Choice via Dual Ownership
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Economics. The day in economics: theory, mechanisms and the evidence beneath them.
Ruth: We're looking at the paper: "Choice via Dual Ownership".
Warren: Here's the gist of the paper. We consider a model of collective choice where for each menu there is an agent who is accountable for the choice made Specifically, we assume that two agents (individuals or selves) allocate choice tasks over different menus using an ownership rule: the owner of the decision over a menu picks what maximizes his own preference and is accountable for it.
Ruth: We characterize three distinct ownership rules under different (nested) choice environments Keywords: menu-dependent choice, collective choice, accountability, bounded rationality.
Warren: Model Formalization The model introduces ownership rules where for each menu attribute the choice to one agent between two; then the owner of the menu makes a decision that maximizes his preference.
Ruth: Formally, an ownership rule (over menus) ω: P(X) → O associates a menu in P(X) to a preference in O.
Paper summary: Ruth: So, to recap, "Choice via Dual Ownership" introduces ownership rules where an agent accountable for a menu makes the choice maximizing their preference. They set up this rule across three different nested environments: minimal preferences function, standard orders function, and rich orders correspondence.
Warren: The core thesis is characterizing the specific conditions a choice function must satisfy to respect these rules in each environment. It’s about finding those necessary and sufficient conditions for choice functions to align with that particular ownership structure.
Iris: The paper emphasizes that the agent doesn't have to be an individual; it could be two selves competing, like present versus future, or two rationales competing over a decision process leading to a choice. That’s what they mean by the owner being accountable for their selection according to their own preference.
Ruth: It moves beyond just looking at single agents making decisions; it incorporates the idea that different entities can hold ownership over different menus based on how those menus are structured. The examples they use, like a child and a parent having preferences about welfare, illustrate this menu-dependent aspect.
Warren: The environments themselves are crucial because they define the assumptions we make about the underlying binary relations—whether we stick to strict preferences or broader orders correspondence. This context is what makes the problem different for each study.
Iris: They focus on revealed preferences, which means they’re looking at what actual choices people make and seeing how those choices reveal something about these ownership rules in these contexts. The paper connects this to earlier work on path independent choice correspondences, like that of Aizerman and Malishevski.
Ruth: It feels like they are providing a comprehensive map for understanding bounded rationality in these collective choice situations by showing exactly what structures the behavior must fit into. It helps us see the constraints imposed by how the problem is set up.
Conclusion: Warren: Thinking about this whole thing, "Choice via Dual Ownership" is really about how the menu structure influences which binary relation we use to decide over a menu, and this paper lays out those rules clearly. The authors explore these three distinct ownership rules across minimal, standard, and rich environments.
Ruth: It’s not just about finding one perfect rule; it’s about understanding that the required conditions change depending on the environment you are in. We see how simple conditions like weak alpha are needed in one setup while a much larger set of requirements appears for another.
Iris: The implication for us is seeing how complex choice behavior can be when you introduce multiple rationales or multiple selves into the decision-making process. It grounds these abstract concepts in testable implications within specific environments, which is what they aim for with their revealed preference characterizations.
Warren: And they point out that while they found some independence between the axioms used for the standard and rich environments, it still shows how sensitive these results are to the exact setup of the problem. It’s not a simple universal solution.
Ruth: The paper provides a framework for analyzing collective choice where accountability is central, showing how bounded rationality manifests depending on what kind of ownership rule is in play. It’s a deep dive into the structure of choice itself.
Université de Montréal · Università Ca’ Foscari Venezia
econ.TH
Submitted: 2026-10-05
Updated: 2026-11-12
License: http://creativecommons.org/licenses/by/4.0/
Importance score: 72/100
The gist: We consider a model of collective choice where for each menu there is an agent who is accountable for the choice made Specifically…
Key concepts
- Ownership Rule
- This rule assigns a specific preference to each menu based on which agent owns the decision. It dictates that the owner of a menu selects the option that maximizes their personal preference, making them responsible for that outcome.
