Professional Documents
Culture Documents
Full Chapter Web and Internet Economics 16Th International Conference Wine 2020 Beijing China December 7 11 2020 Proceedings Xujin Chen PDF
Full Chapter Web and Internet Economics 16Th International Conference Wine 2020 Beijing China December 7 11 2020 Proceedings Xujin Chen PDF
Full Chapter Web and Internet Economics 16Th International Conference Wine 2020 Beijing China December 7 11 2020 Proceedings Xujin Chen PDF
https://textbookfull.com/product/web-and-internet-economics-10th-
international-conference-wine-2014-beijing-china-
december-14-17-2014-proceedings-1st-edition-tie-yan-liu/
https://textbookfull.com/product/web-and-internet-economics-12th-
international-conference-wine-2016-montreal-canada-
december-11-14-2016-proceedings-1st-edition-yang-cai/
https://textbookfull.com/product/web-and-internet-economics-14th-
international-conference-wine-2018-oxford-uk-
december-15-17-2018-proceedings-george-christodoulou/
https://textbookfull.com/product/web-and-internet-economics-15th-
international-conference-wine-2019-new-york-ny-usa-
december-10-12-2019-proceedings-ioannis-caragiannis/
Theory and Applications of Models of Computation 16th
International Conference TAMC 2020 Changsha China
October 18 20 2020 Proceedings Jianer Chen
https://textbookfull.com/product/theory-and-applications-of-
models-of-computation-16th-international-conference-
tamc-2020-changsha-china-october-18-20-2020-proceedings-jianer-
chen/
https://textbookfull.com/product/bio-inspired-information-and-
communication-technologies-12th-eai-international-conference-
bict-2020-shanghai-china-july-7-8-2020-proceedings-yifan-chen/
https://textbookfull.com/product/information-systems-
security-16th-international-conference-iciss-2020-jammu-india-
december-16-20-2020-proceedings-salil-kanhere/
https://textbookfull.com/product/combinatorial-optimization-and-
applications-14th-international-conference-cocoa-2020-dallas-tx-
usa-december-11-13-2020-proceedings-weili-wu/
https://textbookfull.com/product/distributed-computing-and-
internet-technology-16th-international-conference-
icdcit-2020-bhubaneswar-india-january-9-12-2020-proceedings-dang-
Xujin Chen
Nikolai Gravin
Martin Hoefer
Ruta Mehta (Eds.)
ARCoSS
LNCS 12495
Web and
Internet Economics
16th International Conference, WINE 2020
Beijing, China, December 7–11, 2020
Proceedings
Lecture Notes in Computer Science 12495
Founding Editors
Gerhard Goos, Germany
Juris Hartmanis, USA
Web and
Internet Economics
16th International Conference, WINE 2020
Beijing, China, December 7–11, 2020
Proceedings
123
Editors
Xujin Chen Nikolai Gravin
Academy of Mathematics Shanghai University of Finance
and Systems Science and Economics
Chinese Academy of Sciences Shanghai, China
Beijing, China
Ruta Mehta
Martin Hoefer University of Illinois at Urbana-Champaign
Goethe University Frankfurt Urbana, IL, USA
Frankfurt am Main, Germany
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
This volume contains all regular papers and abstracts presented at the 16th Conference
on Web and Internet Economics (WINE 2020). WINE 2020 was held as an Internet
event during December 7–11, 2020, organized at Peking University, Beijing, China.
Over the last 16 years, the WINE conference series has become a leading inter-
disciplinary forum for the exchange of ideas and scientific progress across continents
on incentives and computation arising in diverse areas, such as theoretical computer
science, artificial intelligence, economics, and applied mathematics. WINE 2020 built
on the success of the previous editions of WINE (named Workshop on Internet and
Network Economics until 2013) which were held annually from 2005 to 2019.
The Program Committee was composed of 45 active researchers from the field.
They reviewed 136 submissions and decided to accept 42 papers. Each paper had at
least three reviews, with additional reviews solicited as needed. We are very grateful to
all members of the Program Committee for their insightful reviews and discussions. We
thank EasyChair for providing a virtual platform to organize the review process. We
also thank Springer for providing the proceedings and offering support for Best Paper
and Best Student Paper Awards.
In addition to the contributed talks, the program included four invited talks by
leading researchers in the field: Eric Budish (University of Chicago, USA), Jose Correa
(Universidad de Chile, Chile), Yiling Chen (Harvard University, USA), and Con-
stantinos Daskalakis (MIT, USA).
Our special thanks go to the general chair Xiaotie Deng and the local organization
team, as well as the poster chair Umang Bhaskar.
