door, which is an exemption from the duty
not to break down doors. An example for a paired privilege is my right to
move my body. I can choose to do so or not, it gives a discretion. Claim
rights are rights of the logical form A has the right that B f s, and imply that
B has a duty to A to f . They can have three different functions for their
holders: protection against harm, provision of a need or the per f ormance of
a agreed upon action. Powers are rights to change the claims and privileges
of other persons or oneself and thus gives the holder a certain authority to
4
In the debate on the concept of rights, theorists often use examples that refer to
moral rights. I shall do the same in my examples. I believe the analysis of this
article can be extended or adopted to other classes of rights of conduct, but doing
5
6
7
so would go beyond the scope of this paper.

Wenar, The Nature of Rights.
Wenar, The Nature of Rights, 224.
Wenar, The Nature of Rights, 224-237.
53

change the normative situation. These powers are limited to the extent that
others have immunities: Immunities are rights that, like claims, protect the
right-holder. From the examination of several exemplary rights assertions,
Wenar inductively concludes that all rights are Hohfeldian incidents, whereby
some rights are a combination of Hohfeldian incidents. Indeed, Hohfelds
framework is widely accepted as the internal structure of rights. I agree with
this position. All rights are Hohfeldian incidents.
Besides this consensus, theorists disagree on whether or not all Hohfeldian incidents are rights. Will and interest theorists hold that only those Hohfeldian incidents are rights which perform some specific function, although
they disagree on what that function is. Wenar agrees with the will and interest theorists in that we shall point towards the function of rights to elucidate
which Hohfeldian incidents actually are rights. Yet, Wenar rejects the will
and the interest theory as the correct account of what rights do for the rightholder. He does so in the following way.
First, Wenar rejects the will theory. According to Herbert Harts version
of that theory, the function of a right is to endow the holder with discretion
over the duty of another; it makes the rightholder a small scale sovereign.8
Translated into the Hohfeldian framework, this means that will theorists recognise as rights only paired powers.9 Wenar rejects this theory on the grounds
that it is too narrow and has counter-intuitive implications. Some important
rights, such as the right not to be tortured to death, would not count as rights
under the theory, since the right-holder has no legal power to waive it. He
argues further that the theory incorrectly limits the class of potential rightholders: Children or people otherwise incapable of exercising powers would
be excluded.
Next, Wenar rejects the interest theory, which states that rights have the
single function of promoting the holders interest. Wenar argues that, though
this theory overcomes some weaknesses of the will theory it is also at odds
8
9
Hart, Essays on Bentham: Studies in Jurisprudence and Political Theory, 183.

This interpretation of the Will theory as recognizing only those Hohfeldian incidents which are powers as rights is common.
54
Franziska Poprawe
with our ordinary understanding of rights. Many rights that we want to recognize as such are not in the interest of the holder, e.g. the right of a judge
to sentence a criminal. Therefore, he rejects the interest theory as the correct
account of what rights do for the right-holder.
One alternative to these two theories constitutes the Any-Incident T heory
according to which all Hohfeldian incidents are rights. Wenar rejects it for
being too inclusive: It counts as rights some Hohfeldian incidents that we
ordinarily would not.10 He uses the following example: We have no duty to
lie to people. Therefore, each of us has a single privilege not to lie to people according to the Hohfeldian framework. According to the Any-Incident
Theory, we therefore each have a right not to lie to people. This seems odd.
Wenar argues that the Any- Incident Theory fails to exclude such counterintuitive results because it recognizes as rights incidents that have no function.
He therefore concludes that not all Hohfeldian incidents are rights, and the
correct theory to specify which one are, has to be a theory of the function
of rights. Having rejected both, the will and interest theory as candidates for
such a theory, he proposes the Several-Function T heory (now: SFT).
The SFT according to Wenar states: Rights are all those Hohfeldian incidents that perform one or more of six specific functions: exemption, protection, authority, provision, performance, discretion.11 Wenar points out a number of, as I shall call them, attributes of the SFT that he believes establish its
merits and make it superior to both, the will theory and interest theory. First,
the SFT captures rights recognised by the will theory and rights recognised
by the interest theory. In this sense, it solves the conflict between the will and
the interest theory. Second, the SFT excludes incidents that are intuitively
not rights. For example, the privilege not to lie to people is not a right under
the SFT, because it does not serve a function (it does not exempt you from a
general duty because there is no such duty in to begin with). Third, the SFT
acknowledges that all rights have some function, but does not restrict it to one
single function. Wenar holds that there is no one thing that rights do for the
10 Wenar,
11 Wenar,
The Nature of Rights, 244.

55

right-holder. Rights have no fundamental normative purpose in this sense.12
Fourth, he argues that the SFT best captures our ordinary understanding of
rights. He argues that the standard for settling this dispute [on the concept
of rights] is an ordinary understanding of what rights are and what roles they
play in our lives.13 As opposed to the will and the interest theory, the SFT
can supposedly not be accused of having many counterintuitive implications.
Furthermore, Wenar claims that the SFT does not support a particular
normative theory, and constitutes a conceptual analysis that is detached from
normative arguments. Why is this a good feature of a theory of rights? Wenar
emphasises the difference between stipulating within a moral or legal theory
what the word right means on the one hand, and purely descriptive, conceptual analysis on the other. The latter should be free of the former. He points
out that the will and interest theories fail to do justice to this demarcation
between conceptual analysis and normative theorising.
Will theorists and interest theorists have erred in adopting analyses framed to favour their commitments in normative theory.
This has turned the debate between them into a proxy for the
debate between Kantianism and welfarism.14
In contrast, the SFT does not support any normative position but seeks only
to establish conceptual clarity. The final attribute I would like to mention
is related to the previous ones: Wenar argues that the SFT offers a standard
against which we are to measure the interpretations of rights of diverse normative theories. The will and the interest theory, as theories that are set up to
fit particular normative theories, cannot serve this purpose. He writes:
An analysis that is faithful to an ordinary understanding of rights
will not reduce the justificatory burden on any normative theory
12 Wenar,

14 Wenar, The Nature of Rights 224.
13 Wenar,
56
Franziska Poprawe
of rights. Instead, such an analysis will partly determine the justificatory burden that a normative theory must discharge.15
Concluding this section, I grant that the SFT is superior to the will and interest
theories with regard to the first three of the above aspects. In the remainder of
this paper I challenge primarily the last three of the mentioned merits Wenar
ascribes to the SFT. Does the SFT really provide us with an adequate answer
to the question what are rights? that satisfies our ordinary understanding
of rights, but does not rely on commitments to one or the other normative
theory? I argue that it does.
II. The Several-Function Theory challenged
In this section, I argue that the SFT does not present us with a theory-neutral,
purely descriptive conceptual analysis of what rights are. I argue that we
cannot always determine whether or not a Hohfeldian incident classifies as a
right under the SFT without presupposing or at least implying some normative
claims about that incident. I show that there may be rights which would not be
recognised by the SFT without the reasons for them. I do so by investigating
how we can specify what classifies as a right according to the SFT by applying
the theory to a rights assertion. Let us consider the following example from
the context of enslavement:
(R) Person A has a claim against person B that B does not buy A.
Is R a right under the SFT? Firstly, we have to check whether it is a Hohfeldian
incident. The rights assertion has the form: A has a claim against B that B
f s. To say that a person A has a claim against B that B f s, translated into
Hohfeldian terms implies that B has a duty to A to f . Here, B has a duty to
A not to buy A. Given the immorality of slavery, this seems a sensible thing
to say, so the assertion R is a Hohfeldian claim right.
Further, we can say it is also an immunity: If B bought A, thereby enslaving A, that would change most of As (and some of Bs) claims and rights
15 Wenar,

57

tremendously. Therefore, A has also has an immunity against B that B does
not buy A. In the Hohfeldian language, this implies that B has no power to
buy A. This seems also a sensible thing to say in light of the rights assertion R. Thus, we are dealing with a Hohfeldian claim right and a Hohfeldian
immunity. According to the SFT, this is not yet enough to say that the Hohfeldian instances we identified actually refer to a right. It is not enough to
say that if A has such a claim and immunity, A has a right. Before we may
draw this conclusion, we have to ask: does R have a function? The function
that an immunity and a claim have in common is protection. So we have to
ask: Would the claim and the immunity protect person A, if A was entitled to
them?
Many would probably with no hesitation respond in the affirmative. If a
person A has a claim against another person not to be bought, this would protect person A. Also, if no one has the power to buy anyone and thus everyone
has an immunity against being bought, this would be so to protect everyone
from being bought. Therefore, the Hohfeldian incident expressed in the rights
assertion R has a function, the function of protection. Thus, under the SFT,
the rights assertion R refers to a right. If we say a person has a claim as in R,
we could equally say that that person has a right, lets call it, a right against
enslavement.
Several-Function Theory: R $ Person A has a right against enslavement.
There is more to say here. But before doing so, let me pause the analysis for
a moment and consider the following example: the case of Mike and Sam.
Suppose Sam wants to be Mikes slave. What makes him most happy in
life is to serve Mike. Sam has absolutely no pleasure in choosing himself
what to do, in fact, it makes him miserable. If he could not be Mikes slave,
doing whatever Mike wants him to, he is going to be miserable for the rest
of his life. (Even if Mike treated him from a common sense perspective badly, Sam would still prefer to be a slave, since he is, and has always been,
masochistic. No doubt we can imagine such a case.) Also, Mike would like
58
Franziska Poprawe
be the master of a slave. It would make him truly happy to own Sam. What
ought they to do? As it so happens, Mike and Sam are strong proponents
of utilitarianism. Under a simple reading of utilitarianism, the morally right
action is the one that increases happiness in the world. In the case of Sam
and Mike, the morally right act for Sam would then be to sell himself to
Mike. This brings about the greatest happiness in the world (other things
being equal). Therefore, from a utilitarian point of view, Sam morally ought
to enslave himself. If Sam wants to sell himself to Mike, and Mike wants to
buy Sam, if that is what makes them both happy (all things being equal), then
Sam should do exactly that. In their society, there is no reason not to, and
every moral reason to follow this course of action. But what about the right
against enslavement?
The SFT shall present us with a purely conceptual analysis of rights. It
should tell us what rights are, without making normative claims about these
rights. To put it in the words used in the outset of this paper, the SFT concerns
only the descriptive question I) what classifies as a right, and not question
II), whether or not the right is ascribed to people. That in turn means that,
once we recognize a Hohfeldian incident as a right under the SFT, normative
judgements can only inform our answer to whether or not we ascribe the
identified right to person X in situation Y. Taking a normative stance with
regard to the right should not alter our conception of it as a right.
In the case of Sam and Mike, the previously identified right against enslavement must therefore be at least conceptually conceivable as a right, even
if, in fact, we were to take a normative position denying that it is held by
anyone, including Sam. We arrived at the following claims:
(1) If a person A has a claim against B that B does not buy A, A has a
right against enslavement with the function of protection, according to
the SFT .
(2) The morally right act for Sam is to sell himself into slavery to Mike.
The first claim is the result of the analysis of the rights assertion R according
to the SFT. The second claim arises from the case of Mike and Sam in their
59

utilitarian framework. These claims imply that:
(3) If Sam had a claim against Mike that Mike does not buy him, he
would have a right, called the right against enslavement, which protects
him.
(4) Because Sam morally ought to sell himself to Mike, if Sam had the
right against enslavement, it would protect him from doing what he is
morally required to do.
(3) follows from applying the general case of A and B to Sam and Mike. (4)
is implied by (2) and (3). This claim seems odd. It is not sensible to say that
someone is protected from doing what he morally ought to do. One can be
hindered from doing what one morally ought to do, but that is a different thing
to say. Protection from X means that X can only be filled with something
that, in whatever way and for whatever reason, ought not to be the case, e.g.
harm. But here, it is filled with something that ought to be the case (because
Sam wants to and is morally required to sell himself into slavery to Mike).
Claim (4) is conceptually incoherent. How could we solve this incoherence?
We can either reject claim (1) or claim (2).
To reject claim (2) would be to reject utilitarianism. But we are engaging
in a conceptual analysis of rights and want to refrain from taking a position
on the correctness of a moral theory. The application of the SFT in our conceptual analysis of the rights assertion R should not require us to reject a
normative theory on the whole to retain coherence. What rights are according
to the SFT is no fundamentally normative question (cf. attribute five of the
SFT as listed above). It is important to note that, in the above example, there
is no claim that Sam should be entitled the right R. Though claim (2) is based
on utilitarian grounds and certainly normative, the argument does not assume
that Sam should be ascribed the right not to be enslaved. In fact, because the
conception of R as a right under utilitarianism is incoherent, to say that Sam
should be ascribed the right not to be enslaved would not be sensible.
Thus, claim (1) has to be false: The above analysis has yielded this result:
the incident of the assertion R is a right under the SFT. Rejecting (1) would
60
Franziska Poprawe
mean that something has gone wrong in the analysis of the rights assertion
using the SFT. And that is indeed so: After identifying that we are dealing
with a Hohfeldian incident, I stipulated the answer to the question concerning the function of the right, jumping to the conclusion that if person A had
the claim that B does not buy A, then A would have a right that protects A.
That seemed very intuitive and little objectionable and I do not doubt its correctness. However, pursuing a purely conceptual analysis of rights here, we
may not argue this way. In particular, we have to be careful not to presuppose
normative judgements that, if altered, would change the conceptual analysis
of the rights assertion by altering the answer to whether or not the Hohfeldian incident has a function. Bearing this in mind, let us ask again: Would
the claim and the immunity protect person A, if ascribed to A? In the case of
Sam and Mike, we ask: Would the claim and the immunity protect Sam, if
Sam was entitled to them?
In the case as it is described above, the answer would certainly be no.
Ascribing Sam a claim against Mike that the latter does not buy him could
not be understood as protecting Sam, it would rather hinder him to do what
he desires and what is morally required, that is, to sell himself into slavery
to Mike. Therefore, here, the identified Hohfeldian incident (R) has no function,16 and is therefore, according to the SFT, not a right. Again, this result
does not rely on the claim that under Utilitarianism, Sam is not ascribed the
16 One
might object that under utilitarianism, the function of the incident R is not
protection, but something else, and that the SFT does recognize it as a right. Fortunately for my argument, this route is not successful. The SFT attributes to claims
three possible functions: it entitles the right-holder to protection, provision of a need
and/or the specific performance of some agreed-upon, compensatory or legally or
conventionally specified action (Wenar 2005, p. 229). In the case of Mike and
Sam, the claim that Mike does not buy Sam does not provide Sam with some need.
Quite the opposite, it hinders him from doing what he needs! Also, if R was a right
and Sam was entitled to it, Sam would not be entitled to some performance of the
kind Wenar discusses. If any action was agreed upon or conventionally specified, in
the case at hand, it would be the contrary action (Mike buys Sam). Therefore, the
argument withstands this objection.
61

