Search Theory-Summary

You might also like

Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 19

SEARCH

THEORY-SUMMARY
Madhurima Das
 Following the footsteps of Herbert Simon(1955,1984,197,2000), a brief idea of limited cognition and
cognitive ability is observed, hence the paper by Caplin and Dean(2011), moves from decision
theory/choice theory to search theory, where the choices of individuals depend on the information
availability as well identification of the object that he/she is about to choose. Hence we move from
complete information to incomplete information. Search theory depicts how the individual chooses a
specific object(s) based on information, identification, and complete search.
 What are two theories developed here?

 ABS, alternative based search and RBS, reservation based search which is a refinement on ABS. to
understand these models experimental data is collected instead of standard data, these data depends on an
experimental framework where individuals are given unlimited choices with limited time, and
researchers observe the changes in choices as per the contemplation of time.
 Alternative Based model:

 Assumptions-

 Decision maker(DM) chooses from a set of objects sequentially

 There is no search order

 Every item searched is compared to a fixed utility function.

 Behavioural pattern is not considered initially

 Preference means the DM is unaware of other alternatives.

 The choice process should be consistent, and acyclic(doesn’t show a cyclical pattern of choices).
 For obtaining revealed preference information on final choice, search should be complete.

 ‘List order’ of objects are provided that are observable to outsiders as well(those who are not
included in the experiment).

 Final choice is obtained through a choice procedure, ‘list order’ is used to break ties.

 Sequentially every object in set is searched, this search process is inclusive of recall
factor(repeating the elements)where DM contemplates on the ever-expanding choices of objects
hence the search correspondence, S, is non-decreasing.

 DM chooses best object out of identified objects.

 Any chosen object from given choice set, is utility maximizing in nature.
 Theory:

 Stochasticity helps in developing the model with technically sound statements, RBS is an extended part of ABS model.

 The unified approach to theory and experiment rests on two key premises.

 Premise 1. ABS and RBS represent broad styles of search that may be undertaken in a wide variety of different decision
making environments.

 Premise 2. It is conceptually and experimentally feasible to collect data on the evolution of “intended” choice with
contemplation time.

 ABS is more of a natural process of selection, in any environment particularly when there is a cost of switching attention
from one alternative to another, or if items can be understood only in their entirety(that means if there’s complete
information about the item, is available through repetitive search.)

 The enriched data set used here is called choice process data, Rather than record only the alternative that is finally chosen
by the DM, choice process data track how choice evolves with contemplation time.

 Premises the theory is built on are with defined stochastic choice processes X, and defined utility function.
 Let us trace the premises one by one;

1.  
 This explains that the deterministic choice process of the DM , helps him/her to search all the objects as per
contemplation of time from set A, where A belongs to X which are possible alternatives.
1. The choice process (X,C) is entitled with utility and search correspondence (u,S), which is the ABS
representation where the individual faces non-decreasing search sequences and maximizes his/her utility from
the choice process, C which is based on set A has been contemplated for time, t.

CA(t) = argmax u(x), x belongs to S A(t).

 Note:

 Firstly, the DM always maximizes his/her utility from set of choices and has a fixed utility function perceived to
compare with the searched objects.

 Secondly, choices remain non-decreasing and are recalled as well, so the sets are either fully searched or not
searched.

 Thirdly, DM’s preferences weakly improve over time.

 Fourthly, restricting or limiting behavior not observed.


 ABS shows notions of revealed preference in the final choice. Now, if behavioral patterns are considered
say Bi, then Bi is a proper subset of A, so sequential search would lead to the final choice.

1. There could be symmetric binary relation where preference and indifference can be observed from the
search sequences given t>=1 such that if x is found t+sth time, and y should be found in tth time so that
both objects are completely searched and has equal probability of being selected in the final
consumption bundle.
2. Lemma 1, this one deals with one pre-requisite: reveal preference information should be acyclic in
nature. This lemma states if x is P to y then u(x) > u(y) depicts SARP, while if both are chosen the x I y,
u(x)=u(y) which is the first theorem itself, where we can see that if u(x) is maximized then u(y) is not
but the set is thoroughly searched and y is rejected based on preference. Conversely, as search
correspondence keeps on expanding, if both x and y are identified and y is preferred over x at time t+s,
while y and x were indifferent and generated same level of utility in time t, then it can be said that yPIx
and satisfies OWC as this change in utility explains about choice transformation over time.
 Times when ABS doesn’t hold:
 RBS- reservation based search:

 RBS is a refinement over ABS that explains about stopping rule which is based on the reservation utility level.