General Chair
Xiaotie Deng Peking University, China
Steering Committee
Xiaotie Deng Peking University, China
Paul Goldberg Oxford University, UK
Christos Papadimitriou Columbia University, USA
Paul Spirakis The University of Liverpool, UK
Rakesh Vohra University of Pennsylvania, USA
Andrew Yao Tsinghua University, China
Yinyu Ye Stanford University, USA
Program Committee
Bo An Nanyang Technological University, Singapore
Siddharth Barman Indian Institute of Science, India
Xiaohui Bei Nanyang Technological University, Singapore
Simina Branzei Purdue University, USA
Yang Cai Yale University, USA
Shuchi Chawla University of Wisconsin-Madison, USA
Jing Chen Stony Brook University, USA
Xi Chen Columbia University, USA
Yukun Cheng Suzhou University of Science and Technology, China
Jose Correa Universidad de Chile, Chile
Edith Elkind University of Oxford, UK
Dimitris Fotakis National Technical University of Athens, Greece
Vasilis Gkatzelis Drexel University, USA
Kira Goldner Columbia University, USA
Tobias Harks University of Augsburg, Germany
Zhiyi Huang The University of Hong Kong, China
Thomas Kesselheim University of Bonn, Germany
Max Klimm Humboldt University Berlin, Germany
viii Organization
Additional Reviewers
Matching
Markets
Markets for Efficient Public Good Allocation with Social Distancing. . . . . . . 102
Devansh Jalota, Marco Pavone, Qi Qi, and Yinyu Ye
Fairness
Fair Division with Binary Valuations: One Rule to Rule Them All . . . . . . . . 370
Daniel Halpern, Ariel D. Procaccia, Alexandros Psomas,
and Nisarg Shah
Learning
Abstracts
The Influence of One Strategic Agent on the Core of Stable Matchings . . . . . 463
Ron Kupfer
How Many Citizens Have Already Voted? The Effect of (Interim) Turnout
Rate Polls in Elections. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 464
Akaki Mamageishvili and Oriol Tejada
1 Introduction
Recent years have seen a dramatic increase in electronic marketplaces, both
in quantity and scale. Many of these are two-sided markets, meaning that the
market makes matches between two sets of agents (homeowners and guests for
Airbnb, employers and workers for Upwork, drivers and riders for Lyft and Uber,
etc.), each of whom has preferences over the other. This is in contrast to tra-
ditional resource allocation (cake cutting, Fisher markets, auctions, etc.) where
only one side of the market has preferences. Although envy-freeness and relax-
ations thereof have been studied extensively in one-sided resource allocation (this
research area is typically referred to as “fair division”), we are aware of just one
paper considering envy-freeness for two-sided preferences [22].1
There are two primary motivations for our work. The first is to simply study
fair division for two-sided preferences. The second is that in some ways, two-
sided electronic marketplaces like Airbnb, Upwork, Lyft, and Uber are actually
1
Although [22] is conceptually similar, it is in different setting of recommendation
algorithms, and so it is technically quite different.
c Springer Nature Switzerland AG 2020
X. Chen et al. (Eds.): WINE 2020, LNCS 12495, pp. 3–16, 2020.
https://doi.org/10.1007/978-3-030-64946-3_1
4 S. Gollapudi et al.
to our knowledge has not been considered for two-sided markets. We can define
EF1 equivalently for two-sided preferences: simply replace “there exists a good”
with “there exists a match”, etc.
The less obvious question is how to adapt EF1 to the repeated setting. In this
paper, we assume time is divided into discrete steps, where on each step, some
set of matches occur. Each match consists of two agents, one from each side of
the market. We would like the cumulative matching after each time step to be
EF1. We will also assume that each side of the market has the same number of
agents (if not, add “dummy” agents to the smaller side).
We consider two different versions of this model. In the first version, each time
step consists of just a single match, so we are effectively requiring the cumulative
matching to be EF1 at every point in time. We call this EF1-over-time.
However, asking for fairness at every point in time may be too strong.
Furthermore, in most real-world applications, matches would be happening in
parallel anyway. Conversely, EF1-over-time poses no restrictions on how many
matches agents receive. In real life, agents often have similar “capacities” (e.g.,
most cars have a similar number of seats) and thus should arguably receive
matches at similar rates. These concerns motivate a second version of the model,
where on each time step, each agent is matched exactly once (i.e., we select a
perfect matching). Thus each time step represents a “round” of matches (which
may happen in parallel). We still require that the cumulative matching is EF1
after time step, and we call this EF1-over-rounds.
We use vi (j) to denote agent i’s value for agent j, and assume valuations are
additive. We say that valuations are symmetric if vi (j) = vj (i) for all agents
i, j, and binary if there exists a ∈ [0, 1) such that vi (j) ∈ {a, 1} for all i, j.3
It is worth noting that for symmetric valuations, negative values for vi (j) are
subsumed in the following sense: if the algorithm ever says to match agents i
and j where vi (j) < 0, we simply ignore this and never make the match, which
gives both agents value 0 for the “match”.4
We now describe our results. Due to space constraints, all proofs are deferred
to the full version of the paper [16].
shows that for this class of valuations, fairness can be achieved without sacrificing
economic efficiency. Our algorithm runs in time O(n2.5 ) per time step.
The class of symmetric and binary valuations is somewhat restricted, but is
important to keep several things in mind. First, it is often hard to elicit more
complex valuations. Agents can easily answer binary questions such as “Would
you be happy with this match?”, but may not be able to provide a real number
value for potential matches. Second, the best interpretation of our result (in our
opinion) is that agents’ preferences are not truly binary, but that our algorithm
is guaranteeing EF1 with respect to a binary projection of the preferences. That
is, ask each agent to label each possible match as “good” or “bad”, and guarantee
EF1 with respect to those preferences. This interpretation is reinforced by the
fact that the cumulative matchings computed by our algorithm will be EF1
uniformly across all possible values a, and agents can even have different values
of a. See Sect. 3.1 for a discussion of this.
Symmetry, however, is a significant assumption on the agents’ preferences.
There are reasons to believe real-world preferences are largely symmetric: a rider
is likely to prefer a driver closer to her, and vice versa. However, a natural
question is whether this assumption is necessary.
Counterexamples. Our next set of results shows that the symmetry assumption
is in fact necessary. First, we show that for dynamic binary valuations, EF1-
over-rounds alone is impossible (Theorem 4.1). Second, for non-dynamic (i.e.,
valuations do not change over time) binary valuations, it is impossible to satisfy
EF1-over-rounds while guaranteeing a maximum weight matching for each time
step (Theorem 4.2). These impossibility results suggest that EF1-over-rounds
may be too much to ask for in the setting of two-sided repeated matching.
However, our counterexamples do not rule out the possibility of EF1-over-time,
even for general additive valuations. We leave this as our primary open question.
Beyond Binary Valuations. In a similar vein, we show that when one side of the
market has two agents, the binary assumption can also be relaxed. Specifically,
we give an algorithm which is EF1-over-time for any additive valuations.
Fair Division. Fair division has a long history. In fact, the Bible documents
Abraham and Lot’s use of the cut-and-choose protocol to fairly divide land.
The formal study of fair division was started by [27] in 1948, and envy-freeness
was proposed in 1958 [15] and further developed by [13]. A full overview of the
fair division literature is outside the scope of this paper (we refer the interested
reader to [5,21]), and we discuss only the work most relevant to our own.