right against enslavement, though it is arguable consistent with it. In the utilitarian context of Sam and Mike, the Hohfeldian incident R is conceptually
not a right, which means it cannot be ascribed to anyone.17
One might think that the initial stipulation that the right against enslavement is a right that protects the right-holder had great intuitive appeal. I agree
that the stipulation was not all wrong. Indeed, we could easily modify the
case of Mike and Sam so that we account for this intuition. Let us assume
now another version of the case of Mike and Sam, in which, instead of being
proponents of utilitarianism, let us now assume Sam and Mike are Kantians.
Kantianism, among other theories, ascribes inherent moral value to a persons
dignity, autonomy and freedom. Selling oneself into slavery would mean to
disregard ones dignity. From this Kantian perspective, Sam ought not to sell
himself to Mike. The enslavement of Sam would infringe Kants Humanity
principle,18 at least. Because Sam would treat himself not as an end in itself,
but a means to do what he desires. There is no need to elaborate further to
conclude that, from a Kantian perspective it is therefore morally impermissible (even impossible) for Sam to sell himself into slavery (Kant, 1870).19
Sam ought morally not to sell himself, and Mike ought not to buy Sam. Lets
ask one last time: Would the claim and the immunity protect Sam, if Sam
was entitled to them? Now, the answer seem inevitably yes. If Sam had the
claim against Mike that the latter does not buy, this would protect him. Would
Mike buy Sam, this would be morally wrong because it would infringe Sams
autonomy (amongst others). So in this modified case, the Hohfeldian incident
R we are dealing with has a function, the function of protection at least. Thus
we may conclude that it is conceptually a right under the SFT.
17 The
argument could also be run assuming left-libertarianism instead of utilitarian-
ism.
18 Rawls, Kant. Rawls quotes the Metaphysic of Morals, trans. M.Gregor as The
Doctrine of Virtue (New York: Harper and Row.), indicated my Mds.
fact, for Kant Sam even cannot sell himself to Mike, because in doing so, he
19 In
would lose his autonomy and thus his status a moral agent who can be bound by
contracts. Thus, by selling himself to Mike, he simultaneously cannot be bound by
that contract any more.
62
Franziska Poprawe
Summarising the last paragraphs, the analysis of the rights assertion R
under the SFT has exhibited the following: In the first version of the case
of Mike and Sam which assumes utilitarianism, R is not a right, but merely
a Hohfeldian incident without a function. In the second version of the case,
assuming Kantianism, the incident can be assigned the function of protection
and thus classifies as a right under the SFT.20
This means that whether a Hohfeldian incident classifies conceptually as a
right according to the SFT is contingent on normative presuppositions. These
disturb the conceptual analyses. In particular, they induce whether or not a
Hohfeldian incident has a function or not. We have to be very careful that we
do not implicitly assume commitments to normative theories, without which
the analysis would yield a different result. In the beginning of the section, I
stipulated that it is a sensible thing to say that the claim against enslavement
would protect the right- holder. This stipulation leads to the odd result of
the conceptual analysis, demonstrating the necessity of maintaining diligence
with respect to the clarification of ones normative presuppositions. Only by
setting the analysis of the rights assertion R into the contexts of the case of
Mike and Sam in its two versions, I made general underlying normative presuppositions explicit. This approach exposed how our normative theories bear
upon the conception of rights assertions. They determine whether and explain
why we ascribe (or do not ascribe) a function to a Hohfeldian incident. They
therefore also determine what rights are according to the SFT. Kramer and
Steiner (2007) have accused Wenars SFT of being inaccurate in the assignment of functions to rights. I add that the reason for this lack of determination
of the SFT consists of the contingency of it on our fundamental judgmental
assumptions running in the background of the analysis. Thereby, I merely
20 For
the reader stumbling at the idea of rights within utilitarianism, a brief clar-
ificatory note: One might object that taking a utilitarian standpoint to approach
rights obviously leads to trouble, because utilitarianism does not, per se, account
for rights. Yet this is not the point here. According to Wenar (and others), what
rights are is a conceptual question unrelated to the question whether or not anyone
should be entitled to them (what rights there are).
63

say that the assignment of functions bears some normative charge, without
limiting it to a particular position.
I conclude that the SFT cannot lead to a concept of rights that is fully
detached from our normative commitments. It cannot be a theory-neutral
standard of what rights are, because it does not allow us to explain why an
identified Hohfeldian incident is a right or not without making or implying
normative claims. Were we to explicitly refrain from any normative position
and ask does the rights assertion XY refer to a right, the SFT may happen
to leave us with the unsatisfactory answer it depends. I have shown this
without making any statements about the entitlement of specific rights. It
may be our general normative basis that determines whether or not something
classifies as a right according to the SFT. The result of my analysis does not
rely on whether or not we hold that the particular putative right in question
should be held by anyone.
III. On the One Concept of Rights

In this final section, I will elaborate on implications of the conclusion drawn
in the previous section. I will also explore the possibility of generalising the
thesis I established there to the wider thesis the concept of rights is necessarily
bound to normative theorising.
In the outset of this article I mentioned the division of the philosophical
discourse on rights into two main terrains: one is on the concept of rights
(I), and the other on the normative question of who is entitled to what rights
(II). I started my analysis of the SFT from the same starting point that Wenar
takes when establishing the SFT, focusing on question I) and disregarding
question II). However, the result I obtained from my analysis overrides the
demarcation between the two questions. Consider the following: Question
II) asks for a normative answer. Under the plausible assumption that we aim
for consistency among our normative claims, this normative answer, whatever
it may be, should be directly implied by or at least derivable from our general
commitments in normative theory. According to the analysis just given, the
64
Franziska Poprawe
SFT relies on our general normative presuppositions in the background of our
reasoning. Thus, it follows that, if we accept the SFT as the concept of rights,
we cannot answer question I) without presupposing or at least implying an
answer to question II).
This result might strike some as odd. After all, the division of the discourse on rights has a reason. In light of the significance and controversy of
the debates about what rights people (should) have, a conceptual analysis of
what rights are seems essential to avoid that the normative debate collapses
in misunderstandings. In other words, we want to establish conceptual clarity
of the subject that we talk about, rights, before we talk about it, that is, before engage in a normative debate on who should have which rights. Driven
by this laudable aspiration, theorists concluded that what sets the stage for
the normative discourse on what rights people have is an understanding of
what rights conceptually are. The debate on the nature of rights (question I)
emerged. Within it, the contest of establishing the concept of rights is based
on the assumption that there is a concept of rights that is unrelated to the normative question of what rights people (should) have (question II). The long
debate between the many rights theorists assumes monism relative to the concept of rights and its units, i.e. rights.21 In the same way, Wenars endeavour
of reconciling the conflict between the interest theory and will theory by constructing a new theory of the nature of rights that captures the advantages of
both, namely the SFT, is based on the assumption that we can truthfully contribute oneness to the concept of rights22 and that there is a non-normative
notion of rights. I acknowledge that, for any theorist, the notion of a nonnormative concept of rights has great intuitive appeal. Again, we want to
establish conceptual precision of any subject of interest, before we make normative claims about it. But to take this intuition as sufficient for believing that
such a purely conceptual account is possible on the subject of rights would
21 I
am borrowing here from Van Duffer (2012), who points out that the debate be-
tween the interest and will theories is based on the assumption that it is possible to
provide a monist analysis of the direction of duties. I generalise this thesis here.
Schaffer (2014).
22 Cf.
65

be wishful thinking. It is also not sufficient to deem the result of my analysis
of the SFT implausible. Instead, the result of my analysis should be seen as
a reason to question the assumption that there is a non-normative concept of
what rights are. Now the question arises: How plausible is the assumption
that there is one non-normative concept of rights? Taking into consideration
the consensus that all rights are Hohfeldian incidents, but that not all Hohfeldian incidents are rights, we can ask: Is it possible to formulate a theory that
(i) determines which Hohfeldian incidents constitute a right, but that (ii) does
not rest on normative grounds?
Such a theory should also exclude rights which we would intuitively not
consider rights, solve the conflict between the will and the interest theorists at
least (in some sense) and should not narrow the normative purpose of rights.
As noted by Wenar, the SFT theory fulfils these requirements we demand of
the concept of rights. I acknowledged this in section 1. I can add now that
the SFT fulfils these demands of the concept of rights because it does not
directly en f orce any particular normative position. It thereby overcomes one
weakness of the will and the interest theory. And yet, my analysis has shown
that the SFT requires some normative position to determine which Hohfeldian incidents are rights. These considerations raise the doubt that there is a
theory of what rights are which does not rely on normative theorizing, and
has the merits of the SFT. If all analyses of rights rest on general normative
dispositions, either in the way the SFT does or in the way the will and interest theories do, then the concept of rights must be inherently normatively
charged, and an explanation of what rights are is necessarily dependent on
normative claims.
I believe this result should be taken seriously. That is, I believe the conclusion from the analysis of the SFT is generalizable to the following thesis:
whether or not a Hohfeldian incident is a right depends on the normative load
that we carry when conducting the analysis. A full concept of rights is necessarily bound to normative theorising. This is to say that we have to reject
the assumption that there is one concept of rights, and to acknowledge that
a pure, non-normative conceptual analysis of rights stops at the level of the
66
Franziska Poprawe
consensus that all rights are Hohfeldian incidents. The Hohfeldian framework as the logical structure of rights remains unrefuted. It should be used to
explain what we mean by rights that are otherwise justified. But beyond that,
we cannot explain what rights are, without at least implicitly relying on the
reasons for them. A full proof of this generalised, speculative thesis would
go beyond the scope of this essay. It is worth noting, however, that Kramer
and Steiner23 reach a conclusion which I take to be consistent with my speculation: they advise rights theorists of developing further the will and interest
theories, while denying that there can be a third way.
Whether we stick to the original conclusion of section 2, namely that
what rights are according to the SFT depends on normative commitments,
or whether we accepts its generalisation, namely that the concept of rights
is necessarily bound to normative theorising, the analysis of this paper suggests the following insights which, if proven true, would require us to alter
significantly the way we approach the debates on rights.
First, the result suggests that the standard to judge any concept of rights is
not an ordinary understanding of what rights are. Wenar seems to derive this
from the fact that the final accusation of will and interest theorists against
each other has always been that the other sides theory yields counterintuitive
results.24 However, if the concept of rights is subordinated to normative theories in the sense that what rights are relies on normative commitments, then,
necessarily, the standard by which the will and interest theories (and any other
theory of the concept of rights) judge each other cannot be an ordinary understanding of rights. Kramer and Steiner have also objected to Wenars extremely heavily reliance on ordinary usage25 on the grounds that there is not
one unique ordinary understanding of rights,26 but rather a number of conflicting understandings of what rights are. I agree with this and believe this
is so precisely because whatever our ordinary understandings of rights are,
23 Kramer
and Steiner, Theories of Rights: Is there a Third Way?

The Nature or Rights, 250.
25 Kramer and Steiner, Theories of Rights: Is there a Third Way?, 295.
26 Wenar, The Nature of Rights.
24 Wenar,
67

they are inflicted with potentially conflicting normative commitments. Therefore, the standard to judge a concept of rights against is not congruence with
ordinary usage, but coherence of the concept with ones respective normative
theory.
Second, the analysis suggests that, contrary to Wenar opinion, the concept
of rights does not determine the justificatory burden that a normative theory
must discharge (Wenar 2005, p. 251). The SFT, though maybe on first sight
roughly in line with what many believe rights are, relies on the normative
charge that we carry when investigating which rights assertions actually refer to rights. The SFT only has analytic power with respect to a particular
normative system. It puts the justificatory burden of what rights are on the
normative theories themselves. Therefore, contrary to Wenars assertion, the
SFT cannot serve as a standard against which we are to measure the interpretations of rights of diverse normative theories.27 Put bluntly, the last two
paragraphs say: The standard for the concept of rights is not ordinary understanding, but the moral theories themselves and the concept of rights is not
the ground for the conflict between normative theories.
In light of these considerations, it becomes apparent that the starting point
for the normative debate of what rights people should have is not a nonnormative the concept of rights, but due diligence in making underlying normative claims explicit. Philosophical usage of the term rights needs to be
refined and the reasoning behind its ascribed meaning revealed. If the concept of rights is secondary to normative theorising (even if not immediately
and always obviously so), this approach allows establishing as much conceptual clarity as possible in the normative discussion on what rights people
ought to have.
Conclusion
Wenar proposes the Several Function Theory as a theory of the function of
rights in order to solve the conflict between the will and the interest theory.
27 Wenar,
The Nature of Rights, 250f.

68
Franziska Poprawe
In this article, I argue that Wenars proposal does not provide us with the analytic framework it purports to be. Though the Several Function Theory does
not directly support any particular normative position, and thereby overcomes
serious conceptual shortcomings of both, the will and the interest theory, it
requires some normative position to explain what rights are. In particular, I
have shown that whether or not a Hohfeldian incident classifies conceptually
as a right according to this theory is contingent on our fundamental normative presuppositions running in the background of the analysis. The Several
Function Theory, therefore, cannot lead to a purely descriptive non-normative
concept of rights.
I explored the possibility of generalising this result to the thesis that the
concept of rights is necessarily bound to normative commitments in one way
or the other. My speculation reads: what we mean my rights assertions can
be analysed using the purely descriptive Hohfeldian framework. But what we
believe rights are, i.e. a complete concept of rights, cannot be fully detached
from normative theorising.
69

References
[1] Hart, H. Essays on Bentham: Studies in Jurisprudence and Political Theory. Oxford: Clarendon Press, 1982.
[2] Kramer, M.H. and Steiner, H. Theories of rights is there a third way?
Oxford Journal of Legal Studies 2007; 27(2): 281-310.
[3] Rawls, J. Kant. Lecture on the History of Moral Philosophy|, ed. Barbara Hermann. Cambridge, Massachusetts: Harvard University Press,
2007: 143-322.
[4] Van Duffel, S. The Nature of rights debate rests on a mistake. Pacific
Philosophical Quarterly 2012; 93 (1):102-123.
[5] Wenar, L. Rights. Standord Encyclopedia of Philosophy, 2011.
[6] Wenar, L. The Nature of Rights. Philosophy and Public Affair 2005;
33(3): 223-252.
Franziska Poprawe is a MA student of Philosophy and Economics at the

University of Bayreuth, Germany. Her main fields of interest are on rights and
normativity. She hopes to continue research in the future. You can contact
her at [franziska.poprawe@gmail.com].
70
Rerum Causae, 7 (2015), 72-81
How Ought We Treat Family in Political

Liberalism?
Implications of the Separation of
Parent-Child Relationship and Marriage
Julia-Maria Franz
Abstract
This essay intends to place family into Political Liberalism according to the debate among Okin and Rawls. I suggest the reconsideration of family as consisting of two distinct aspects adult relationships
as private associations and reproductive relationships between children
and their parents as an institution of the basic structure. This enables to
address concerns about effects of a gender biased society without implementing comprehensive doctrines in the political sphere. I criticise
common justifications of state supported forms of adult partnerships.
The replacement of reproductive relationships within the basic structure
suggests a priority of basic rights of future citizens over parental rights.
I want to thank Alex Voorhoeve for his helpful comments. Im grateful to JanWillem van der Rijt, Lukas Beck and Aline Biedermann for their thoughts on earlier
versions of this essay.
72
Julia-Maria Franz
I. Rereading the Debate Between Okin and Rawls
John Rawls Political Liberalism is motivated by the idea of developing principles of justice within a society of equal and free individuals. He identifies
the basic structure, the main social and political institutions and their distribution of social goods, as the scope within the principles of justice are to be
applied. By distinguishing between political and private spheres of life, the
individual is perceived as a private person enabled to enter and exit associations on the one hand and as a political citizen on the other side.1
Susan Okins writing addresses the question of whether the Rawlsian
framework sets women and their children in the position of free and equal
citizens and if Political Liberalism can provide a basis for overcoming contemporary effects of gendered oppression.2 By identifying the family as the
linchpin3 of persisting gender inequalities, she challenges the sharp demarcation between the public and the private in Rawls work. She advocates for a
direct application of the principles of justice to familial relationships.4 Four
aspects of the mechanisms of the gendered family can be condensed in order
to illustrate her argument: The denial of equal rights by legal unification of
spouses (1) and the gendered division of labor and its anticipation (2) refer to the first part of Okins critique: Political Liberalism fails to consider
women as free and equal citizens. The Political Virtues argument (3) and the
Incongruence argument (4) pose a threat to the consistency of liberalism by
questioning the systems stability.5 The latter criticises the ambivalence of
imposing great importance on the human psychological development in order
for society to be stable on the one hand,6 but putting almost no emphasis on
1
2
3
4
5
6
Rawls, Political Liberalism 48-54.