- Minimal Choice Environment
- This environment assumes preferences are strict (asymmetric). In this setting, a monotonic ownership rule partitions menus into upper and downsets. A choice function respects this rule if and only if it satisfies the weak-alpha condition.
- Rich Choice Environment
- This environment allows for set-valued choices (correspondences) instead of strict preferences. Here, Borda ownership is studied, requiring a complex set of conditions (PC, ATT, EFF, etc.) for a choice correspondence to respect it.
Terminology
Summary
We consider a model of collective choice where for each menu there is an agent who is accountable for the choice made Specifically, we assume that two agents (individuals or selves) allocate choice tasks over different menus using an ownership rule: the owner of the decision over a menu picks what maximizes his own preference and is accountable for it. p. 0 We consider a model of collective choice where for each menu there is an agent
who is accountable for the choice made. Specifically, we assume that two agents (individuals
or selves) allocate choice tasks over different menus using an ownership rule: the owner of the
decision over a menu picks what maximizes his own preference and is accountable for it. We
characterize three distinct ownership rules under different (nested) choice environments. Keywords: menu-dependent choice, collective choice, accountability, bounded rationality
JEL Classification Numbers: D10, D71, D90
1 INTRODUCTION
Since the seminal contribution by Kalai et al.Where the paper says it
Where the paper says it
Model Formalization
The model introduces ownership rules where for each menu attribute the choice to one agent between two; then the owner of the menu makes a decision that maximizes his preference. Formally, an ownership rule (over menus) ω: P(X) → O associates a menu in P(X) to a preference in O. p. 3 Formally, consider a pair of preferences O = {≻1,≻2} on X. An ownership rule
(over menus) ω : P(X) → O associates a menu in P(X) to a preference in O. An ownership rule
(over menus) ω : P(X) → O associates a menu in P(X) to a preference in O. The
choice behavior generated via an ownership rule acts as a two-step process: first, it calls
on ω to select which preference to deploy; second, it chooses as if maximizing it. Definition 1. A choice function c on P(X) respects single ownership if there exists a
pair of preferences O = {≻1,≻2} on X and an ownership rule ω : P(X) → O such
that
c(A) = max(A,ω(A))
for any A ∈ P(X).Where the paper says it
Where the paper says it
Where the paper says it
Choice Environments
The paper operationalizes ownership rules over two binary relations in three different (nested) choice environments commonly studied in abstract choice theory. Table 1 labels the three nested environments and summarizes their assumptions on the two binary relations and on the observability of choice. p. 6 Table 1 labels the three nested environments and summarizes their assumptions on
the two binary relations and on the observability of choice. binary relations choice map
Environment ≻1,≻2 c(·)
1.Where the paper says it
-
Minimal preferences function assumes that c(·) is a choice function and that the two binary relations ≻1,≻2 are preferences (i. p. 6
e., asymmetric). p. 3Where the paper says it
Rich orders correspondence Table 1: Three choice environments 3. 1 MINIMAL CHOICE ENVIRONMENT The minimal choice environment assumes that c(·) is a choice function and that the two binary relations ≻1,≻2 are preferences (i.
Where the paper says it
Following Cherepanov et al. [2013], a preference is an asymmetric binary relation ≻ on X; and a connected4 , transitive preference is an order.
-
Standard orders function assumes orders instead of preferences but keeps a single-valued choice function. p. 7
Where the paper says it
This fits well with the parent-child paradigm, where one expects the first to exhibit more rationality than the second: the choices of the parent, who decides over larger menus, are rationalized by an order while the choices of the child over smaller menus by a preference. 2 STANDARD CHOICE ENVIRONMENT We now turn to a standard choice environment, assuming orders instead of preferences but keeping a single-valued choice function.
-
Rich orders correspondence assumes orders but allows set-valued choices, so that c is now a correspondence. p. 8
Where the paper says it
This special case can be characterized similarly to the above: one has to drop Condition 2 and require each menu Ai to have size at least 3 in the definition of prevalence. 3 RICH CHOICE ENVIRONMENT The rich choice environment assumes orders but allows set-valued choices, so that c is now a correspondence.
Characterizations of Choice Behavior
The paper provides revealed preference characterizations for each environment.