There are two main differences between our work and that of traditional fair
division. First, we study two-sided preferences instead of one-sided preferences.
Second, we study a repeated setting, where we must make an irrevocable decision
on each time step; most fair division research considers a “one-shot” model where
all the goods are allocated at once.
We briefly overview some key results in the one-sided one-shot model of fair
division for indivisible items7 . The EF1 property was proposed for this model
by [6]. EF1 allocations always exist, and can be computed in polynomial time,
even for general combinatorial valuations [20]8 . This sweeping positive result for
the one-sided one-shot model lies in stark contrast to our negative results for
the two-sided repeated model. It was later shown that for additive valuations,
maximizing the product of valuations yields an allocation that is both EF1 and
Pareto optimal [8].
Envy-Freeness up to Any Good (EFX). The same paper suggested a new fairness
notion that is strictly stronger than EF1, which they called EFX9 . The first for-
mal results regarding EFX allocations were given by [23]. A major breakthrough
recently proved the existence of EFX allocations for additive valuations and three
agents [9], but despite ongoing effort, the question of existence remains unsolved
for more than three agents (or more complex valuations). This is perhaps the
most significant open problem in the fair division of indivisible items.
In many contexts (especially for additive valuations), it is common to modify
the requirement to be that whenever i envies j, removing any good which i
values positively from j’s bundle is sufficient to eliminate the envy [8].10 Under
the latter definition, for {0, 1} binary valuations, EF1 and EFX coincide. This
is because the only positively valued goods are the maximum value goods. In
this sense, our positive results for {0, 1} binary valuations immediately extend
to EFX as well.
Repeated Fair Division. There are a smattering of recent works studying envy-
freeness for one-sided preferences in a repeated (i.e., not one-shot) setting; see [1]
7
Indivisible items, such as cars, must each go entirely to a single agent. In contrast,
divisible items, such as a cakes, can be split between multiple players. Fair division
studies both of these settings, but the indivisible case is more relevant to our work.
8
The algorithm of [20] was originally developed with a different property in mind.
9
An allocation is envy-free up to any good (EFX) if whenever i envies j, removing
any good from j’s bundle eliminates the envy.
10
The reason this is less common when considering non-additive valuations is that for
general combinatorial valuations, it is less clear what “values positively” means.
8 S. Gollapudi et al.
for a short survey. One example is [2], which focuses on minimizing the maximum
envy (i.e., the maximum difference between an agent’s value for her own bundle
and her value for another agent’s bundle) at each time step. Despite the growing
interest in repeated one-sided fair allocation, the literature on the analogous
two-sided problem remains sparse.
The paper proceeds as follows. Section 2 describes the formal model. Section 3
presents our main result: an algorithm for symmetric and binary valuations such
that (1) the sequence of cumulative matchings is EF1-over-rounds, (2) a max-
imum weight matching is chosen for each time step, and (3) this holds even
for dynamic valuations (Theorem 3.1). Section 4 presents our counterexamples:
Theorem 3.1 cannot be extended to non-symmetric binary valuations. Specifi-
cally, without symmetry, (1) and (3) together and (1) and (2) together are both
11
For a broad overview of this topic, see [24].
Almost Envy-Free Repeated Matching in Two-Sided Markets 9
impossible (Theorems 4.1 and 4.2, respectively). The rest of the results (and all
of the proofs) can be found in the full version of the paper [16].
2 Model
Let N and M be two sets of agents. We assume that |N | = |M | = n; if this
is not the case, we can add “dummy” agents (i.e., agents i such that vi (j) =
vj (i) = 0 for all j) to the smaller side of the market until both sides have the
same number of agents. We will typically use odd numbers for the elements
of N and even numbers for the elements of M , i.e., N = {1, 3, . . . , 2n − 1} and
M = {2, 4, . . . , 2n}. A matching X assigns a multiset of agents in N to each agent
in M , and a multiset of agents in M to each agent in N . For each i ∈ N ∪ M , we
will use Xi to denote agent i’s bundle, i.e., the multiset of agents she is matched
to. Throughout the paper, we will use standard set notation for operations on
the multisets Xi . For example, Xi ∪ {j} increments the multiplicity of j in Xi ,
i.e., the number of times j occurs in Xi . In order for X to be a valid matching,
the multiplicity of j in Xi must be equal to the multiplicity of i in Xj for each
i ∈ N, j ∈ M.
Each agent i also has a valuation function vi , which assigns a real number
to each possible bundle she might receive. We will use v to denote the valuation
profile which assigns valuation vi to agent i. We say that vi is additive if for any
bundle Xi ,
vi (X) = vi ({j})
j∈X
“round” into n time steps, each containing one match, in an arbitrary order. We
know that at the end of each round of n time steps, the cumulative matching
is EF1. Within each round, each agent only gains one additional match, and we
can always remove that match to return to an EF1 state.
Our goal will be to show the existence of (and efficiently compute) a sequence
x1 , x2 , x3 . . . such that the induced sequence X is EF1-over-time and/or EF1-
over-rounds. For brevity, if an algorithm is guaranteed to produce a sequence X
that is EF1-over-time (resp., EF1-over-rounds), we simply say that the algorithm
is EF1-over-time (resp., EF1-over-rounds).
is never changed. The algorithm starts with a tentative matching, and changes
tentative matches until it is satisfied for the current time step (see Fig. 1), at
which point the matches are confirmed with MakeMatch. The algorithm then
proceeds to the next time step and never changes pairings from previous time
steps. Note also that the algorithm uses no information about valuations for
future time steps.
N: 1 3 1 3
M: 2 4 2 4
Lemma 3.1. Let t ≥ 1 be any time step, and suppose that X t−1 is (1 − a)-
envy-bounded and has no envy cycles. Then X t is (1 − a)-envy-bounded and
has no envy cycles. Furthermore, the chosen matching xt is a maximum weight
matching (with respect to the valuations on that time step).