Okin, Reason and Feeling in Thinking about Justice 229.
Okin, Justice, Gender and the Family 138.
Okin, Forty Acres and a Mule for Women: Rawls and Feminism 242-244.
deWijze, The Family and Political Justice The Case for Political Liberalisms
260-264.
Rawls, Political Liberalism 81-85, 140-144.
73
How Ought We Treat Family in Political Liberalism?

the circumstances of their development on the other.7 The Political Virtues
argument focuses on two particular requirements to be an autonomous citizen. She needs self-worth in order to promote her interest in the public and
must be able to accept the burdens of judgment in public discourse.8 Women
in particular will lack a conception of the self as a virtuous person in the public deliberation when growing up under sexist comprehensive doctrines in the
private sphere.9
Rawls argues in response to Okin that the private public distinction leads
to the wrong line of demarcation between the institutions of society. Family must be integrated into the basic structure by its reproductive function.10
Nevertheless he differentiates between two kinds of institutions within the basic structure: political in contrast to private associational. The first refers to
institutions in which we can only enter by birth and exit by death and hence,
demands the principles of justice to be applied directly into its internal structure. The second group sees the principles of justice applied on its structure.
Rawls identifies family as belonging to the latter because it is a voluntarily
chosen group membership for the sake of living reasonable comprehensive
doctrines. Thereby, he argues that one would not want the principles [to]
apply directly into the internal life of our associations.11
In this essay I argue that both, Okin and Rawls, fail to place the conception of family properly within the framework of Political Liberalism. First,
I propose a separation of the entity family into two distinct aspects: adult
relationship (e.g. marriage) and reproductive aspects in the parental-child relationship. Conversely, this distinction will lead to the placement of adult
relationships in the associational-private realm. On the other hand, the analysis of the parent child relationship outlines its influence on the basic structure.
The replacement of the reproductive aspect of family life will result in suggesting the adaptation of a child-centred approach to parental rights in order
7
Okin, Political Liberalism, Justice, and Gender 27-30.

Rawls, Political Liberalism 54- 58.
9 Okin, Political Liberalism, Justice, and Gender 37-40.
10 Rawls, Justice as Fairness: A Restatement 162-168.
11 Rawls, Justice as Fairness: A Restatement 165.
8
74
Julia-Maria Franz
to prioritise parents obligations over their interest.
II. The Legal Separation Between Adult and Parent Child Relationships
Both Okin and Rawls put emphasis on describing the effects of the gendered
family on the society but both neither explicitly define the core aspect of the
term family nor justify its standing as a distinct institution.12 Instead, they
vary in using marriage, shared household and parenting synonymously. In order to place family within the framework of the basic structure, Okin uses all
three of the characteristics in her argument. Rawls focuses only on reproductive aspects for justification. Hence, the two of them implicitly use different
premises: Rawls argues that the core characteristic of family consists of reproduction and Okin subsumes both, adult relationships, such as marriage in
shared households, and reproduction as being necessary components in talking about family.
Both are wrong in approaching the conception of family in this way. It is
very questionable to use a single conception for the term family at all. In
a liberal society, one would expect citizens to deliberate on a definition that
would be used for policy and law making. The mere empirical observation
of children being born out of or within marriage and what counts as an adult
relationship might be sufficient for an objection here and leads us to analyse
the standing of relational and procreative aspects of family separately.
III. Adult Relationship as Private Association
Considering family as marriage in adult relationships will imply its placement
in the associational realm of the society. By following this understanding
of family, one might wonder why the state should provide for and protect
this institution at all; and secondly, why feminist critique on Rawls should
not take into consideration other forms of commitment other than marriage.
These questions shall be discussed in the following section.
12 Metz,
Untying the Knot 73- 83.

75

Marriage contains two rather uncontroversial defining components; it is
an agreement among adults who voluntarily choose to enter and exit the institution. Both parts of the definition are essential for liberalism to respect
the parties as equal and free citizens. Like other forms of associations, the
content of the agreement plays a subordinated role, as long as basic rights are
not violated and the right to exit is not endangered. All further narrower definitions will end up in illiberal lines of argumentation. It would be possible
to take the monogamous heterosexual relationship as being the ideal standard
of a marriage and to use it as an optimal account for accomplishing human
flourishing and care provision within this private association. This would require an overlapping consensus in the deliberation among citizens, but the
case of marriage is not and has not been perspicuous and is a centre of debate whether it supports all individual comprehensive doctrines in the society
or not. Like Brake points out, the core of marriage has always been an issue of philosophical debate in the past.13 Today, care-networks, polyamorous
relationships and asexuals may legitimately ask why the state provides recognition for dyadic (homo- and heterosexual) relationships but not for others.14
Additional, an overlapping consensus requires a possibility that the state has
any relevance in marriage, monogamy, same-sex-marriage or any approach
to offer the possibility of right and asset share in a single institution based on
comprehensive doctrines of the good life. To discuss whether these problems
will lead to the abolishment of legal recognition of adult relationships or to a
limitation of its conceptions, is not of major importance for this contention.15
I conclude that Rawls is right in limiting marriage and adult relationships
in the same way as churches, companies or private universities as associations, although it is a topic of debate whether the state shall prioritise certain
forms of adult relationship over others at all. This account of an associational character of marriage and civil union neither justifies actions that limit
the possibility to exit, nor measures that do not consider the partners legal
13 Brake,
Minimizing Marriage 14-20.

65-108.
15 Chambers, The Family as a Basic Institution 134-143.
14 Ibid.,
76
Julia-Maria Franz
independencies from each other. Thereby, the first two principles of Okins
critique, according to the inequality in rights and labor, are considered and
confuted. Nevertheless, this means that partners legally may have full status
of equal citizenship but believe in a comprehensive doctrine that promotes
an unequal status among them. Okin criticises that the political virtue and
the incongruence argument may bring women in a position to live in such
ambivalent conceptions of themselves.16 They may end up assigning their
disregarded status in the partnership to the public realm and will lack selfrespect in articulating their own conception of the good.17 I reply that these
problems primarily arise out of socialisation during childhood within the reproductive aspect of family and should be solved in that sphere. I take political virtues, once developed in childhood, as constituting a degree of stability
in the personal character and are not endangered to be lost through marriage.
In addition, although the indirect pressure in the private may play a significant role in reality, a tool to measure it is required. Defining a threshold
between unjust pressure and just influence on women initiated a necessary
debate on the concept of adaptive preferences that lies beyond the coverage
of this essay.18
IV. The Child Centred Approach: Implications of Placing Parenthood in
the Public Realm
The consideration of only reproduction and parenting in family separately
leads us to push family back into the realm of the political institutions of the
basic structure. The qualifying criteria for entry and exit to and from the
basic structure are birth and death. Hence, the requirement of arbitrariness,
which finds recognition in the ideal theory of the veil of ignorance, plays a
constitutive role at this point. From the childs perspective, the birth in any
particular family is as arbitrary as other factors. Furthermore, it is the con16 Okin,
Reason and Feeling in Thinking about Justice 173.

Justice as Fairness: A Restatement 59. Rawls, Political Liberalism 77.
18 For a discussion of the role of adaptive preferences on autonomy in the context of
17 Rawls,
marriage review
77

stitution for many other factors like genetic outcomes, region of birth and
social status during childhood. Rawls argument of establishing all aspects of
the family in the associational sphere is rightly criticised by Chambers as a
red-herring.19 The treatment of parent child relationships cannot be justified
by the parental interest in living a life according to their comprehensive doctrines, even if this will further their human flourishing. Rawls outlines earlier:
the basic structure is not an association because it is structurally distinct in its
entry by birth and by its missing ends of comprehensive doctrines.20 Consequently, associational and political parts of the family have to be regarded as
distinct when applying legal and policy questions.
In addition to his distorted setting of the familial institutions within the
basic structure, Rawls account faces a second criticism on the constitution
of the parent-child-relationship. The ultimate prioritising of adults interests
over the childs in Rawls? work conflicts with his conception of basic rights
as equal citizens21 during childhood. These universal rights contain aspects
of welfare and integrity of the person and cannot be limited by adults interest. Rawls ambivalent conceptions in the moral status of the person during
childhood may be called parent-centred perspective. It perceives children as
parental property since Rawls priority of the adult privacy is not justified by
the effects it poses on the childs development. Once agreed that the two
domains of the family are not one entity, the conflict of autonomy between
the adults interest in their living of a comprehensive doctrine and the childs
interest of moral development diminishes. In case of conflict of the political
basic structure and private comprehensive doctrines the first sets constraints
on the latter. Hence, the paradox of restraining the parents autonomy by intervening in the private sphere in order to prevent the future autonomy of the
child is indeed not a paradox, because parents never possessed the right to
impose their comprehensive doctrine on the child if it endangered its moral
development in the first instance.
19 Chambers,
The Family as a Basic Institution 76.

Political Liberalism 40.
21 Rawls, Justice as fairness: A restatement 166.
20 Rawls,
78
Julia-Maria Franz
Nevertheless, one may identify the particular problem that children cannot
yet possess and sue for their own rights based on missing capacities. If this
would be true, it does not follow that the child does not have rights. It is
the state in its guardianship position of the childs moral development that
allocates the entitlement for parental rights to the adult. The reproductive
aspect of the familys standing in the basic structure may be redefined as
a transfer of rights by the state that is in guardianship of the childs moral
development.
The evaluation of the implications on adult partnerships proved, that Rawls
policy recommendation goes along with what follows out of the separation of
the two spheres of family life. This is different for the case of the parent child
relationship. The political character of the childs development is defined by
the societys interest in the childs development. Rawls reduces the scope
of application of the interest only to the function as freedom of abuse and
equality of opportunity (e.g. equal basic rights for primary education for girls
and boys in their welfare rights).22 However, his conception of the human
moral development requests more than that. In order for one to acquire the
features of being reasonable and having moral sensibility, Rawls introduces
four conditions: willingness to cooperate in the society, the recognition of
the burdens of judgment (acknowledgment of different reasonable comprehensive doctrines than ones own), self-respect, and a moral psychology with
capacities for rational argument.23 In order to develop all four of these conditions a society finds an overlapping consensus on the question of how this
may be achieved. A discussion of these means seems not to be necessary
in detail here, but what can be assumed to find consensus, is that moral development requires a learning process that relies on the parental capacity to
teach. Especially the first two features are directed to society and require a
learning process independent of the family. Learning to take part in the public discourse includes having learned the rules in a space where one meets
others with different conceptions of the good. This is what I call the positive
22 Ibid.,
165.
Political Liberalism 81.
23 Rawls,
79

requirement of the public space for the child to learn. The negative requirement, which citizens will agree on, is the child being in need of a place where
the development of self-respect is supported. Surely, not every child will be
raised in conditions that pose a threat to the development of this feature, but
with the public sphere, we might introduce an instance of control that will
lessen concerns about the moral problem, called adaptive preferences, that is
connected to Okins concern about the effect on the child living under sexist
comprehensive doctrines. What practical implications can my argument pose
on adults sphere of authority? One may advocate for the abolishment of private or ecclesiastic schools, as the common education of religious thought in
these institutions and under certain empirical restraints the duty to use public
daycare.
Conclusion
This essay provides a concept for the integration of the family within Political Liberalism. It addresses feminist concerns about the threats of the genderbiased society and develops an argument to consider adult- relationships distinct from the parent-child framework. The legal and moral separation of
the two aspects leads to adaptations of both Rawls and Okins recommendations. I argue that the scope of rights and acknowledgment of particular adult
agreements like marriage and civil unions must be infringed. Furthermore,
the integration of the reproductive aspect of the family speaks in favour of a
child-centred approach to their rights.
80
Julia-Maria Franz
References
[1] Brake, E. Minimizing Marriage. Oxford: Oxford University Press,
2012.
[2] Chambers, C. The Family as a Basic Institution, in: Abbey Ruth (ed.)
Feminist Interpretations of John Rawls, 1st ed. Harrisburg: Penn State
University Press, 2013. 75-95.
[3] Khader, S. Adaptive Preferences and Womens Empowerment. Oxford:
Oxford University Press, 2011.
[4] Metz, T. Untying the Knot. Princeton: Princeton University Press, 2010.
[5] Okin, S. Reason and Feeling in Thinking about Justice, ETHICS 1989,
99(2), 229-249.
[6] Okin, S. Justice, Gender, and the Family. New York: Basic Books, 1989.
[7] Okin, S. Political Liberalism, Justice, and Gender, ETHICS 1994,
105(1), 23-43.
[8] Okin, Susan (2005). Forty Acres and a Mule for Women: Rawls and
Feminism, Politics, Philosophy & Economics 2005, 4(2), 233-248.
[9] Rawls, J. Justice as Fairness: A Restatement. Boston: Harvard University Press, 2001.
[10] Rawls, J. Political Liberalism. New York: Columbia University Press,
2005.
[11] deWijze, S. The Family and Political Justice The Case for Political
Liberalisms, The Journal Of Ethics 2000; 4(3): 257-282.
Julia-Maria Franz is a BA student in Philosophy and Economics at the University of Bayreuth (2012-2015). Her main areas of interest are in political
philosophy and philosophy of action. You can contact her at [jm.franz@posteo.de].
81
Rerum Causae, 7 (2015), 82-104
The Problem of Old Evidence in Bayesian