 Assumptions-

 any set with below reservation level utility should be searched completely.
 Final choices contains information about revealed preference.
 Reservation utility is ρ
 X ρu = the set of choices with above reservation utility, X/ X ρu= objects that are searched but has low ρ.

 Here limitation is that, there should be no items with higher reservation utility, is left to be searched, so that
utility is maximized in final choice of DM.
 No restrictions on behavioral patterns
 R1 and R2(given below) can satisfy individually.
 ABS leads to RBS if the sets are fully searched.
 How to prove RBS exists?

 R1. If A Ո X ρu= Փ that means no search correspondence is left within choice process A, that has object
with above reservation level of utility.

 R2. If A Ս X ρu = not null, then search continues until object(s) with above reservation level utility is
obtain.
 Condition to get only ABS but not RBS:
 Conditions to attain RBS:

 For any set D belonging to X, there are binary relations that explains ABS+RBS, given D is set with
below reservation utility.
 Hence this explains that a. DM stops searching if reservation utility is crossed by the chosen object. b.
DM keeps searching if reservation levels are not crossed. These sets are called as direct non-terminal
sets, XN.

 To follow up with binary relations as well as OWC, the search correspondence trace both direct non-
terminal sets and indirectly non-terminal sets XIN, then the theorem 2, leads to the stochastic models.
 Stochastic models:

 ABS-

 Search orders are non-observable.


 Allowing stochasticity(randomness) leads to a choice that generates maximum utility against a stochastic search sequence.
 Probability space considered here is Z.
 Sets are open sets or cylinder sets where additional objects are introduced over time and the topology follows σ-algebra.
 Every cylinder sets has two probabilities – selected or not selected, hence probability space values at 2n .

 Stochastic choice process or SCP comprises finite set of choices, where search set never shrinks and there are non-
decreasing elements where utility maximizes at the final selection.

 ABS representation is similar to RBS if we consider stochastic process and deterministic process are same, with joint and
replacement sets between x and y, say, Jxy and Rxy where strict indifference of x to y and strict preference from x to y is
observed, respectively.

 Again binary relation between search process of alternative cylinder sets, should follow ABS relations of preference,
indifference and OWC.
 RBS:

 The measures here followed with tracing both directly non-terminal and indirectly non-terminal sets
which includes , in such a way that both x and y are identified in different search sequences. Hence
complete search and binary relation is traced among all the sets with utility level below reservation and
above reservation. Final choices could be from any set and would be maximizing levels of utility of the
DM as per preference and behavioral interpretation.

 Note:

 Here the sets are constructed in two ways, compression and decompression. Compressed sets basically
delays the switching process from choosing one element to the other by removing all constancy and
repetitive elements making the choice process finite and limited, so ABS and RBS jointly survives under
compression.

 While decompression induces recall, and includes all the repetitive elements. That’s stated by theorem 3
and 4 that implies, search should stop within a limited and finite time period.
 RBS and non-standard behavior:

a. Framing effect- the search process depends on search cost, if search cost is high then low reservation
utility and likewise if search cost is low, then high reservation utility level. These will create an impact
on preferences, here if is SCP on (Z,F) then binary relation follows OWC, also it is found that from a
family of reservation utility the final choice is ‘less directly’ observed given the choice process data. So
basically framing effects explains about cognitive bias a DM faces while choosing an object presented
with negative or positive connotations.
b. Status quo- it deals with DM’s choice with highest utility object from the given set, but if DM chooses
from below reservation set, unless the object chosen has the highest utility within that set, the DM will
not chose the object. Status quo has 2 effects, either object is chosen or status quo will not impact on
DM’s choice.
c. Stochastic choice- the given choice set will be divided in two parts, below reservation set & above
reservation set, search orders will be equally probable of getting selected from above reservation set &
final choice will be randomly selected given the stochasticity of choice.
 Conclusion:

 Incomplete information explains about why at times individual does not choose utility maximizing
objects, the experimental design identified the cognitive aspects of one’s choice process.

 Queries that were not pointed out:

1. Can an incentive impact the choice preferences as compared to initial preference? (I can prefer a coke,
but I will prefer it more if I get an additional offer of getting a pack of chips and dips with it.)
2. How does the external environment impacts cognition?(in world cup season I prefer more snacks than
compared to entire year)
3. There may not be search orders initially but in the long run needs based search can there(as per my
need I will buy the products). These were addressed by Azzopardi(2021), where takes up these
questions and answers through theoretical framework of cognitive bias and information processing by
individuals or DM’s.
This Photo by Unknown Author is licensed under CC BY-NC

You might also like