Before diving into the proof of Lemma 3.1 (and the runtime analysis), we
briefly show that (1 − a)-envy-boundedness will actually give us the result we
want:
The role of a. Before diving into the main proof, we briefly discuss the role of
a. For Theorem 3.1, we assume that there exists a ∈ [0, 1) such that vi (j) ∈ {a, 1}
for all i, j ∈ N or i, j ∈ M . For ease of notation, we assume that all agents have
the same value of a, but this is in fact not necessary. In fact, Algorithm 1 will
be EF1-over-rounds simultaneously for all values of a.
Lemma 3.3. Assume (i, j) is (1 − a)-envy-bounded and that |Xit | = |Xjt |. Let
a ∈ [0, 1), and define a new valuation vi such that vi (k) = a whenever vi (k) = a,
and vi (k) = 1 otherwise. Then (i, j) is (1 − a )-envy-bounded with respect to vi .
Note that the assumption of |Xit | = |Xjt | is always satisfied when working
with EF1-over-rounds, since we will match every agent once on each time step.
Therefore we can actually just choose an arbitrary value of a ∈ [0, 1) and run
Algorithm 1. Lemma 3.3 implies that the resulting sequence of matchings will be
EF1-over-rounds simultaneously for all values of a, even if different agents have
different values of a. That said, if we need to include dummy agents in order to
equalize the sizes of N and M , a = 0 probably makes the most sense.
4 Counterexamples
A natural question is whether Theorem 3.1 can be extended to all dynamic binary
valuations (i.e., not necessarily symmetric). The answer is unfortunately no,
which we show in two different ways. First, for dynamic binary valuations, EF1-
over-rounds alone is impossible (Theorem 4.1). Second, for non-dynamic binary
valuations, it is impossible to guarantee both EF1-over-rounds and maximum
weight matching for each time step (Theorem 4.2).
t=1 t=2
1 3 1 3
2 4 2 4
Theorem 4.1 uses the instance in Fig. 2. Essentially, after the first round,
either agents 1 and 3 form an envy cycle, or agents 2 and 4 form an envy cycle.
14 S. Gollapudi et al.
After the second round of matching, one of the agents in the envy cycle will
become even more envious, violating EF1.
Theorem 4.1. For dynamic and binary valuations, there is no algorithm which
is EF1-over-rounds.
1, 3 5
2 4 6
Theorem 4.2 uses the instance in Fig. 3. For some intuition, note that are two
cycles of desire: (1, 4, 5, 6) and (3, 4, 5, 6). Like in the previous counterexample,
these cycles will cause problems, but here we have the additional consideration
that agents 1 and 3 are competing for agent 4. We show that the frequency with
which agents 4 and 5 are matched is at least the frequency with which either
agent 1 or agent 3 is matched with agent 4. For example, if agents 1 and 3 have
each been matched to agent 4 twice, then agents 4 and 5 will have been matched
4 times. This leads to agents 1 and 3 increasingly envying agent 5, until EF1 is
violated.
The assumption of maximum weight is necessary only to prevent agents 2
and 5 from ever being matched: if agents 2 and 5 can be matched, the above
argument can be circumvented.
5 Conclusion
of EF1-over-time, even for general additive valuations. More broadly, future work
could investigate other possible fairness notions for this setting.
Another possible future direction concerns more general study of two-sided
preferences. Envy-freeness is an example of a topic that has been widely studied
for one-sided resource allocation, but not for two-sided markets. We wonder if
there are other such topics that are worthy of study for two-sided preferences.
References
1. Aleksandrov, M., Walsh, T.: Online Fair Division: A Survey. arXiv preprint
arXiv:1911.09488 (2019)
2. Benade, G., Kazachkov, A.M., Procaccia, A.D., Psomas, C.A.: How to make envy
vanish over time. In: Proceedings of the 2018 ACM Conference on Economics and
Computation, EC 2018, pp. 593–610. ACM, New York (2018)
3. Bokányi, E., Hannák, A.: Evaluating Algorithm Fairness in an Agent-based Taxi
Simulation (2018, working paper)
4. Bokányi, E., Hannák, A.: Ride-share Matching Algorithms Generate Income
Inequality (2019, working paper)
5. Brandt, F., Conitzer, V., Endriss, U., Lang, J., Procaccia, A.D.: Handbook of
Computational Social Choice. Cambridge University Press, Cambridge (2016)
6. Budish, E.: The combinatorial assignment problem: approximate competitive equi-
librium from equal incomes. J. Polit. Econ. 119(6), 1061–1103 (2011)
7. Calo, R., Rosenblat, A.: The taking economy: uber, information, and power.
Columbia Law Rev. (2017)
8. Caragiannis, I., Kurokawa, D., Moulin, H., Procaccia, A.D., Shah, N., Wang, J.:
The unreasonable fairness of maximum Nash welfare. In: Proceedings of the 2016
ACM Conference on Economics and Computation, pp. 305–322. ACM (2016)
9. Chaudhury, B.R., Garg, J., Mehlhorn, K.: EFX exists for three agents. arXiv
preprint arXiv:2002.05119 (2020)
10. Damiano, E., Lam, R.: Stability in dynamic matching markets. Games Econ.
Behav. 52, 34–53 (2005)
11. Echenique, F., Oviedo, J.: A theory of stability in many-to-many matching markets.
Theor. Econ. 1(2), 233–273 (2006)
12. Fieseler, C., Bucher, E., Hoffmann, C.P.: Unfairness by design? The perceived
fairness of digital labor on crowdworking platforms. J. Bus. Ethics 156(4), 987–
1005 (2019)
13. Foley, D.K.: Resource allocation and the public sector. Yale Econ. Essays 7(1),
45–98 (1967)