Confirmation Theory:
A Solution Based on the Difference
Between Update and Justification
Chlo de Canson*
Abstract
In this paper, I discuss the problem of old evidence in Bayesian Confirmation Theory and attempt to build on a suggestion by Lange (1999)
to provide a satisfactory solution. I present the problem, argue that it
cannot be solved by the standard appeal to counterfactuals, and show
that a modification of the Bayesian framework along the lines of what
Lange proposes is necessary. Then, within that new framework, I show
how the problem of old evidence can be solved.
Introduction
The confirmation of scientific hypotheses is a central topic in philosophy of
science. The main idea behind the different accounts of confirmation of scientific hypotheses is that pieces of evidence can lend support to theories, they
*
For helpful discussions on the topic and comments on this paper, I am greatful to
John Worrall, Olivier Roy, Luc Bovens and Alexandru Marcoci.
82
Chlo de Canson
Figure 1: Eddingtons Test

can confirm them. This is usually understood as meaning that, in light of
some evidence, hypotheses can become more plausible. Let us illustrate this
with an example. Let a hypothesis H be all ravens are black, and let the evidence E be the observation of a raven that is indeed black. Intuitively, coming
to know that the raven is black makes the hypothesis more plausible, and
consequently an agents willingness to accept the hypothesis might increase.
This intuition holds for theories considerably more complicated than H. For
example, Einsteins General Relativity Theory (GRT), published in 1915, entails that light bends around the sun. This in turn entails that stars must appear
further apart during the day than they do during the night to an observer on
earth (Figure 1). Arthur Eddington observed in 1919 that this was indeed the
case. This was taken to be a major argument in favour of GRT. Philosophers
of science, and, more importantly, scientists themselves, would say that this
evidence confirms GRT.
One influential systematic account of theory confirmation is Bayesian
Confirmation Theory (BCT). One of the charges against this account is that
it has the clearly unacceptable consequence that already known evidence
can never confirm a newly proposed theory. Bayesianism and the alleged
problem of old evidence are the focus of this paper. In Section I, I outline
Bayesian confirmation theory, and I show how the problem of old evidence
arises within it. In Section II, I go on to examine two versions of a solution
83
The Problem of Old Evidence in Bayesian Confirmation Theory

that has been proposed to the problem of old evidence, the counterfactual
solution, and show that both are unsatisfactory. In Section III, I draw on a
proposal by Marc Lange to put forward a different, although related, solution
to the problem. To do so, I begin by arguing that the traditional Bayesian
framework needs to be modified in order for it to better fit scientific practice,
as proposed by Lange. I then show that, in this modified framework, with the
addition of a proviso, the problem of old evidence can be solved.
I. The Problem of Old Evidence in Bayesian Confirmation Theory
The Bayesian approach to confirmation relies on the probability calculus, and
particularly on a central theorem proven by Bayes that is a direct consequence
of the probability axioms:
p(A|B) =
p(B|A).p(A)
, given p(B) 6= 0.
p(B)
Bayesian epistemology aims at giving an account of how a rational agent

ought to update her system of beliefs in the light of new evidence. The standard account is called Bayesian conditioning. According to that model, an
epistemic agent starts off with a prior probability distribution p1 , her credences, over a given set of propositions. The requirement that her degrees
of belief satisfy the probability calculus embodies the Bayesian constraint of
coherence. She then updates that prior distribution by conditioning on the
propositions that she learns to obtain a posterior probability distribution p2 .
The conditioning consists of using the rules of probability, including notably
Bayes theorem, in order to incorporate the proposition that is learnt into the
set of background knowledge in such a way that the agent remains coherent. Formally speaking: a probability function p2 results from a probability
function p1 by conditioning on a proposition e iff for all propositions a,
p2 (a) = p1 (a|e)
Effectively, at time t2 , the agent has the proposition e within her background knowledge. If an agent begins her epistemic life with a coherent
probability distribution over her beliefs, and then updates by Bayesian con84
Chlo de Canson
ditionalisation, her resulting credence distribution will also satisfy the coherence requirement. Reiteration of the conditionalising process may then occur.
Bayesian confirmation theory uses this epistemological framework to provide a criterion of confirmation. Let p1 (H|E) be called the posterior probability of H, p1 (H) its prior probability, and p1 (E) the prior probability of E.
The Bayesian criteria of confirmation are the following: at some time ti , for
any piece of evidence E and any hypothesis or theory H,
E confirms H if and only if pi (H|E) > pi (H).
E disconfirms H if and only if pi (H|E) < pi (H).
E is neutral towards H if and only if pi (H|E) = pi (H).
Intuitively, this is appealing, because, since, by definition, pi (H|E) =
pi+1 (H), the above account implies that, if some piece of evidence E confirms
H, then pi+1 (H) > pi (H) the agents credence in H increases as she learns
E. Furthermore, given that, as stated above, if a prior distribution satisfies the
coherence requirement, then so does the posterior distribution obtained by
Bayesian conditioning, we can introduce new evidence after having already
updated, and this opens the possibility of incremental confirmation. For the
sake of clarity, the subscripts related to probability functions will be omitted
in the rest of the discussion if all the occurrences of p in a formula refer to
the same function.
To illustrate how BCT works, let us show how it handles the historical
example given in the introduction. Our hypothesis H is GRT, and our evidence
E is the result of Eddingtons test. We want to know whether E confirms H.
As explained in the introduction, H entails E1 and therefore the likelihood of
E, p(H|E), equals 1. Let us assume that, given our background knowledge,
p(E) = 0.1 and p(H) = 0.05.2 Thus, according to Bayes theorem,
1
Of course, H does not entail E on its own, it does so in conjunction with a large set
of auxiliary hypotheses, including for example some laws of optics. This is known
as Duhems problem.
One major problem of Bayesian confirmation theory, called the problem of the pri85
p(H|E) =
p(E|H).p(H) 1 0.05
=
= 0.5.
p(E)
0.1
We have p(H|E) > p(H), so E confirms H - as per our intuitions. Bayesian

confirmation theory is an attractive account of confirmation, not only because
it has such intuitive results in cases like the above, but also because it can
give some solutions to several issues in the philosophy of science. For example, it can account for our intuition that if a hypothesis (p(H) = 0.05) makes
two predictions, one that seems unlikely (p(E1 ) = 0.1) and the other likely
(p(E2 ) = 0.7), the observation of the former will have a greater confirming
impact than the latter. Indeed,
p(H|E1 ) =
p(H).p(E1 |H) 0.05 1

=
= 0.5
p(E1 )
0.1
p(H|E2 ) =
p(H).p(E2 |H) 0.05 1

=
= 0.07
p(E2 )
0.7
So, the following holds. Both E1 and E2 confirm H, but E1 does so (considerably) more than E2 :
p(H) < p(H|E2 ) < p(H|E1 )
But notwithstanding its successes, BCT has some problems. This paper
aims at developing a solution to one of them, called the problem of old evidence, first raised by Clark Glymour.3 Further to the result of Eddingtons
test, another piece of evidence is taken to lend very strong support to GRT:
the explanation of the anomaly in Mercurys perihelion. According to Newtonian physics, the orbit of a given planet, to a first approximation, traces out
an ellipse with the sun at one of its focus. The point where a planet is closest
to the sun is called its perihelion (see Figure 2).
A number of factors cause the position of planets perihelions to shift, the
ors, is how to determine the values of the prior probabilities of E and H. Curd and
Cover (2012) give a good review of the debate around this issue. The values that
I have assigned to p(E) and p(H) are however not completely arbitrary. Indeed, I
could not have assigned a lower prior to E than I have to H, because since H entails
3
E, and H 6= E, p(E) > p(H) must hold.

Glymour, Why I Am Not a Bayesian 63-93.
86
Chlo de Canson
Figure 2: the Perihelion of a Planet

most important of which is the interference with other planets gravitational
fields. Newtonian mechanics, once it took these extra factors into account,
succeeded in predicting the positions of all planets perihelions with remarkable accuracy, except for that of Mercury, which had an anomalous advance
of around 43 arcseconds per tropical century.4 This was acknowledged to be
a problem for celestial mechanics in 1859. Many hypotheses were brought
forward in order to account for this precession in Mercurys perihelion, including Urbain Le Verriers postulate of a new planet, Vulcan, between Mercury and the Sun, but all such attempts at solving the puzzle turned out to be
unsuccessful. Einsteins GRT, which was constructed for reasons completely
independent of this issue, was later discovered to entail exactly an extra advance of 43 arcseconds per tropical century in Mercurys perihelion, as well
as the same shifts in its position that Newtonian mechanics predicted. This
was taken to lend significant support to GRT.
The problem of old evidence arises in BCT as it is unable to account for
this case. The anomaly (it was anomalous within the Newtonian framework)
in the perihelions precession, call it E, was discovered half way through the
19th century. It would seem then that, on any sensible analysis, any rational
4
An arcsecond is a measure of angle. One arcsecond corresponds to 1/3600 of a

degree. A tropical year is the time that the sun takes to return to the same position
in the cycle of seasons, as seen from the Earth. One hundred of those compose a
tropical century.
87

agent knew in 1915 that E held and therefore ascribed it a probability equal
to 1. Since GRT entails E, we have p(E|GRT ) = 1. Regardless of the value
of p(GRT ), the following then holds:
p(GRT |E) =
p(E|GRT ).p(GRT ) 1 p(GRT )

=
= p(GRT )
p(E)
1
This result is more general: if p(E) = 1, then for any H, whatever its
logical relation to E, the following is true:
p(H|E) =
p(E|H).p(H) 1.p(H)
=
= p(H)
p(E)
1
The reason why, if E is known (ie. p(E) = 1), then p(E|H) also equals
1, is that E|H is a subset of E (it is the intersection of E and H), and so E|H
is implied by E. Therefore, if one knows E, and is a perfect logician, as is
assumed in the Bayesian framework, then one also knows E|H.
On BCT, old evidence cannot confirm theories. Intuitively, this is very
implausible it goes against very firm intuitive judgments in particular cases,
like the orbit of Mercury, and in general there seems to be no good reason
why the fact that a piece of evidence is known should, on its own, rule out
that evidence from being confirmatory. The next sections of this discussion
will aim at modifying BCT in such a way that it can account for the intuition
that E confirms GRT.
II. The Counterfactual Response and Its Problems
Two attempted solutions of the problem of old evidence have been discussed
in the literature, namely i) adopting a new criterion for confirmation, or ii)
denying that p(E) = 1. Many have argued for a solution of the first type, for
example by claiming that we should conditionalise on the recognition (which
did not of course occur before 1915) that E is implied by H, rather than on
the fact that E is true.5 I focus here on the second type of solution. The most
common response of this type is the counterfactual response: in a nutshell,
5
A review of such proposals can be found in Curd et. al., Philosophy of Science 626.
88
Chlo de Canson
this response claims that we should assign to p(E) the value that it would
have had, had E not been known. Martin Curd and J.A. Cover6 identify
two types of counterfactual responses, the historical proposal and the present
proposal. I examine them in turn and show that neither works as a response
to the problem of old evidence.
a. The Historical Proposal
Glymour, after introducing the problem of old evidence, concedes that a natural response would be to introduce counterfactuals. His suggestion is that
we should assign to p(E) the value that it had before we came to know E.
He draws an analogy with a coin-flipping situation. Suppose that you toss
a coin, and the outcome is heads. The prior probability of the coin to yield
outcome heads should be, not 1, but 12 - the probability that was assigned to
that outcome before the experiment. Applied to the example of GRT and the
precession of Mercurys perihelion, he proposes that one could take the value
of p(E) to be that before the precession was observed.
However, Glymour notes two main problems with such a proposal. The
first is that there would be a technical difficulty linked to the Bayesian calculus. We cannot simply use the degree of belief in E that we had at a previous
time and retain our other present-day degrees of belief in all other propositions, because that would most probably violate the coherence requirement.
But even if this difficulty could be overcome, Glymour raises another, more
serious philosophical objection to the historical proposal: choosing which
historical period is relevant is not as straightforward as it seems. The discovery of the precession in Mercurys perihelion took place in stages between the
mid nineteenth century and 1912, with estimated values changing and confidence in the methods of measurement varying. Thus, at the different times
between these two dates, the value of p(E) fluctuated. How do we decide
which date to choose as the basis for p(E) in our counterfactual approach? It
seems that any such decision would be arbitrary.
6
Curd et. al., 624.

89

I argue that there is a third, related problem with the historical proposal.
Suppose that it were possible to determine which period to choose as a basis
for assigning a value to p(E). Even then, this value would be arbitrary in that
it would depend on factors unrelated to either the present beliefs of the agent
or the evidence itself. Suppose that it had been decided that the appropriate
historical period to determine p(E) was in 1619, at the time of the publication
of Keplers three laws of planetary motion. The paradigm in which scientists
were operating at the time comprised a claim about the constancy of the area
speed of planets. Thus, their credence in E would have been extremely small.
It seems arbitrary to use this value for p(E) today, because their reasons for
attributing such a small prior to E have nothing to do either with observations of Mercurys orbit, nor with any theories that are accepted today. Had
the value of p(E) determined by an agent in 1619 been greatly influenced
by observations of the phenomenon, or by hypotheses that are still accepted
today as plausible, it could have made some sense to use it as a counterfactual
value, but this was not the case. I used a rather extreme example to illustrate
my point it is unlikely that, if a period for determining the prior of E should
be used, it would be 1619 but the fact is that the same criticism can apply to whatever period is chosen, because the accepted hypotheses in science
vary greatly with time. The historical proposal being inadequate, I turn to the
present proposal.
b. The Present Proposal
The present proposal, which is upheld notably by Colin Howson and Peter Urbach,7 consists in calculating the prior of E with respect to all present
beliefs except E. There are two steps: the first is to delete E from the background knowledge B, and the second is to calculate p(E) given the new background knowledge B\E. This proposal is not vulnerable to any of the criticisms against the historical proposal, because there is no talk of historical
period. However, it has some problems of its own.
7
Howson and Urbach, Scientific Reasoning

90
Chlo de Canson
The first is brought forward by Howson and Urbach themselves. The
issue is that the content of B\E depends on the way that B and E are formulated. Suppose that an agents background knowledge contained the propositions a and b. Because Bayesian agents are assumed to be perfect logicians,
the proposition a&b is also part of the agents background knowledge. Simply removing a from B thus makes little sense, unless it is removed from an
axiomatisation of B, in which case all occurrences of a would be removed.
However, there are many different equivalent axiomatisations of B. For example, even in a simple case where intuitively all we know is a and b, B could
be axiomatised as either {a, b} or {a, a ! b}. But removing a from these sets
leaves us with two different sets: {b} and {a ! b} and there seems to be
no non-arbitrary way of deciding which one to use between the two.
Even if this problem could be solved (Howson and Urbach point to a

paper by Satoru Suzuki8 which claims to do so), there still remains a problem,
brought forward by Salmon.9 It is that of calculating the prior of E, pc (E),10
in terms of B\E. The suggestion brought forward by Howson and Urbach,
which is rather standard, is that, since pc (E) = ki=1 pc (E|Hi ).pc (Hi ) holds
where Hi is the set of all alternatives to H including H, it suffices to calculate
ki=1 pc (E|Hi ).pc (Hi ). That means, for all alternatives H1 , H2 , . . . , Hk , that:
pc (E) = pc (E|H1 ).pc (H1 ) + pc (E|H2 ).pc (H2 ) + ... + pc (E|Hk ).pc (Hk )
+pc ((H1 _ H2 _ . . . _ Hk )).pc (E|(H1 _ H2 _ . . . _ Hk ))
But, as Salmon points out, the last term is intractable. This is because,
even if all the alternative hypotheses considered by an agent are mutually
8
Suzuki, The Old Evidence Problem and AGM Theory 105-126.