3.1 Minimal Choice Environment studies the testable implications of a monotonic ownership rule in this minimal choice environment. p. 6 This last assumption is
weaker than usual, but for instance it features prominently in the model of choice by
sequential application of multiple criteria studied by Manzini and Mariotti [2007]. We study the testable implications of a monotonic ownership rule in this minimal
choice environment. An ownership rule is monotone if, for any three menus A ⊂ B ⊂C, ω(A) ̸=
ω(B) implies ω(B) = ω(C). A monotonic ownership rule partitions the set of menus into two blocks, respectively called upper set and downset. Our first result shows that Weak-α characterizes monotone
choice functions in the minimal environment. A choice function respects monotone ownership in a minimal environment
if and only if it satisfies weak-α.Where the paper says it
Where the paper says it
Where the paper says it
3.2 Standard Choice Environment characterizes choice behavior via simple ownership in a standard environment. p. 7 Formally, we say that the
ownership rule is simple if ω(A) = ω(B) for all menus A,B of size 2. We characterize choice behavior via simple ownership in a standard environment. The
independence of the two axioms is shown in Appendix B. A choice function respects simple ownership in a standard environment if
and only if it satisfies PC and AP.Where the paper says it
Where the paper says it
3.3 Rich Choice Environment characterizes choice behavior via Borda ownership in a rich environment. p. 9 Borda ownership, instead, fits a requirement of accountability: given that
someone has to be responsible for making a choice, it assigns ownership to minimize
the non-owner’s discontent. We characterize choice behavior via Borda ownership in a rich environment. 3 RICH CHOICE ENVIRONMENT
The rich choice environment assumes orders but allows set-valued choices, so that c is
now a correspondence. In a simple case like the “top and top” procedure, the (joint)
ownership rule is a constant function mapping each menu to the pair {≻1,≻2}.Where the paper says it
Where the paper says it
Independence of Axioms
The paper collects examples that show the independence of axioms used in Theorem 2 and in Theorem 3. For instance, B. p. 3 Many examples in the introduction can be formalized by suitable restrictions on
the ownership rule. For instance, consider a situation where two agents (individuals or
selves) assign choice ownership according to the size of the menu: one agent handles
4 A binary relation ≻ is connected if, for any x ̸= y, either y ≻ x or x ≻ y. Our next result characterizes choice behavior under a simple ownership rule. The
independence of the two axioms is shown in Appendix B.Where the paper says it
Where the paper says it
Conclusion and Future Directions
The notion of an ownership rule models how the structure of a choice problem affects which binary relation is used to decide over a menu We conjecture that the general case of n binary relations is much harder to tackle. Future research directions include studying ownership rules in structured settings such as consumer choice or choice under uncertainty. p. 1 future) competing to own a consumption/saving
decision, or there may be two rationales competing to own the process leading to a
choice. Therefore, we say simply that an ownership rule assigns a preference to each
menu, and leave it unspecified to whom or what the preference refers. Our general notion of an ownership rule provides a unified way to study such choice procedures. Manzini and Mariotti
[2018] studies the Dual Random Utility model (D-RUM), where the stochastic choice
function is generated by probability distribution(s) over two linear orders. Their menudependent version of the D-RUM can be seen as a stochastic analogue of an ownership
rule. In this paper, we operationalize ownership rules over two binary relations in three
different (nested) choice environments, commonly studied in abstract choice theory.Where the paper says it
Where the paper says it
Where the paper says it
Proofs and Results
The proofs involve complex inductive constructions to show that the defined orders rationalize the choice correspondence under Borda ownership. The proof of Theorem 3 involves an inductive construction using a lexicographic order for adding alternatives. p. 2 More recently, Eliaz et al. [2011] and
De Clippel and Eliaz [2012] have studied a few important cases where at most two
alternatives are chosen using one or two linear orders. As the examples suggest, sometimes the context itself is a source of bounded rationality because some features (e. , size, domain,
complexity) interfere with maximizing a single preference over all choice problems. Because
a violation reveals an ownership change, monotonicity requires no further ownership