14. Gale, D., Shapley, L.S.: College admissions and the stability of marriage. Am.
Math. Mon. 69(1), 9–15 (1962)
15. Gamow, G., Stern, M.: Puzzle-math. Viking Adult (1958). http://www.worldcat.
org/isbn/0670583359
16. Gollapudi, S., Kollias, K., Plaut, B.: Almost Envy-free Repeated Matching in Two-
sided Markets. arXiv preprint arXiv:2009.09336 (2020)
17. Hannák, A., Wagner, C., Garcia, D., Mislove, A., Strohmaier, M., Wilson, C.:
Bias in online freelance marketplaces: evidence from Taskrabbit and Fiverr. In:
Proceedings of the 2017 ACM Conference on Computer Supported Cooperative
Work and Social Computing, CSCW 2017, pp. 1914–1933. ACM, New York (2017)
16 S. Gollapudi et al.
18. Haruvy, E., Ünver, M.U.: Equilibrium selection and the role of information in
repeated matching markets. Econ. Lett. 94(2), 284–289 (2007)
19. Konishi, H., Ünver, M.U.: Credible group stability in many-to-many matching
problems. J. Econ. Theor. 129(1), 57–80 (2006)
20. Lipton, R.J., Markakis, E., Mossel, E., Saberi, A.: On approximately fair allocations
of indivisible goods. In: Proceedings of the 5th ACM conference on Electronic
commerce, pp. 125–131. ACM (2004)
21. Moulin, H.: Fair Division and Collective Welfare. MIT Press, Cambridge (2004)
22. Patro, G.K., Biswas, A., Ganguly, N., Gummadi, K.P., Chakraborty, A.: FairRec:
two-sided fairness for personalized recommendations in two-sided platforms. Proc.
Web Conf. 2020, 1194–1204 (2020)
23. Plaut, B., Roughgarden, T.: Almost envy-freeness with general valuations. In: Pro-
ceedings of the 29th Annual ACM-SIAM Symposium on Discrete Algorithms, pp.
2584–2603. Society for Industrial and Applied Mathematics (2018)
24. Roth, A.E., Sotomayor, M.: Two-Sided Matching: A Study in Game-Theoretic
Modeling and Analysis. Cambridge University Press, Cambridge (1990)
25. Sotomayor, M.: Three remarks on the many-to-many stable matching problem.
Math. Soc. Sci. 38(1), 55–70 (1999)
26. Sotomayor, M.: Implementation in the many-to-many matching market. Games
Econ. Behav. 46(1), 199–212 (2004)
27. Steinhaus, H.: The problem of fair division. Econometrica 16(1), 101–104 (1948)
28. Sühr, T., Biega, A.J., Zehlike, M., Gummadi, K.P., Chakraborty, A.: Two-sided
fairness for repeated matchings in two-sided markets: a case study of a ride-hailing
platform. In: Proceedings of the 25th ACM SIGKDD International Conference on
Knowledge Discovery & Data Mining, KDD 2019, pp. 3082–3092. ACM, New York
(2019)
29. Wolfson, O., Lin, J.: Fairness versus optimality in ridesharing. In: 2017 18th IEEE
International Conference on Mobile Data Management (MDM), pp. 118–123. IEEE
(2017)
Another random document with
no related content on Scribd:
This mystical error is distinctly characterized in the first chapter of
this gospel, and is there met by the direct assertions, that in Jesus
Christ, the Word, and the God, was not only life, but that the life
itself was the light of men;――and that John the Baptist “was not
the Light, but was only sent to bear witness of the Light;” and
again, with all the tautological earnestness of an old man, the aged
writer repeats the assertion that “this was the true Light, which
enlightens every man that comes into the world.” Against these
same sectaries, the greater part of the first chapter is directed
distinctly, and the whole tendency of the work throughout, is in a
marked manner opposed to their views. With them too, John had
had a local connection, by his residence in Ephesus, where, as it is
distinctly specified in the Acts of the Apostles, Paul had found the
peculiar disciples of John the Baptist long before, on his first visit to
that city; and had successfully preached to some of them, the
religion of Christ, which before was a strange and new thing to them.
The whole tendency and scope of this gospel, indeed, as directed
against these two prominent classes of heretics, both Gnostics and
Sabians, are fully and distinctly summed up in the conclusion of the
twentieth chapter;――“These things are written, that ye might
believe that Jesus is the Christ, the Son of God, and that in
believing on him, ye might have life through his name.”