Curd et. al., 624.
10 The reader is reminded that, in this subsection, p(E) is the agents counterfactual
9
credence in E, as if E was not known. To make this clear, I will use the denotation
pc to refer to this function throughout the subsection. However, keep in mind that
the fact that we are concerned with counterfactual values only affects the value of
p(E), and the values of p(E|x) for all x. That is, pc (E) differs from preal (E) and
pc (E|x) from preal (E|x), but pc (x) does not differ from preal (x) for all other x.
91
H3
H1
H2
(H1 v H2 v H3 v H4)
H4
Figure 3: Intractability of pc (E)
exclusive, they will never exhaust the possible hypotheses to explain a phenomenon, and most importantly, it will always be impossible for that agent
to know how big a proportion of possible explanations he has taken into account. Let us illustrate this on a diagram. Suppose that an agent can think
of four mutually exclusive hypotheses H1 , H2 , H3 , H4 , represented below as
rectangles whose areas represent the credence that the agent has in these hypotheses (Figure 3). She has no way of knowing how large the area representing the possibilities that she has not thought of, exactly because she has
not thought of them. Thus, according to Salmon, even if there was a way to
delete E from B, it would still be impossible to deduce pc (E) from it.
However, Howson and Urbach object to this response by recalling that
the Bayesian framework interprets probabilities in a subjective manner, probabilities are degrees of belief. Thus, since the agent does not take (H1 _
H2 _ . . . _ Hk ) into account in her system of beliefs (she has not considered
it and so her credence in (H1 _ H2 _ . . . _ Hk ) is zero), the above diagram is
not an adequate representation of the problem, and the dotted line and everything in it should be deleted. Thus, that last term may be discarded and pc (E)
becomes:
pc (E) = pc (E|H1 ).pc (H1 ) + pc (E|H2 ).pc (H2 ) + ... + pc (E|Hk ).pc (Hk )
92
Chlo de Canson
This is no longer intractable. To illustrate their point, they explain that, in
1915, the only serious alternative to GRT was Classical Gravitation Theory
(CGT), which was part of Newtonian Mechanics. Thus, according to them,
the probability of the precession in Mercurys perihelion was:
pc (E) = pc (E|CGT ).pc (CGT ) + pc (E|GRT ).pc (GRT )
Therefore, Salmons criticism fails to show that it is impossible, if E can be
removed from B, to calculate the prior probability of E.
I argue however that, although the calculation is possible, it leads to a conceptual difficulty linked to the standard Bayesian framework. The present
proposal requires that, at some time t, the value of p(E) be taken to be less
than 1, when it had been fixed at 1 at a previous time t-1. But this deletion of
knowledge is impossible in the standard Bayesian framework which works as
a series of updates by Bayesian conditionalisation: the calculus does not allow
for an agent to lower her probability from 1 to a lesser value. Furthermore,
the traditional model does not allow for an agents beliefs (even if counterfactual) not to correspond to a point in the Bayesian updating. The present
counterfactual move is therefore inconsistent with the assumptions of the traditional Bayesian framework. What options do we have to go forward? On
the one hand, we could reject the present counterfactual move altogether. On
the other, we could modify the traditional Bayesian framework, so as to make
it consistent with a solution to the problem of old evidence that, just like the
present proposal, is based on the idea that there is a reasonable prior for E
that does not construe it as old. In Section III, I opt for the latter strategy.
III. Rational Reconstruction versus Updating: Towards a Satisfactory
Solution
In this third section, I argue that, if we enlarge the scope of traditional Bayesianism in the way suggested by Mark Lange,11 the present proposal can be modified so as to provide a satisfactory answer to the problem of old evidence.
11 Lange,
Calibration and the Epistemological Role of Bayesian Conditionalization.
294-324.
93
B0, po
B1, p1
B2, p2
B3, p3
t1
t2
t3
t4
E1
E2
E3
E4
time
Figure 4: Bayesian Updating

First, I explain how Lange differentiates between updating and justification,
and how he uses that distinction to expand the standard Bayesian framework
into another that is richer and that coincides better with our intuitions about
confirmation. Then, I outline how Lange claims to use this extended framework to solve the problem of old evidence, I show some limitations to his
proposal and suggest a way in which to build on it in such a way that it becomes satisfactory.
a. Justification versus Updating, an Extended Bayesian Framework
In this subsection, I present and explain Langes suggestion of differentiating
update and justification by building a new Bayesian framework that incorporates both the diachronic and the synchronic aspects of Bayesian epistemology. In order to illustrate the arguments I make in this subsection, I introduce
an example. Alice presents some symptoms of tuberculosis and her doctor,
Bob, wants to know whether she has contracted the illness. He performs a
series of tests on Alice. In order for tuberculosis to be diagnosed, three tests
must come back positive.
In the previous section, I explained how the process of Bayesian updating via Bayesian conditionalisation works: the epistemic agent begins with
a prior probability function over the set of her beliefs, and this function is
updated according to Bayesian conditionalisation at each step at which she
encounters new evidence (where each Ei is taken to be the strongest evidential claim she learns at that stage). This can be represented as on Figure 4.
Here, the sets Bi are the sets of beliefs of the agent, and the probabilistic
94
Chlo de Canson
functions pi are her credence distributions over Bi . At time t1 , the agent
learns evidence E1 and updates her credence distribution, so that it remains
consistent as it changes from p0 to p1 . If we apply this to the example, we
get the following. Bob, before the first test, has credence say 0.3 in the claim
that Alice has contracted tuberculosis, in view of the symptoms. He then, at
t1 , gets the results of the first test, incorporates this evidence into his set of
beliefs, and updates his credence in Alice having tuberculosis using Bayes
theorem to, say 0.5. This process is repeated at times t1 , t2 , and t3 , until
his credence in that proposition becomes 1 (or very close to 1) at time t4 ,
supposing that this is when he gets the result of the third test.
A Bayesian rational reconstruction differs from such updates. Indeed,
unlike the latter, a justificatory argument is synchronic. It consists, as Lange
explains, of several steps. It begins with a primary probability distribution
over a chosen set of beliefs. The person justifying her credence in a proposition chooses which beliefs she would like to take as primary this is the
background knowledge that she considers relevant (B0 in the figure 5 below)
and she assigns a coherent probability distribution over them (p0 ). Then,
at each step si , she brings in new evidence Ei and updates her credence
function over her set of beliefs according to Bayesian conditionalisation. She
then ends up with Bi+1 and pi+1 , and can reiterate the process. Lange himself raises the obvious issue of how the agent should determine her initial, or
primary, set of beliefs B0 , and the associated probability function p0 . He
responds that this is no more problematic than the problem of the priors in
standard Bayesian confirmation theory. This is, I believe, not a satisfactory
response. Of course, just like the prior distribution at the beginning of an
agents epistemic life (supposedly then, at birth), the primary distribution in
a justificatory argument is subject to the problem of the priors. But one of
the widely accepted arguments against the significance of that problem for
standard Bayesian epistemology is the washing out of the priors namely
the fact that, as evidence accumulates, the value of the prior influences that
of the posterior less cannot be transposed to the case of justificatory arguments, because justifications are much shorter than epistemic histories, and
95

therefore evidence cannot accumulate enough. The Bayesian who accepts
this expanded framework will need to provide an additional proviso on the
primary probability distribution of justifications. I will propose one later in
the paper.
There are two things to note about the relation between updates and justificatory arguments. The first is that, in a justificatory argument, the person
justifying her belief can choose which evidence to incorporate in her argument. In the diagram above, for example, the agent has omitted E3 . The
second thing to note is that the set of beliefs B3 that the agent has at the end
of her justificatory argument is a subset of the set of beliefs that the agent really has after t4 , B4 . Indeed, B4 is the set containing all of the agents beliefs,
and B3 is only one or few of such beliefs. Let us now illustrate this Bayesian
justificatory argument using the example of Alice. Bob justifies at t4 his belief that Alice has tuberculosis. He begins by claiming that his original set
of relevant beliefs B0 contains the belief that Alice has tuberculosis and he
assigns a probability 0.3 to this proposition. Then, he brings in the result of
the first test, E1 , incorporates it into his justificatory beliefs and updates his
credence distribution so that his credence in Alice being ill becomes 0.5.
Suppose that E3 was the name of Alices mother. Learning that evidence, although it updated some of Bobs beliefs, had no incidence on his belief that
Alice had tuberculosis. Therefore, it is reasonable for him not to bring it in as
evidence in the justificatory argument. He ends up with a set of justificatory
beliefs B3 (here, that Alice has contracted tuberculosis) which is a subset,
here proper, of the set B4 of all his beliefs.
b. Solving the Problem of Old Evidence
When you ask whether observing a black raven confirms the hypothesis that
all ravens are black, you really could be asking two questions. The first is:
as you discovered that an object was a black raven, did your credence in the
hypothesis increase? This is a question about update, and it is a factual and
historical question: the agent is asked whether the update took place when
she learnt the new evidence. The second question differs: can you give a
96
Chlo de Canson
B0, po
B1, p1
B2, p2
B3, p3
B4, p4
t1
t2
t3
t4
E1
E2
E3
E4
time
B'3, p'3
s3
E4
s2
E2
s1
E1
B'2, p'2
B'1, p'1
B'0, p'o
Figure 5: Bayesian Updating and Justification

satisfactory12 justification for your degree of belief in a hypothesis, within
which your degree of belief increases from p(H) to a higher p(H|E)? Thus,
in the new framework, there are really two questions of whether old evidence
can confirm a hypothesis. The first is whether this can happen in the update,
that is, in the real epistemic history of an agent, and the second is whether an
agent can give a justification that old evidence confirms a hypothesis. In this
last subsection of the paper, I show how the introduction of the new framework which contains both updates and justifications leads to the dissolution
of both problems of old evidence. First, I uphold that in the updating, there
can indeed be no confirmation of a new theory by old evidence, and explain
why this might seem counterintuitive though in fact, it is not. Second, I show
how, by a move similar to the counterfactual present proposal, old evidence
can confirm a hypothesis in a justificatory context. Finally, I discuss some
problems, including those that applied to the counterfactual solutions (both
historical and present), and show that none of threaten the proposed solution.
12 I
will discuss what makes for a satisfactory justification at a later stage.

97

In a context of Bayesian updating, there can be no confirmation by old
evidence. I explained why this was the case above: if the prior probability of
E at the time H is proposed is 1, then the posterior probability of H, p(H|E),
is equal to its prior, p(H). As stated above, this might seem very counterintuitive, because it seems that in many examples, such as that of Einsteins GRT
and the precession of Mercurys perihelion, old evidence can confirm new
hypotheses. But I think that this is only counterintuitive because there is a
conflation between the two types (updating and justificatory) of confirmatory
arguments: we can justify (I will explain how in the next paragraph) that the
evidence from Mercury can confirm GRT, and therefore think there must also
be confirmation in the update. But actually, it fits our intuitions to deny this
latter claim. Indeed, there is no moment in the history of the agents epistemic life where the probability of H is increased by conditionalising on E.
However, although there is no confirmation, the fact that E was already known
has an impact on the probability H is assigned when it is proposed. Remember that the whole framework assumes that the agent is a perfect logician and
that her credences respect the coherence requirement. Suppose that the agent
already knew E, her credence in that evidence is 1. As she was introduced to
the hypothesis H, she had to assign a probability value to H, which had to be
consistent with the rest of her beliefs. Since she already knew E, and since
she is a perfect logician so she knows that it is entailed by H, she must assign
a relatively high prior probability to H in order to be coherent. Thus, the prior
probability of H, p(H), is affected by the fact that the agent knew E. Had she
not known E, the probability of H might not have been so high because the
coherence requirement would not have demanded it. Thus, although there is
no confirmation in the update, E has an impact on the prior of H and this
corresponds to intuitions.
There is however confirmation in the rational reconstruction. As Lange
claims, the agent can construct a justificatory argument in which E counts
as confirming evidence for H. The way that she does this is rather similar to
the earlier present proposal. She takes as her primary set of relevant beliefs a subset of her actual set of beliefs which contains the hypothesis H and
98
Chlo de Canson
all auxiliary hypotheses which she accepts do support H. Over this set, she
builds a coherent credence distribution such that her degree of belief in E is
less than 1. She then uses Howson and Urbachs proposal to determine the
value of p(E), given that primary set of beliefs. This is similar to the present
proposal of the counterfactual response in that the agent has, at the introduction of her justificatory argument, a coherent probability distribution over a
set of beliefs which corresponds to what she would know that is relevant to H
if she didnt know E. Once the agent has fully specified this primary state, at
stage 1, she introduces the evidence E as known p(E) = 1. She must then
update her original credence distribution by Bayesian conditionalisation, so
as to be consistent with this new evidence. In doing so, it will presumably
be the case that p(H|E) > p(H), and thus E confirms H in the context of a
justificatory argument. The old evidence that Mercurys perihelion is precessive can still confirm GRT for a contemporary agent: she need only give a
rational reconstruction of how she arrived at her current high degree of belief
in GRT, introducing the evidence about Mercurys orbit at some non-primary
state of the argument.
Let us consider an important issue with Langes suggestion. It might seem
that, using such rational reconstructions, any evidence can be made to confirm any hypothesis. For example, an agent could start a justification of the
Ptolemaic model of celestial movement, without construing the data of planetary motion in the primary set of beliefs as known. One could then introduce
it at some stage of the argument, in which case it would count as confirming
evidence. But intuitively, data from the stars does not confirm the Ptolemaic
model, because the parameters of that model were constructed exactly to fit
the existing data. This issue arises because there are absolutely no constraints
on what the primary probability distribution in a justification should be. As
mentioned before, Lange says about them that they need to be unbiased and
impartial, but that any further discussion would be an attempt to solve the
problem of the priors, which he does not intend to do.13 It is indeed true that
fully resolving this issue would amount to solving the problem of the priors,
13 Lange,
303.
99

but I nevertheless believe that, although a full resolution is complicated to
say the least, some improvement in the case of rational reconstruction is possible. Although, as we have seen already, the justification of a belief need
not be identical to the historical arrival to that belief, there is a case for arguing that it should be a possible way of actually arriving at that belief using
Bayesian updating. Given an initial credence distribution (here clearly the
original problem of the priors applies), there are a number of different orders
in which the evidence that was acquired could have been acquired. In my
example above, it could have been that Alices second test results came out
before the first ones, and thus Bob would have acquired his final credence
in Alices having tuberculosis in a different way. I argue that the steps of a
justification should be one such possible order, and that this suitably restricts
the possible primary distributions. There is something about justifications, or
rational reconstructions, that gives them a particular epistemic status, and this
is that, for them to be satisfactory, they need to be anchored in reality in some
way. A justification of Alices having tuberculosis that would appeal to her
mothers name, for example, would be unsatisfactory. The reason why this
might be is that, during Bobs actual Bayesian updating, the name of Alices
mother did not have an impact on his credence in Alice having tuberculosis,
and it could not have because the two are unrelated. Given this remark, I
propose the following proviso: it must be the case for a justification to be satisfactory that each step of the justification, including the primary stage, could
have been a step in the actual update, given the agents actual initial credence
distribution at t0 . Furthermore, the more plausible the justification, the more
satisfactory it is. Plausibility here is to be understood as degree of closeness
to reality, and it could be measured probabilistically: a justification would be
plausible at degree 1 if it was identical to the actual update, and 0 if it was
logically impossible for it to have been an actual update.
I argue that this entails a probabilistic version of Elie Zahar14 and John
Worralls15 criterion for confirmation by old evidence. Their original criterion
14 Zahar,
Why Did Copernicuss Research Programme Supersede Ptolemys?