All that has been said on the character and the objects of the
gospel, may be exactly applied to this very similar production. So
completely does it resemble John’s gospel, in style, language,
doctrines and tendencies, that even a superficial reader might be
ready to pronounce, on a common examination, that they were
written in the same circumstances and with the same object. This
has been the conclusion at which the most learned critics have
arrived, after a full investigation of the peculiarities of both,
throughout; and the standard opinion now is, that they were both
written at the same time and for the same persons. Some reasons
have been given by high critical authority, for supposing that they
were both written at Patmos, and sent together to Ephesus,――the
epistle serving as a preface, dedication, and accompaniment of the
gospel, to those for whom it was intended, and commending the
prominent points in it to their particular attention. This beautiful and
satisfactory view of the object and occasion of the epistle, may
certainly be adopted with great propriety and justice; but in regard to
the places of its composition and direction, a different view is much
more probable, as well as more consistent with the notion, already
presented above, of the date and place of the gospel. It is very
reasonable to suppose that the epistle was written some years after
John’s return to Ephesus,――that it was intended, (along with the
gospel, for the churches of Asia generally, to whom John hoped to
make an apostolic pastoral visit, shortly,) to confirm them in the faith,
as he announces in the conclusion. There is not a single
circumstance in gospel or epistle, which should lead any one to
believe that they were directed to Ephesus in particular. On the
contrary, the total absence of anything like a personal or local
direction to the epistle, shows the justice of its common title, that it is
a “general epistle,” a circular, in short, to all the churches under his
special apostolic supervision,――for whose particular dangers,
errors and necessities, he had written the gospel just sent forth, and
to whom he now minutely commended that work, in the very opening
words of his letter, referring as palpably and undeniably to his
gospel, as any words can express. “Of that which ‘was from the
beginning, of the Word,’ which I have heard, which I have seen with
my eyes, which I have looked upon, which my hands have
handled,――of the Word of Life” &c.; particularizing with all the
minute verbosity of old age, his exact knowledge of the facts which
he gives in his gospel, assuring them thus of the accuracy of his
descriptions. The question concerns his reputation for fidelity as a
historian; and it is easy to see therefore, why he should labor thus to
impress on his readers his important personal advantages for
knowing exactly all the facts he treats of, and all the doctrines which
he gives at such length in the discourses of Christ. Again and again
he says, “I write,” and “I have written,” recapitulating the sum of the
doctrines which he has designed to inculcate; and he particularizes
still farther that he has written to all classes and ages, from the
oldest to the youngest, intending his gospel for the benefit of all. “I
have written to you, fathers,”――“unto you, young men,”――“unto
you, little children,” &c. What else can this imply, than a dedication of
the work concerning “the WORD,” to all stations and ages,――to the
whole of the Christian communities, to whom he commits and
recommends his writings;――as he writes “to the
fathers――because they know him who was from the
beginning,”――in the same way “to the young men, because they
are constant, and the Word of God dwells in them,” and “that the
doctrine they have received may remain unchangeable in them,” and
“on account of those who would seduce them.” He recapitulates
all the leading doctrines of his gospel,――the Messiahship, and the
Divinity of Jesus,――his Unity, and identity with the divine
abstractions of the Gnostic theology. Here too, he inculcates and
renewedly urges the great feeling of Christian brotherly love, which
so decidedly characterizes the discourses of Jesus, as reported in
his gospel. So perfect was the connection of origin and design,
between the gospel and this accompanying letter, that they were
anciently placed together, the epistle immediately following the
gospel; as is indubitably proved by certain marks in ancient
manuscripts.
It was mentioned, in connection with a former part of John’s life, that this epistle is
quoted by Augustin and others, under the title of the epistle to the Parthians. It seems very
probable that this may have been also addressed to those churches in the east, about
Babylon, which had certainly suffered much under the attacks of these same mystical
heretics. It is explained, however, by some, that this was an accidental corruption in the
copying of the Greek.――The second epistle was quoted by Clemens Alexandrinus, under
the title of “the epistle to the virgins,” προς παρθενους, which, as some of the modern critics
say, must have been accidentally changed to παρθους, by dropping some of the syllables,
and afterwards transferred to the first (!) as more appropriate;――a perfectly unauthorized
conjecture, and directly in the face of all rules of criticism.
These are both evidently private letters from John to two of his
intimate personal friends, of whose circumstances nothing whatever
being known, except what is therein contained, the notice of these
brief writings must necessarily be brief also. They are both honorably
referred to, as entertainers of the servants of Jesus Christ as they
travel from place to place, and seem to have been residents in some
of the Asian cities within John’s apostolic circuit, and probably
received him kindly and reverently into their houses on his tours of
duty; and them he was about to visit again shortly. The second
epistle is directed to a Christian female, who, being designated by
the very honorable title of “lady,” was evidently a person of rank; and
from the remark towards the conclusion, about the proper objects of
her hospitality, it is plain that she must have been also a person of
some property. Mention is made of her children as also objects of
warm affection to the aged apostle; and as no other member of her
family is noticed, it is reasonable to conclude that she was a widow.
The contents of this short letter are a mere transcript, almost
verbatim, of some important points in the first, inculcating Christian
love, and watchfulness against deceivers;――(no doubt the
Gnostical heretics,――the Cerinthians and Nicolaitans.) He
apologizes for the shortness of the letter, by saying that he hopes
shortly to visit her; and ends by communicating the affectionate
greetings of her sister’s children, then residents in Ephesus, or
whatever city was then the home of John. The third epistle is
directed to Gaius, (that is, Caius, a Roman name,) whose hospitality
is commemorated with great particularity and gratitude in behalf of
Christian strangers, probably preachers, traveling in his region.
Another person, named Diotrephes, (a Greek by name, and probably
one of the partizans of Cerinthus,) is mentioned as maintaining a
very different character, who, so far from receiving the ministers of
the gospel sent by the apostle, had even excluded from Christian
fellowship those who did exercise this hospitality to the messengers
of the apostle. John speaks threateningly of him, and closes with the
same apology for the shortness of the letter, as in the former. There
are several persons, named Gaius, or Caius, mentioned in apostolic
history; but there is no reason to suppose that any of them was
identified with this man.
For these lucid views of the objects of all these epistles, I am mainly indebted to Hug’s
Introduction, to whom belongs the merit of expressing them in this distinctness, though
others before him have not been far from apprehending their simple force. Michaelis, for
instance, is very satisfactory, and much more full on some points. In respect to the place
whence they were written, Hug appears to be wholly in the wrong, in referring them to
Patmos, just before John’s return. Not the least glimmer of a reason appears, why all the
writings of John should be huddled together in his exile. I can make nothing whatever of the
learned commentator’s reason about the deficiency of “pen, ink and paper,” (mentioned in
Epistle ii. 12, and iii. 13.) as showing that John must still have been in “that miserable
place,” Patmos. The idea seems to require a great perversion of simple words, which do not
seem to be capable of any other sense than that adopted in the above account.
To this period of his life, are referred those stories of his miracles
and actions, with which the ancient fictitious apostolic narratives are
so crowded,――John being the subject of more ancient traditions
than any other apostle. Some of those are so respectable and
reasonable in their character, as to deserve a place here, although
none of them are of such antiquity as to deserve any confidence, on
points where fiction has often been so busy. The first which follows,
is altogether the most ancient of all apostolic stories, which are not in
the New Testament; and even if it is a work of fiction, it has such
merits as a mere tale, that it would be injustice to the readers of this
book, not to give them the whole story, from the most ancient and
best authorized record.
For this series of fables I am indebted again to the kindness of Dr. Murdock, in whose
manuscript lectures they are so well translated from the original romances, as to make it
unnecessary for me to repeat the labor of making a new version from the Latin. The sight of
the results of abler efforts directly before me, offers a temptation to exonerate myself from a
tedious and unsatisfactory effort, which is too great to be resisted, while researches into
historical truth have a much more urgent claim for time and exertion.
The only one of all these fables that occurs in the writings of the Fathers, is the first,
which may be pronounced a tolerably respectable and ancient story. It is narrated by
Clemens Alexandrinus, (about A. D. 200.) The story is copied from Clemens Alexandrinus
by Eusebius, from whom we receive it, the original work of Clemens being now lost.
Chrysostom also gives an abridgement of the tale. (I. Paraenes ad Theodosius) Anastasius
Sinaita, Simeon Metaphrastes, Nicephorus Callistus, the Pseudo-Abdias, and the whole
herd of monkish liars, give the story almost verbatim from Clemens; for it is so full in his
account as to need no embellishment to make it a good story. Indeed its completeness in all
these interesting details, is one of the most suspicious circumstances about it; in short, it is
almost too good a story to be true. Those who wish to see all the evidence for and against
its authenticity, may find it thoroughly examined in Lampe’s Prolegomena to a Johannine
Theology (I. v. 4‒10.) It is, on the whole, the best authorized of all the stories about the
apostles, which are given by the Fathers, and may reasonably be considered to have been
true in the essential parts, though the minute details of the conversations, &c., are probably
embellishments worked in by Clemens Alexandrinus, or his informants.
The rest of these stories are, most unquestionably, all unmitigated falsehoods; nor does
any body pretend to find the slightest authority for a solitary particular of them. They are
found no where but in the novels of the Pseudo-Abdias, and the martyrologies. (Abdiae
Babyloniae episcopi et Apostolorum discipuli de Historia, lib. V., St. John.)
his death.
Respecting the close of his life, all antiquity is agreed that it was
not terminated by martyrdom, nor by any violent death whatever, but
by a calm and peaceful departure in the course of nature, at a very
great age. The precise number of years to which he attained can not
be known, because no writer who lived within five hundred years of
his time has pretended to specify his exact age. It is merely
mentioned on very respectable ancient authority, that he survived to
the beginning of the reign of Trajan. This noblest of the successors
of Julius, began his splendid reign in A. D. 98, according to the most
approved chronology; so that if John did not outlive even the first
year of Trajan, his death is brought very near the close of the first
century; and from what has been reasonably conjectured about his
age, compared with that of his Lord, it may be supposed that he
attained upwards of eighty years,――a supposition which agrees
well enough with the statement of some of the Fathers, that he died
worn out with old age.
Jerome has a great deal to say also, about the age of John at the time when he was
called, arguing that he must have been a mere boy at the time, because tradition asserts
that he lived till the reign of Trajan. Lampe very justly objects, however, that this proof
amounts to nothing, if we accept another common tradition, that he lived to the age of 100
years; which, if we count back a century from the reign of Trajan, would require him to have
attained mature age at the time of the call. Neither tradition however, is worth much. Our old
friend Baronius, too, comes in to enlighten the investigation of John’s age, by what he
considers indubitable evidence. He says that John was in his twenty-second year when he
was called, and passing three years with Christ, must have been twenty-five years old at the
time of the crucifixion; “because,” says the sagacious Baronius, “he was then initiated into
the priesthood.” An assertion which Lampe with indignant surprise stigmatizes as showing
“remarkable boldness,” (insignis audacia,) because it contains two very gross
errors,――first in pretending that John was ever made a priest, (sacerdos,) and secondly in
confounding the age required of the Levites with that of the priests when initiated. For
Baronius’s argument resting wholly on the very strange and unfounded notion, that John
was made a priest, is furthermore supported on the idea that the prescribed age for entering
the priesthood was twenty-five years; but in reality, the age thus required was thirty years,
so that if the other part of this idle story was true, this would be enough to overthrow the
conclusion. Lampe also alludes to the absurd idea of the painters, in representing John as a
young man, even while writing his gospel; while in reality all writers agree that that work was
written by him in his old age. This idea of his perpetual youth, once led into a blunder some
foolish Benedictine monks, who found in Constantinople an antique agate intaglio,
representing a young man with a cornucopia, and an eagle, and with a figure of victory
placing a crown on his head. This struck their monkish fancies at once, as an
unquestionable portrait of John, sent to their hands by a miraculous preservation.
Examination however, has shown it to be a representation of the apotheosis of Germanicus.
But even here, the monkish inventors have found room for new
fables; and though the great weight of all ancient testimony deprives
them of the opportunity to enter into the horrible details of a bloody
and agonizing death, they can not refuse themselves the pleasure of
some tedious absurdities, about the manner of his death and burial,
which are barely worth a partial sketch, to show how determined the
apostolic novelists are to follow their heroes to the very last, with the
glories of a fancifully miraculous departure.
Protoclete.――Hammond claims this peculiar honor for Philip, with great zeal. (See
his notes on John i. 43.)
The testimony of the Fathers on this point, is simply this. Eusebius (Church History, III.
31,) quotes Polycrates, bishop of Ephesus, who, in his letter to Victor, bishop of Rome,
(written A. D. 195, or 196,) makes mention of Philip in these exact words: “Philip, who was
one of the twelve apostles, died in Hierapolis;” (in Phrygia;) “and so did two of his
daughters, who had grown old in virginity. And another of his daughters, after having
passed her life under the influence of the Holy Spirit, was buried at Ephesus.” This certainly
is a most perfect identification of Philip the apostle with Philip the deacon; for it is this latter
person who is particularly mentioned in Acts, xxi. 8, 9, as “having four daughters who did
prophesy.” He is there especially designated as “Philip the evangelist, one of the seven,”
while Polycrates expressly declares, that this same person “was one of the twelve.”