New Evidence for Old
15 Worrall,
100
Chlo de Canson
states that evidence must not have been used in the construction of the theory
for it to confirm that theory. Above, I mentioned that using this framework,
it seems that the Ptolemaic model in 1543 can be confirmed by planetary
data: the agent could simply introduce the theory in the primary stage of her
justification and the data at subsequent steps. But, given the proviso from the
above paragraph, such a justification would be extremely unsatisfactory. This
is because the intricacy of the Ptolemaic model makes it extremely unlikely
for someone to conceive of it without knowledge of the data, and according
to the proviso, a justification based on such unlikely facts would have very a
very low plausibility. This can be understood as a probabilistic version of the
Zahar-Worrall criterion: if the evidence used in the construction of the theory
is introduced after the theory in a justificatory argument, because this is so
unlikely to have happened in reality, the plausibility (which, as explained in
the previous paragraph, can be understood probabilistically), and therefore,
the satisfactoriness of the confirmatory justification is very low. Therefore, if
a good justification must be a plausible possible update, the only admissible
primary degree of belief in any E that was used in the construction of some
theory H is 1.
This proposal is not vulnerable to any of the criticisms brought forward
against the counterfactual response. All the criticisms against the historical
proposal dissolve immediately as there is no reference to any historical period in a justificatory argument. This proposal is also immune to the criticisms
against the present proposal. As we saw in Section II.a., the two serious objections are that there is no non-arbitrary way of deleting E from background
knowledge, and that even if there were, it would be inconsistent with the traditional update-only Bayesian framework to do so. In the update+justification
framework, unlike in the update-only framework, there is no need to delete E
from background knowledge, there is only a need to construct a primary credence distribution such that the credence in E is not 1. Thus, the agent might
choose the primary credence distribution that she pleases and here the only
remaining problem is that of the priors. The update+justification framework
also resolves the issue of the internal inconsistency related to the deletion of
101

E from background knowledge. Indeed, in that framework, the agent can simply fail to include E in the primary set of beliefs of her justification provided
it respects the proviso. This is perfectly coherent with the general framework.
Before concluding, I consider a potential problem for using justification
as a basis for confirmation. It is possible to construct two justifications for a
belief, one in which there is confirmation, and another in which there is not.
For example, Bob could give two justifications of his belief in Alice having
tuberculosis: one in which he begins by introducing the results of the three
tests and then formulates the hypothesis that Alice is ill, and another in which
he firsts introduces his hypothesis and then brings in the evidence. There
would therefore be confirmation of the hypothesis by the evidence only in the
second justification. Lange does not consider that the same belief could be
justified by different rational reconstructions, and thus cannot directly deal
with such a problem. However, once the proviso I put forward is added to
Langes proposal, this problem can be solved. One of those two justifications
is more satisfactory because it differs less from reality, here, the second one.
Because, in the better justification, there is confirmation, that is the result
we should consider. Can such a move really be satisfactory? In historical
cases of old evidence, like GRT and the advance in Mercurys perihelion, the
best justification according to my proviso is the one that is closest to reality,
namely the one that begins by introducing the evidence and then introduces
the theory. Thus, in the best justification, there is no confirmation. Is that
problematic? As I have explained above, it is actually not counterintuitive to
think that there is no actual confirmation in such cases. Furthermore, there
is an almost imperceptibly less plausible (that means, extremely plausible
since the one that copies reality has a plausibility of 1) justification in which
confirmation does occur. I think this accounts for our intuitions about cases
of old evidence.
102
Chlo de Canson
Conclusion
In this paper, I have sought to present a solution to the Bayesian problem of
old evidence, building on a suggestion by Marc Lange. To do so, I began, in
the first section, by outlining the basic tenets of Bayesian confirmation theory, and by showing how the problem of old evidence arises within it. In
the second section, I went on to examine the counterfactual response to this
problem, by investigating the historical and present counterfactual proposals.
I concluded that neither of them is satisfactory. Finally, in the third section,
I drew on Langes suggestion of a Bayesian rational reconstruction distinct
from updating to elaborate a new more intuitively appealing Bayesian framework comprising both aspects, and I showed that, in this framework and with
a proviso, the problem of old evidence could be solved in a way that is immune to the criticisms brought forward against the counterfactual response.
103

References
[1] Curd, M., J.A. Cover, and C. Pincock. Philosophy of Science: The Central Issues. Second International Student Edition. New York: W. W.
Norton & Company, 2012.
[2] Glymour, C. Why I am not a Bayesian. Theory and Evidence. Chicago:
University of Chicago Press, 1981. 63-93.
[3] Howson, C., and P. Urbach. Scientific Reasoning: The Bayesian Approach. Open Court Publishing, 2006.
[4] Lange, M. Calibration and the Epistemological Role of Bayesian Conditionalization. Journal of Philosophy 96, no. 6 (1999): 294324.
[5] Suzuki, S. The Old Evidence Problem and AGM Theory. Annals of the
Japan Association for Philosophy of Science 13, no. 2 (2005): 10526.
[6] Worrall, J. New Evidence for Old. In In the Scope of Logic, Methodology and Philosophy of Science, Eds. Peter Gardenfors, Jan Wolenski
and Katarzyna Kijania-Placek., 191-209. London: Kluwer Academic,
2002.
[7] Zahar, E. Why Did Copernicuss Research Programme Supersede
Ptolemys? In Lakatos, I., The Methodology of Scientific Research Programmes, Vol. 1. Cambridge University Press, 1978.
Chlo de Canson has just graduated from LSE with a BSc Philosophy, Logic
and Scientific Method (2012 - 2015). From October 2015, she will be reading
the MPhil Philosophy at the University of Cambridge. Her main areas of
interest are in philosophy of science, philosophy of probability and logic.
You can contact her at [chloe.decanson@gmail.com].
104
Rerum Causae, 7 (2015), 106-116
Evolutionary Biology, Moral Intuitions and

the Is-Ought Problem
Geoff Keeling
Abstract
In this paper, I argue that deriving normative conclusions from moral
intuitions commits Humes is-ought fallacy. First, I argue that moral intuitions evolved to maximise reproductive success and not to track moral
truth. I use this to support the claim that moral intuitions are better seen
as biochemical states of affairs, rather than indicators of moral truth.
Second, I argue that this provides reason to think that deriving normative conclusions from moral intuitions commits Humes is-ought fallacy.
I defend this against two objections: (1) that the argument commits the
genetic fallacy, and (2) that the argument pre-supposes meta-ethical objectivism.
I argue that deriving normative conclusions from moral intuitions commits

Humes is-ought fallacy. First, I outline a biological account of moral intuitions. I then argue that moral intuitions, as characterised, evolved to maximise reproductive success and not to track moral truth. Second, I argue that
these considerations support the claim that moral intuitions are biochemical
Special thanks go to Farbod Akhlaghi-Ghaffarokh and Ben Lange for many helpful
discussions. Further thanks go to Dr Orsolya Reich for an informative commentary at the LSE-Bayreuth Conference, and to Dr Jonathan Birch for recommended
reading.
106
Geoff Keeling
states of affairs and not indicators of moral truth. Third, I argue that this
provides reason to think that deriving normative conclusions from moral intuitions commits Humes is-ought fallacy. Fourth, I defend this against (1) the
objection that the argument commits the genetic fallacy, and (2) the objection
that the argument pre-supposes meta-ethical objectivism.
First, I outline a biological account of moral intuitions. McMahan characterises moral intuitions as "a moral judgement [...] that is not the result of
inferential reasoning."1 I want to expand this definition, such that moral intuitions refer to the quick-fire moral judgements that correlate with increased
activity in affective or emotion-driven parts of the brain such as the ventromedial pre-frontal cortex.2 I argue that deriving normative conclusions from
moral intuitions, as characterised, commits Humes is-ought fallacy.
There is a strong empirical case to support the claim that emotional neurobiology is responsible for moral intuitions. For example, damage to the prefrontal cortex correlates with impaired moral intuitions.3 In addition, individuals with a self-reported strong sense of justice have heightened activity in
the pre-frontal cortex when viewing morally-loaded stimuli, when compared
with individuals who self-report a weak sense of justice.4 There is not scope
to outline all the relevant data.5 But I think the current data provides sufficient
reason to think that emotional neurobiology underpins moral intuitions.
If moral intuitions do have a neurological basis, then there is reason to
think that moral intuitions are subject to natural selection. The claim that
neurochemistry has a genetic basis is uncontentious.6 But in order to vindicate the claim that moral intuitions evolved to maximise reproductive success
and not to track moral truth, it is necessary to show that there exists a selection
1
2
3
4
5
McMahan, Blackwell Guide to Ethical T heory 104-105.

Shenhav and Greene, Integrative Moral Judgement 4741.
Raine and Yang, Neural Foundations to Social Behaviour 20.
Yoder and Decety, The Good, the Bad, and the Just 4161.
For two recent meta-analyses see Sevinc and Spreng, Contextual and Perceptual
Brain Processes Underlying Moral Cognition and Lisofsky et al., Investigating
Socio-Cognitive Processes in Deception.

Gazzaniga et al., Cognitive Neuroscience : T he Biology o f the Mind 588.
107

pressure such that moral intuitions evolved f or some other reason.
I argue that the reason natural selection shaped moral intuitions into their
current form is that intuitions similar to our own promote co-operation. There
is a strong empirical case from behavioural ecology suggesting that co-operation
improves the fitness of organisms.7 As such, genes coding for moral intuitions that increase the probability of co-operation are likely to be favoured
by natural selection.
This might seem counter-intuitive. On the one hand, natural selection requires a struggle for life, such that it seems to be the case that if morality exists
then it is a veneer for selfish agents to maximise payoffs in terms of reproductive success.8 On the other hand, moral sentiments feel genuine. When
humans act morally, it seems to be the case that we have genuine concern for
the wellbeing of others.9
This conflict can be settled as follows. First, it is not necessary that organisms cognise the underlying fitness benefits of their behaviour. Moral
intuitions evolved through variation and selection according to different combinations of intuitions. If genuine empathy increases fitness, then genes for
genuine empathy will spread through the gene-pool. The next issue is more
complex. How can co-operative sentiments, which often impose a cost to
reproductive fitness, spread through the gene-pool?
There are at least two senses in which co-operation promotes fitness. In
groups with high-relatedness, for example ant colonies with sterile workers
sharing 3/4 of their genes with the queen, altruism is a stable route to genepropagation. The only method for the sterile workers to pass their genes to
the next generation is to service the queen at a cost to their individual fitness.
But in doing so, they promote their inclusive fitness, which is a measure of
the individuals fitness in addition to the weighted sum of the fitness of the
7
8
Nowak, Five Rules for the Evolution of Co-operation 1560-5163.

Joyce, T he Evolution o f Morality 14; Curry, Morality as Natural History 192-
193.
This remains a matter of dispute but for good discussions see Stitch, Evolution,
Altruism and Cognitive Architecture and Sober and Wilson, Unto Others.
108
Geoff Keeling
individuals relatives. As this predisposition towards altruism arises through
emotional neurochemistry (most species do not cognise their actions), there
is reason to think that human moral intuitions developed at some early stage
to maximise cooperation, and by extension fitness, as opposed to tracking
moral truth.
The second mechanism does not require relatedness. Axelrod and Hamilton found that co-operative strategies achieve higher payoffs than spiteful
or selfish strategies in iterated prisoners dilemmas.10 Iterated prisoners
dilemmas often occur in nature, with examples as diverse as reciprocal grooming and the symbiotic relationship between whales and cleaner-fish.11 Though
there is ample opportunity to defect, for example being groomed by another
individual and not grooming them back in response, Axelrod and Hamiltons
insight was to see that co-operative strategies yield higher payoffs in the longterm, and are therefore likely to be favoured by natural selection.
McMahan expresses some concern that humans often display biologically
counter-productive behaviour, and as such, naturalistic explanations of moral
intuitions might be unsatisfactory.12 There are at least third problems with
this argument. First, biologically counter-productive is a point of view. From
the perspective of individual fitness, all altruistic behaviours are counterproductive. But from the perspective of inclusive fitness, these might have
a benefit. Second, a behaviour that imposes a cost at a time t might enhance
the reputation of an organism within a community, such that it incurs a potential benefit at a time ti.13 Third, natural selection does not strive for optimal
solutions, because each incremental change must confer a benefit. As such,
intuitions might be stuck at a local-optimum such that biologically counterproductive behaviour sometimes occurs, but on the whole, the benefit of the
intuition outweighs the cost.
10 Axelrod
and Hamilton, The Evolution of Cooperation; Axelrod and Dion, The
Further Evolution of Cooperation.

11 Trivers, The Evolution of Reciprocal Altruism.
12 McMahan, Blackwell Guide to Ethical T heory 118-119.
13 Haidt, The New Synthesis in Moral Psychology 1000.
109

In light of these considerations, I think there is good reason to think that
moral intuitions evolved to maximise co-operation, and in turn, reproductive
success. Unless the maximisation of fitness is moral truth in some respect, it
is difficult to argue that the evolution of moral intuitions converges on moral
truth. This is because natural selection is a normatively-empty algorithm that
works through differential survival of the fittest. The thesis that human moral
intuitions are categorically different to those of vampire bats punishing nonreciprocating free-riders.14 or capuchin monkeys refusing to participate in
tasks if their partner is unfairly rewarded for the same action,15 is hard to
support. As such, I think there is good reason to think that our non-inferential
judgements are in some respect independent of the natural world.16
These considerations put question to the claim that moral intuitions are
indicators of moral truth.17 In the absence of a strong counter-argument, it
seems that moral intuitions ought to be categorised alongside other neurological responses such as the fight or flight response. That is, at present there is
no good reason to think that moral intuitions provide insight into moral truth.
As such, I argue that moral intuitions ought to be seen as biochemical states
of affairs as opposed to indicators of moral truth. This provides the platform
for the central claim of the paper. If moral intuitions are biochemical state
of affairs, without normative character, then deriving normative conclusions
from moral intuitions commits Humes is-ought fallacy. That is, no normative
conclusion strictly follows from non-normative premises concerning states of
affairs in the world.18 To draw a normative claim from an emotional response
is fallacious reasoning.
This is a contentious claim. A major objection is that the argument behind the claim commits the genetic fallacy.19 Genetic fallacy occurs when
14 Curry,
T he Evolution o f Morality 64; Dawkins, T he Sel f ish Gene 230-231;
Wilkinson, Reciprocal Food-Sharing in the Vampire Bat.

15 Brosnan and de Waal, Monkeys Reject Unequal Pay 297.
16 Gintis et al., Strong Reciprocity and the Roots of Human Morality.
17 Singer, Ethics and Intuitions 337.
18 Hume, Enquiries 335.
19 Joyce, T he Evolution o f Morality 179.
110
Geoff Keeling
historical relations are conflated with logical relations.20 It might be the case
that Marx suffered from a skin condition that caused feelings of alienation.
But this is not sufficient to undermine Marxs work simpliciter, because there
might exist good independent reason to subscribe to Marxs ideas.21 In the
same line of thought, it is not immediately clear that explaining the evolutionary history of moral intuitions is sufficient to undermine the claim that moral
intuitions have normative force. There might be good independent reason to
subscribe to intuition-based moral theories.
I think there is reason to be sceptical about this objection. The argument in this paper aims to use evolutionary considerations to undermine the
justi f ication behind the use of moral intuitions to draw normative conclusions. It might be the case that moral intuitions provide a reliable indication
of moral truth. But this would require an improbable set of circumstances, in
which moral truth is found to equate with an independent evolutionary mechanism that promotes co-operation and ultimately reproductive success. The
claim is that, with the present data, there is no good reason to think that moral
intuitions provide a reliable indication of moral truth. Until data is found
to the contrary, we are not justi f ied in asserting that moral intuitions have
normative weight.
But it might be objected that this response, and indeed the entire paper,
pre-suppose meta-ethical objectivism. The argument states that the evolution
of moral intuitions does not track moral truth. But suppose there is no mindindependent moral truth. If this situation obtains, it is difficult to see the
relevance of the fact that moral intuitions do not track moral truth. In order
to support moral objectivism, it is not possible to appeal to intuition because
the same argument can be made against our intuitive sense of meta-ethical
objectivism. As such, it might be the case that an independent argument for
meta-ethical objectivism is needed for the argument to succeed.
I think this objection fails to consider the exact role of the claim that moral
intuitions did not evolve to track moral truth in the argument. The claim that
20 Cohen
and Nagel, Introduction to Logic and Scienti f ic Method 388-439.

Evolutionary Debunking Arguments 105.
21 Kahane,
111

moral intuitions did not evolve to track moral truth is used to support the
claim that moral intuitions are better seen as biochemical states of affairs,
rather than indicators of moral truth. It seems to be the case that the argument
retains its force, irrespective of whether meta-ethical objectivism obtains.
On the one hand, if objectivism does obtains, moral intuitions did not
evolve to track moral truth and are therefore best seen as biochemical states
of affairs. On the other hand, if objectivism does not obtain, moral intuitions
remain as biological states of affairs. It would take a substantial independent argument to show that moral intuitions indicate moral truth. As such,
irrespective of whether moral objectivism obtains, there is still good reason
to think that deriving normative conclusions from moral intuitions commits
Humes is-ought fallacy.
But perhaps this overstates the importance of normative claims. There
seems to be an implicit association between normative claims and some sort
of metaphysics, such the truth-values of normative claims are logically independent of states of affairs in the world. If we adopt an entirely descriptive
account of the term ought,22 then the fact that moral intuitions are biochemical states of affairs does not entail that their use in deriving normative conclusions commits a fallacy. But this reveals a problem with Humes claim
that ought cannot be derived from is. It is beyond the scope of this paper
to defend this claim, which I think there is at least some reason to doubt. I
intended to show that a certain characterisation of moral intuitions results in
an ought being derived from an is, when normative claims are derived from
these intuitions. And I think there is reason to believe this is true. But whether
this has important implications for ethics is another question that will likely
pivot on the soundness of Humes is-ought argument.
In this paper, I argued that deriving normative conclusions from moral
intuitions commits the is-ought fallacy. First, I outlined a biological account
of moral intuitions. I then argued that moral intuitions evolved to maximise
reproductive success and not to track moral truth. Second, I argued that this
supports the claim that moral intuitions are biochemical states of affairs and
22 Richards,
A Defence of Evolutionary Ethics 286-290.

112
Geoff Keeling
not indicators of moral truth. Third, I argued that this provides reason to think
that deriving normative conclusions from moral intuitions commits Humes
is-ought fallacy. Fourth, I defended this argument against two objections. I
conclude that deriving normative conclusions from moral intuitions commits
Humes is-ought fallacy.
113

References
[1] Axelrod, R. and Dion, D., 1988. The Further Evolution of Cooperation. Science, 242(4884): 1385-1390.
[2] Axelrod, R. and Hamilton, W., 1981. The Evolution of Cooperation.
Science, 211(4489): 1390- 1396.
[3] Brosnan, S. and F. de Waal, 2003. Monkeys Reject Unequal Pay.
Nature, 425(6955): 297-299
[4] Cohen, M. and Nagel, E., 1934. Introduction to Logic and Scienti f ic
Method. 1st ed. New York: Harcourt, Brace and Company.
[5] Curry, O. S., 2004. Morality as Natural History : An Adaptonist
Account o f Ethics, London School of Economics and Political Science
(PhD Dissertation).
[6] Dawkins, R., 1989. T he Sel f ish Gene. New York: Oxford University
Press.
[7] Gazzinga, M., Irvy, R. and Mangun, G., 2002. Cognitive Neuroscience :
T he Biology o f the Mind. 2nd ed. New York: W.W. Norton and Company.
[8] Gintis, H. et al., 2008. Strong Reciprocity and the Roots of Human
Morality. Social Justice Research, 21(2): 241-253.
[9] Haidt, J., 2007. The New Synthesis in Moral Psychology. Science,
Volume 316: 998-1002.
[10] Hume, D., 1989. Enquiries Concering Human Understanding and
Concering the Principles o f Morals. Oxford University Press: New
York.
[11] Joyce, R., 2007. T he Evolution o f Morality. Cambridge, MA: Massachusetts Institute of Technology Press.
114
Geoff Keeling
[12] Kahane, G., 2011. Evolutionary Debunking Arguments. Nous, 45(1):
103-125.
[13] Lisofsky, N., Kazzer, P., Heekeren, H. R. and Prehn, a. K., 2014. Investigating Socio-Cognitive Processes in Deception: A Quantitative
Meta-Analysis of Neuroimaging Studies. Neuropsychologia, Volume
61: 113-122.
[14] McMahan, J., 2013. Chapter 5: Moral Intuition. In: H. LaFollette and
I. Persson, eds. T he Blackwell Guide to Ethical T heory. Blackwell
Publishing Ltd: 103-120.
[15] Nowak, M., 2006. Five Rules for the Evolution of Co-Operation.
Science, Volume 314: 1560-1563.
[16] Raine, A. and Y. Yang, 2006. Neural Foundations to Moral Reasoning and Antisocial Behaviour. Social Cognitive and A f f ective
Neuroscience, 1(3): 203-213.
[17] Richards, R., 1986. A Defence of Evolutionary Ethics. Biology and
Philosophy, 1(3): 265-293.
[18] Sevinc, G. and R.N. Spreng, 2014. Contextual and Perceptual Brain
Processes Underlying Moral Cognition: a Quantitative Meta-Analysis
of Moral Reasoning and Moral Emotions. PloS one, 9(2), p. e87427.
[19] Shenhav, A. and J.D. Greene, 2014. Integrative Moral Judgment: Dissociating the Roles of the Amygdala and Ventromedial Prefrontal Cortex. T he Journal o f Neuroscience, 34(13): 4741-4749.
[20] Singer, P., 2005. Ethics and Intuitions. T he Journal o f Ethics, Volume
9: 331-352.
[21] Sober, E. and D.S. Wilson, 1999. Unto Others : T he Evolution and
Psychology o f Unsel f ish Behaviour. Harvard University Press.
115

[22] Stitch, S., 2007. Evolution, Altruism and Cognitive Architecture: a
Critique of Sober and Wilsons Argument for Psychological Altruism.
Biology and Philosophy, 22(2): 267-281.
[23] Trivers, R., 1971. The Evolution of Reciprocal Altruism. Quarterly
Review o f Biology: 35-57.
[24] Wilkinson, G., 1984. Reciprocal Food-Sharing in the Vampire Bat.
Nature, Volume 308: 181-4.
[25] Yoder, K. J. and J. Decety, 2014. The Good, the Bad, and the Just: Justice Sensitivity Predicts Neural Response During Moral Evaluation of
Actions Performed by Others. T he Journal o f Neuroscience, 34(12):
4161-4166.
Geoff Keeling is a BSc student in the Department of Philosophy, Logic and

Scientific Method at the London School of Economics. He is the Editor in
Chief of the British Journal of Undergraduate Philosophy, and his research
interests include philosophy of biology and cognitive science. He can be
contacted at [g.keeling@lse.ac.uk].
116
Rerum Causae, 7 (2015), 118-128
A Defence of Smiths Solution to The

Moral Problem
Shambhavi Shankar
Abstract
Michael Smiths internalist resolution to The Moral Problem serves
to establish a necessary connection between moral judgement and moral
motivation in the rational agent. Externalists, like Brink, counter Smiths
claim with the figure of the amoralist, whose moral motivation, they argue, is only contingent on antecedently-held desires. In this paper, I
draw a distinction between moral motivation if an agent judges it
right to f , she is, ceteris paribus, motivated to f and acting on moral
motivation if an agent judges it right to f , she actually f s. I show
how the amoralist explanation might account for the latter notion, but
not the former, which ultimately leaves Smiths argument for motivational internalism intact.
Introduction
The central organising question in the field of Metaethics, as contended by
Michael Smith, pertains to the apparent incompatibility between the oft-presupposed
objectivity and practicality platitudes of moral judgements. In his book, The
I wish to thank Katie Steele and Markus Loening for their helpful comments on
earlier version of this paper.
118
Shambhavi Shankar
Moral Problem, Smith aims to overcome the incompatibility by arguing that
moral judgements entail normative reasons for action, which are, by their
very nature, both objective and motivating.1 His solution has since been challenged by many Externalists including David Brink; Brink posits the figure
of the amoralist, a rational being who makes moral judgements, but fails to be
motivated by them. In this paper, I aim to show that Brinks account of moral
motivation does not ultimately challenge the validity of Michael Smiths solution to the Moral Problem.
The pape proceeds as follows: In the first section, I give a brief overview
of Smiths formulation of the Moral Problem, and follow that with an overview
of his proposed solution in the second section. In the third section, I recapitulate Brinks challenge to Smiths solution.In the fourth section, I show how
Brinks argument does not invalidate Smiths solution. I reformulate the conclusion of Smiths argument in such a way as to make its salience against
Brinks criticism apparent. The fifth section concludes.
I. The Moral Problem: An Introduction
The moral problem is constituted by three propositions about morality that are
individually plausible, but jointly inconsistent. The problem, as formulated
by Smith, is as follows:
(1) Moral judgements of the type, It is right that I f express a subjects
beliefs about an objective matter of fact, a fact about what is right for
her to do.
(2) If one judges it right that she f s, then ceteris paribus, she is motivated to f .
(3) An agent is motivated to act in a certain way just in case she has an
appropriate desire and a means-end belief, where belief and desire are,
1
Belief about the rightness of f entails a normative reason to f , which causes the
desire to f , which in turn, motivates the agent to act in accordance with the belief.
119

in Humes terms, distinct existences.2
It is worth explicating exactly wherein the inconsistency arises. In general, traditional views of moral judgements presuppose both objectivity and
practicality what Smith has referred to as moral platitudes. The first proposition is Smiths objectivity thesis. Were this true, moral judgements would
have to be beliefs, since only beliefs are truth-apt and representational of
moral facts. Beliefs, however, are not unto themselves practically motivating.
The second proposition is Smiths practicality requirement. If this were true,
moral judgements would have to be expressions of desire, since only desires
are motivating. The downside is that desires are neither objective and truthapt, nor rationally criticisable. So far, we can see that if moral judgements
are to be both objective and practical, they must be both beliefs and desires.
However, the long-standing Humean view of moral motivation states that beliefs and desires are distinct entities and additionally that a moral judgement
is sufficient to explain action only if it is supplemented with a relevant desire
to realize that judgement. Our traditional view of moral judgements is, thus,
inconsistent.
a. Strategies for Overcoming the Inconsistency
If we espouse the Humean view of belief-desire psychology, then moral judgements cannot be objective and practical simultaneously. A natural strategy to
overcome the inconsistency lies in rejecting one of the three propositions.
If we reject the objectivity thesis, we adopt the non-cognitivist perspective,
under which, we can neither account for the seemingly descriptive nature
of moral discourse, nor allow for the platitude of moral disagreement. By
denying that moral judgements are beliefs, we concede that morality is entirely contingent on the agents desires and thereby, reduce moral discourse
to the goal of convincing others to adopt our desires. If we reject the practicality requirement, we are unable to explain how and why moral motivation
2
Smith, The Moral Problem, 12.

120
Shambhavi Shankar
tracks the agents beliefs about what is right; beliefs, given their mind-toworld direction of fit, are not action-guiding. And yet, when we change our
beliefs about what is morally permitted or prohibited, we find ourselves eager to comply with these new beliefs. How, then, can we account for this?
Lastly, if we reject the Humean picture, we can neither account for our actions as goal-directed, nor explain our actions as fundamentally teleological.
3
We lose much of moralitys substantive characteristics by rejecting one of
the three conflicting propositions. Therefore, Smith concludes, we should

overcome the inconsistency while retaining all three propositions.
II. Normative Reasons: Smiths Solution to The Moral Problem
Smiths solution to the problem is a form of moral rationalism wherein the
right thing for an agent to do is the thing she would do were she fully rational. Smith sides with cognitivists in accepting the objectivity thesis presented
above. That is to say that when an agent makes a moral judgement of the kind,
It is right to f in C, she indeed expresses truth-apt beliefs about objective
moral facts. Smith also accepts the practicality requirement, but makes a distinction between explanatory and normative reasons for action. Explanatory
reasons are psychological states that link beliefs to action only when paired
with a pre-existing desire to act. Normative reasons, on the other hand, apply
to all moral judgements and are such that the belief itself entails a desire to
act (note that beliefs and desires are distinct entities here). When an agent
says, It is right to f in C, she has a normative reason to f in C.
Smith defends the claim that these normative reasons are both objective
and practical. They are objective in that, when arrived at through a process of
rational reflection and argument, they prescribe a common course of action
for all people in the same moral situation. Here, the process of rational
reflection consists of considering what the agents fully rational counterpart
would want her to do in situation C. Said another way, then, the agents belief
that It is right to f in C, is tantamount to saying the agents fully ideal
3
Copp, Belief, Reason and Motivation in the Moral Problem, 35.