Eusebius also, in the preceding chapter, quotes Clemens Alexandrinus as mentioning Philip
among those apostles who were married, because he is mentioned as having had
daughters; and Clemens even adds that these were afterwards married, which directly
contradicts the previous statement of Polycrates, that three of them died virgins, in old age.
Yet Eusebius quotes all this stuff, with approbation.
Papias, (A. D. 140,) bishop of Hierapolis, the very place of the death and burial of Philip,
is represented by Eusebius as having been well acquainted with the daughters of Philip,
mentioned in Acts, as the virgin prophetesses. Papias says that he himself “heard these
ladies say that their father once raised a dead person to life, in their time.” But it deserves
notice, that Papias, the very best authority on this subject, is no where quoted as calling this
Philip “an apostle;” though Eusebius, on his own authority, gives this name to the Philip of
whom Papias speaks. It is therefore reasonable to conclude, that this blunder, betraying
such a want of familiarity with the New Testament history, originated after the time of
Papias, whose intimate acquaintance with Philip’s family would have enabled him to say, at
once, that this was the deacon, and not the apostle; though it is not probable that he was
any less deplorably ignorant of the scriptures than most of the Fathers were.
Now what can be said of the testimony of the Fathers on points where they can not refer,
either to their own personal observation, or to informants who have seen and heard what
they testify? The only way in which they can be shielded from the reproach of a gross
blunder and a disgraceful ignorance of the New Testament, is, that they were right in
identifying these two Philips, and that modern theologians are wrong in making the
distinction. On this dilemma I will not pretend to decide; for though so little reverence for the
judgment and information of the Fathers has been shown in this book, there does seem to
me to be some reason for hesitation on this point, where the Fathers ought to have been as
well informed as any body. They must have known surely, whether, according to the notions
of those primitive ages of Christianity, there was any incompatibility between the apostleship
and the deaconship! If their testimony is worth anything on such points, it ought to weigh so
much on this, as to cause a doubt whether they are not right, and the moderns wrong.
However, barely suggesting this query, without attempting a decision, as Luther says, “I will
afford to other and higher spirits, occasion to reflect.”
This is all the satisfaction that the brief records of the inspired or
uninspired historians of Christianity can give the inquirer, on the life
of this apostle;――so unequal were the labors of the first ministers of
Christ, and their claims for notice. Philip, no doubt, served the
purpose for which he was called, faithfully; but in these brief
sketches, there are no traces of any genius of a high character, that
could distinguish him above the thousands that are forgotten, but
whose labors, like those of the minutest animals in a mole-hill,
contribute an indispensable portion to the completion of the mass, in
whose mighty structure all their individual efforts are swallowed up
forever.
NATHANAEL, BAR-THOLOMEW.
his name and call.
A few very brief notices are given of this apostle by John, who
alone alludes to him, otherwise than by a bare mention on the list. It
is mentioned in his gospel that Nathanael was of Cana, in Galilee, a
town which stood about half-way between lake Gennesaret and the
Mediterranean sea; but the circumstances of his call seem to show
that he was then with Philip, probably at or near Bethsaida. Philip,
after being summoned by Jesus to the discipleship, immediately
sought to bring his friend Nathanael into an enjoyment of the honors
of a personal intercourse with Jesus, and invited him to become a
follower of the Messiah, foretold by Moses and the prophets, who
had now appeared, as Jesus of Nazareth, the son of Joseph. On
hearing of that mean place, as the home of the promised King of
Israel, Nathanael, with great scorn, replied, in inquiry, “Can any good
thing come out of Nazareth?” To this sneering question, Philip
answered by the simple proposition, “Come and see;”――wisely
judging that no argument could answer his friend’s prejudice so well
as an actual observation of the character and aspect of the
Nazarene himself. Nathanael, accordingly, persuaded by the
earnestness of his friend, came along with him, perhaps, partly to
gratify him, but, no doubt, with his curiosity somewhat moved to
know what could have thus brought Philip into this devout regard for
a citizen of that dirty little town; and he therefore readily
accompanied him to see what sort of prophet could come out of
Nazareth.
On the day but one after this occurrence, as John records, Jesus
was in Cana of Galilee, the residence of Nathanael, and was present
at a wedding which took place there. From the circumstance that the
mother of Jesus was there also, it would seem likely that it was the
marriage of some of their family friends; otherwise the conjecture
might seem allowable, that the presence of Jesus and his disciples
on this occasion, was in some way connected with the introduction of
Nathanael to Jesus; and that this new disciple may have been some
way concerned in this interesting event. The manner in which the
occurrence is announced,――it being next specified, that two days
after the occurrences recorded in the end of the first chapter, Jesus
was present at a marriage in Cana of Galilee,――would seem to
imply very fairly, that Jesus had been in some other place
immediately before; and it is probable therefore, that he
accompanied Nathanael home from Bethsaida, or whatever place
was the scene of his calling to the discipleship, along with Philip. The
terms of the statement are not, however, absolutely incompatible
with the idea of this first introduction of these two disciples to Jesus,
in Cana itself, which may have been the part of Galilee into which
Jesus is said to have gone forth, after leaving Bethabara; although,
the reasons above given make it probable that Bethsaida was the
scene. After this first incident, no mention whatever is made of
Nathanael, either under his proper name, or his paternal appellation,
except that when the twelve were sent forth in pairs, he was sent
with his friend Philip, that those who had been summoned to the
work together, might now go forth laboring together in this high
commission. One solitary incident is also commemorated by John, in
which this apostle was concerned, namely, the meeting on the lake
of Gennesaret, after the resurrection, where his name is mentioned
among those who went out on the fishing excursion with Peter. His
friend Philip is not there mentioned, but may have been one of the
“two disciples,” who are included without their names being given.
From this trifling circumstance, some have concluded that Nathanael
was a fisherman by trade, as well as the other four who are
mentioned with him; and certainly the conjecture is reasonable, and
not improbable, except from the circumstance, that his residence
was at Cana, which is commonly understood to have been an inland
town, and too far from the water, for any of its inhabitants to follow