121

counterpart would want her to f in C. This connection to the agents fully
rational self explains the practicality of normative reasons. Smith argues that
if an agents fully rational counterpart wants her to f in C, the agent is then
motivated to f in C. In short, if an agent judges it right to f in C i.e.
possesses a normative reason to f in C because her fully rational self wants
her to f in C then she is motivated to f in C. The upshot of this argument,
and the conclusion at which Brinks criticism is directed, is that if an agent
judges it right to f in C, and fails to be motivated to f in C, her belief-desire
set is incoherent and she is practically irrational by her own lights.4
In essence, Smiths argument commits him to the internalist camp, which
holds that there is a necessary connection between moral judgement and
moral motivation in the good and strong-willed person.5 The motivation
to act in accordance with ones moral beliefs is internal or intrinsic to the
belief itself. It is not contingent upon further considerations like the agents
fundamental proclivity to be moral or wish to be rational. Of course, to remain compatible with Humean psychology, Smith does allow that a person
who judges it right to f in C, may fail to actually f in C. Nevertheless, the
connection between judgement and motivation is mandatory in the rational
person i.e. it is impossible for a fully rational person to make a genuine
moral judgement without being moved to motivation by it. For if this is not
so, the agent fails to have a desire that she herself believes is rational to have,
which by Smiths view, is practically irrational.
III. The Amoralist: An Externalist Challenge to Smiths Solution
Smiths conception of rationality has drawn much criticism from Externalists
like David Brink, who counter with the archetype of the amoralist, the rational, strong willed individual who seemingly makes moral judgments, while
remaining utterly indifferent.6 Brink argues that there is no obvious irra4
5
6
Smith, The Moral Problem, 177.

Ibid.
Rosati, Moral Motivation.
122
Shambhavi Shankar
tionality in failing to be motivated by the moral judgements one makes, because in many cases, moral reasons bear no rational authority.
By moral reasons, Brink means, impartial other-regarding obligations
that do not apply to agents in virtue of their aims or interests.7 Rational authority refers to the prescriptions of an agent-centered conception of practical reason that rests on instrumental or prudential concerns.8 Put differently, in many cases, the dictates of an impartial morality diverge from the
ends of individual interest and make it so that fulfilling ones moral requirements does not advance ones interests. If we accept that rational action is
action that advances the agents interests, as Brink believes we should, in
such cases at least, moral judgements have no rational authority. Relaying
Brinks argument in Smiths terms, we see that in such cases, it would not be
irrational to judge it right to f in C and fail to be motivated to f in C.9 This
is especially true, Brink posits, because Smith provides no reason to favour
moral or normative reasons over instrumental or prudential reasons in ones
judgement-making and decision-making processes.
Brinks argument, if viable, challenges Smiths idea of normative reasons and the necessary and intrinsic connection between moral judgement and
moral motivation they stipulate (externalism, by definition). It shows, instead,
that any moral motivation is contingently derived from an antecedently-held
deeper desire of the agents.
IV. Defeasibility of Moral Motivation: A Defence of Smiths Solution
The upshot of Brinks challenge is that we must deny the original practicality
platitude from our explanation of moral judgements. As discussed previously,
however, rejecting this proposition is troublesome; it leaves us unable to explain how and why moral motivation so reliably tracks the agents beliefs
7
8
9
Brink, Moral Motivation, 19.

Brink, Moral Motivation, 18.
Rightness of f in C $ Normative reason to f in C ! Instrumental reasons >
Normative reason ! Agent not motivated to f in C.
123

about what is right.10 Fortunately, we need not reject the practicality requirement, since Smiths argument stands salient against Brinks criticism. In this
section, I will defend Smiths solution against the amoralist argument by citing the defeasibility of moral action. I will also introduce the notion of moral
irrationality to bridge the gap, as identified by Brink, between instrumental
and moral reasons.
Recall that the crux of Brinks argument lay in the claim that in cases
where moral and practical reasons for action conflict, it might be rational to
judge it right to f and fail to be motivated to f . Brink claims that this explains
why moral judgement cannot be intrinsically motivating. I challenge this
claim by making a distinction between moral motivation and acting upon
moral motivation. In Brinks case, when he says that moral motivation is
contingent upon antecendently-held instrumental desires, I contest that what
he actually means is that the agents acting on moral motivation is contingent
upon antecendently-held instrumental desires.
Per my view, moral motivation is not contingent i.e. it occurs intrinsically once an agent judges it right to f and discovers a normative reason
to f . The action (derived from the motivation), however, is contingent upon
the agents goals and interests and it is this contingency to which Brinks
challenge refers.11 Understood in this sense, we see how Brinks argument
might account for why the agent refrains from ultimately acting upon her
moral judgement (cases in which other-regarding moral requirements do not
advance the agents interests), but does not disprove Smiths claim that the
moral judgement is necessarily motivating. I argue that all moral agents will
be pro tanto motivated to f , at least marginally, after forming a belief about
the rightness of f . However, not all moral agents chief amongst them,
10 In
Moral Judgement and Moral Motivation (1998), Shafer-Landau makes an Exter-
nalist argument for how an underlying desire to be moral, de dicto, can still account
for this tracking, but Smith considers and rejects it as a form of moral fetishism.
The scope of this paper prevents me from covering that vein of argument here.
of f in C $ Normative reason to f in C ! Agent motivated to f in C
11 Rightness
! Instrumental reasons > Normative reason ! Agent does not f in C.

124
Shambhavi Shankar
amoralists will act upon that motivation; only those agents whose instrumental reasons do not override their moral reasons will be moved to action.12
Strong Moral Motivation: Brink attacks Strong Moral Motivation,
which mandates a necessary connection between moral judgement and
action: Agent judges it right to f ! Agent necessarily f s.
Weak Moral Motivation: Smith, however, defends only a weakened
form of moral motivation, which mandates a necessary connection between moral judgement and moral motivation: Agent judges it right to
f ! Agent is necessarily motivated to f ! Agent may or may not f .
Brinks connection between moral judgement and moral action, is stronger
than Smiths original proposition. Smith defends the connection between
moral judgement and moral motivation, but allows for defeasibility between
moral motivation and moral action. Per Smith, if someone judges it right that
she f s, then ceteris paribus, she is motivated to f . This formulation necessitates an all-things-equal connection between judgement and motivation, but
allows that an agent may fail to act on her motivation due to weakness of will
[and] other such psychological failures.13 Brink attacks a stronger version of
Internalism, claiming that only [such a restatement] makes the puzzle genuinely inconsistent.14 Thus, the simplest way to overcome Brinks challenge
is to consider the difference between being motivated to f (moral motivation)
and acting upon ones motivation to f (moral action). Smith, himself, allows
for this consideration via the inclusion of the ceteris paribus clause in his
practicality requirement.
However, I do believe that Brink makes a valid objection to Smiths conception of irrationality. Recall that for Smith, if an agent judges it right to f
in C, and fails to be motivated to f in C, her belief-desire set is incoherent
12 Rightness
of f in C $ Normative reason to f in C ! Agent motivated to f in C
! Instrumental reasons < Normative reason ! Agent f s in C.

13 Smith, The Moral Problem, 119.
14 Brink, Moral Motivation, 8.
125

and she is practically irrational by her own lights.15 Given that there is a
substantive difference between being motivated to f and actually f ing, Externalists, like Brink, urge a redefinition of Smiths conception of irrationality. They argue that when moral and instrumental reasons for action conflict,
Smith provides no reason to favour moral reasons over instrumental reasons.
He, instead, presupposes that we should prefer moral reasons, lest we become
irrational. At this juncture, Externalists assert a distinction between weak and
strong irrationality, where weak irrationality entails acting contradictory to a
reason for action and strong irrationality entails acting contradictory to the
best reason for action.16 Brinks claim follows: since the instrumental reason
for action is often the best reason, it should not be strong irrational to favour
it over the moral reason. I believe that Smith would concede that point (as he
has already allowed for defeasibility of moral action). I argue, however, that
moral considerations should not be relegated to the realm of weak irrationality, since morality is an important common concern. Therefore, we should
introduce the notion of moral irrationality, which entails acting against ones
most compelling moral reason, but in compliance with ones most compelling
practical reason. In this way, if an agent judges it right to f , but fails to actually f , she might be rational in a strong Fsense, but will be irrational in a
moral sense.
Smiths Original Irrationality Proposition
15 Smith,
The Moral Problem, 177.

Moral Judgement and Normative Reasons, 36.
16 Shafer-Landau,
126
Shambhavi Shankar
Amended Irrationality Proposition
The notion would alter Smiths conclusion so as to have it read: if an

agent judges it right to f in C, and fails to actually f in C, her belief-desire
set is incoherent and she is morally irrational.17 This restatement trumps
the Amoralist challenge by allowing for cases in which practical reasons can
override moral reasons (defeasibility of moral action), but retains Smiths Internalist assertion that moral judgements are necessarily motivating.
Conclusion
In this paper, I have showed that Brinks Externalist attack against Smiths
solution to the Moral Problem is invalid. Brink attacks a stronger version
of Smiths solution than Smith himself advocates. We can overcome Brinks
challenge by acknowledging that Smith allows for defeasibility between moral
judgement and action, but not between moral judgement and motivation.
Having conceded that point, we can bolster Smiths position by introducing
the notion of moral irrationality, which is the state of acting against ones
best moral reason for action. This restatement of Smiths conclusion accommodates the amoralist challenge while retaining the intrinsic nature of moral
motivation.
17 Originally:
if an agent judges it right to f in C, and fails to actually f in C, her
belief-desire set is incoherent and she is practically irrational.

127

References
[1] Brink, D.O. Moral Motivation. Ethics 1997; 108(1): 4-32.
[2] Copp, D. Belief, Reason, and Motivation: Michael Smiths "The Moral
Problem". Ethics 1997; 108(1): 33-54.
[3] Rosati, C.S. Moral Motivation. Stanford Encyclopedia of Philosophy (Spring 2014 Edition), Edward N. Zalta (ed.), URL =
<http://plato.stanford.edu/archives/spr2014/entries/moral-motivation/>.
[4] Shafer-Landau, R. Moral Judgement and Normative Reasons. Analysis
1999; 59(1): 33-40.
[5] Smith, M. The Moral Problem. 1994; 1-180.
Shambhavi Shankar is an MA candidate in Philosophy & Economics at the

University of Bayreuth (2014-2016). She also holds a BBA degree in Finance from the Schulich School of Business in Toronto, Canada. Her main
areas of interest concern questions about the philosophy of action, specifically intentional action, agent causation, and incompatibilism. She hopes to
continue research in these areas at the PhD level. She can be contacted at
[shambhavi.shankar@hotmail.com].
128
3rd LSEBayreuth Student

Philosophy Conference
Thursday 7 May, Room LRB.5.05
Friday 8 May, Room LRB.5.05
09.3010.00 Tea & Coffee
09.3010.00 Tea & Coffee
10.0011.10
Somayeh Tohidi, LSE
Commentary: Bryan Roberts, LSE
10.0011.10
Julia-Maria Franz, UBT
Commentary: Alex Voorhoeve, LSE
11.2012.30
Predicting Results in the Checkerboard
Model Using Only Data About Agents
Daniel Mayerhoffer, UBT
Commentary: Kamilla Buchter, LSE
11.2012.30
The Problem of Old Evidence in Bayesian
Confirmation Theory
Chlo de Canson, LSE
Commentary: Olivier Roy, UBT
12.3013.30 Lunch
12.3013.30 Lunch
13.3014.40
Is the Use of QALYs as the Currency of Healthcare Justice Unfair to People With Disabilities?
Yi Li, LSE
Commentary: Jan-Willem van der Rijt, UBT
13.3014.40
Evolutionary Biology, Moral Intuitions and the
Is-Ought Problem
Geoff Keeling, LSE
Commentary: Orsolya Reich, UBT
14.5016.00
Is the Concept of Rights Independent of
the Reasons for Them?
Franziska Poprawe, Bayreuth
Commentary: Leif Wenar, Kings College
14.5016.00
Defending Smiths Solution to the Moral
Problem against Brinks Externalist Challenge
Shambhavi Shankar, UBT
Commentary: Katie Steele, LSE
16.0016.30 Tea and Coffee
16.0016.30 Tea and Coffee
16.3018.00 Keynote Lecture

The Ethics of Nudge
Luc Bovens, LSE
16.3018.00 Keynote Lecture

The Logical Structure of Scanlons Contractualism
Olivier Roy, UBT
18.00 Conference Dinner
18.00 Drinks
We gratefully acknowledge financial support by the LSE Department of Philosophy, Logic and Scientific Method, the LSE
Teaching and Learning Centre, the Frderverein Philosophy & Economics e.V., and the Department of Philosophy, University of
Bayreuth.

Rerum Causae

Uploaded by

Copyright:

Available Formats

You might also like

Rerum Causae

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Rerum Causae

Uploaded by

Copyright:

Available Formats

Rerum Causae

Volume VII, Issue I

All rights reserved.

The Rerum Causae editorial team deeply appreciates the support

Luc Batty is an MSc Philosophy of Science student at the

BA in Philosophy from the University of Warwick (20102013).

His main interests are the philosophy of logic,

mathematics, and language.

You can contact him at

Lewis Humphreys is a BSc Philosophy, Logic and Scientific

Daniel Matthias Mayerhoffer

Rerum Causae, 7 (2015), 1-14

Putnam and the Reality of Time

for their helpful comments.

Putnam and the Reality of Time

Norton, Einstein for Everyone

Putnam and the Reality of Time

Putnam, Time and Physical Geometry 241

Sklar, Philosophy and Spacetime Physics 291.

Putnam and the Reality of Time

R = {(i.i), (u, i), (i, e), (u, u), (i, u)}

Putnam and the Reality of Time

R = {(i, i), (u, i), (e, i)}

and (u, i) 2 R, so u 2 L. That means for y = u, the value of left hand

side (LHS) of NPO is 1. On the other hand we know v(NPO) = 1. That

III. The Second Formulation

Putnam and the Reality of Time

IV. Sklars Criticism

He uses this second formu-

lation only to infer a stronger conclusion about reality of everything by introducing

Putnam and the Reality of Time

Somayeh Tohidi is a MSc student in Philosophy of Science at the London

Rerum Causae, 7 (2015), 16-28

Analytical Approximation of the Results of

Daniel Matthias Mayerhoffer

model is the one famously worked with in the literature.

Analytical Approximation of Schellings Checkerboard Model

Schelling, Models of Segregation.

Daniel Matthias Mayerhoffer

Figure 1: Utility function of the agents with q = 0.5

Here, y is the total number of agent is neighbours and x is the number

Analytical Approximation of Schellings Checkerboard Model

Table 1: Neighborhood segregation (number of own-kind, vertical)

This rate of own-colour neighbours will be called neighbourhood segrega6

Daniel Matthias Mayerhoffer

Table 2: Segregationist preference and number of available neighborhoods

However, the number of available neighbourhood compositions alone does

Analytical Approximation of Schellings Checkerboard Model

Muldoon, Smith and Weisberg, Segregation That No One Seeks.

Daniel Matthias Mayerhoffer

Table 3: Comparison of analytical and simulative results

Table 3 contrasts Q with the simulation results for a certain q . It shows

Analytical Approximation of Schellings Checkerboard Model

Daniel Matthias Mayerhoffer

8m0 withx(m0 ) < q : ui (m0 ) < ui (m )

Analytical Approximation of Schellings Checkerboard Model

Daniel Matthias Mayerhoffer

Analytical Approximation of Schellings Checkerboard Model

Rerum Causae, 7 (2015), 30-48