flips along the same chain. Our first result shows that Weak-α characterizes monotone
choice functions in the minimal environment.Where the paper says it
Where the paper says it
Where the paper says it
Summary of Key Concepts
The core concept is the ownership rule which assigns a preference to each menu. The paper distinguishes between single ownership, joint ownership, and general ownership based on how the decision is imputed to an agent or set of agents. p. 0 We consider a model of collective choice where for each menu there is an agent
who is accountable for the choice made. Specifically, we assume that two agents (individuals
or selves) allocate choice tasks over different menus using an ownership rule: the owner of the
decision over a menu picks what maximizes his own preference and is accountable for it. These environments differ in the rationality of the conditions on the underlying binary
relations (asymmetry, transitivity, etc. ) and in their observability requirements (singleor multi-valued choice). In
the spirit of bargaining, the fallback solution seeks a compromise that takes everyone
into account. Borda ownership, instead, fits a requirement of accountability: given that
someone has to be responsible for making a choice, it assigns ownership to minimize
the non-owner’s discontent. In this paper, we operationalize ownership rules over two binary relations in three
different (nested) choice environments, commonly studied in abstract choice theory. These environments differ in the rationality of the conditions on the underlying binary
relations (asymmetry, transitivity, etc.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
References
AIZERMAN, M. A. AND A. V. MALISHEVSKI (1981): “General Theory of Best Variants Choice: Some Aspects,” IEEE Transactions on Automatic Control, 26, 1030–1040. p. 2 The ownership rules introduced in this paper model precisely such menu
dependence, and thus can be seen as a next step from their work. Our results also relate to early work by Aizerman and Malishevski [1981], who
show that path independent choice correspondences are those generated by selecting
the maximizers of a fixed set of linear orders.Where the paper says it
AMBRUS, A. AND K. ROZEN (2015): “Rationalising choice with multi-self models,” The Economic Journal, 125, 1136–1156. p. 1 One may imagine a two-step choice procedure, where the agent facing the
menu activates a binary relation and then maximizes that binary relation to make her
choice. This sort of menu-dependent dictatorship is very different from the aggregation
of binary relations across all menus studied in Ambrus and Rozen [2015].Where the paper says it
APESTEGUIA, J. p. 2 We conjecture that the general case of n binary relations is much
harder to tackle. Indeed, Apesteguia and Ballester [2010] and Demuynck [2011] show
that the general problem of rationalizing choice behavior is NP-complete, whereas the
main case of polynomial computability (for a specific model of sequential rationalization) in Demuynck [2011, Theorem 3] requires at most two binary relations. We conjecture that the general case of n binary relations is much
harder to tackle. Indeed, Apesteguia and Ballester [2010] and Demuynck [2011] show
that the general problem of rationalizing choice behavior is NP-complete, whereas the
main case of polynomial computability (for a specific model of sequential rationalization) in Demuynck [2011, Theorem 3] requires at most two binary relations. We conjecture that the general case of n binary relations is much
harder to tackle. Indeed, Apesteguia and Ballester [2010] and Demuynck [2011] show
that the general problem of rationalizing choice behavior is NP-complete, whereas the
main case of polynomial computability (for a specific model of sequential rationalization) in Demuynck [2011, Theorem 3] requires at most two binary relations.Where the paper says it
Where the paper says it
Where the paper says it
APESTEGUIA, J., M. A. BALLESTER, AND J. p. 2 We conjecture that the general case of n binary relations is much
harder to tackle. Indeed, Apesteguia and Ballester [2010] and Demuynck [2011] show
that the general problem of rationalizing choice behavior is NP-complete, whereas the
main case of polynomial computability (for a specific model of sequential rationalization) in Demuynck [2011, Theorem 3] requires at most two binary relations. Our general notion of an ownership rule provides a unified way to study such choice procedures. Manzini and Mariotti
[2018] studies the Dual Random Utility model (D-RUM), where the stochastic choice
function is generated by probability distribution(s) over two linear orders.Where the paper says it
Where the paper says it
AU, P. H. AND K. KAWAI (2011): “Sequentially rationalizable choice with transitive rationales,” Games and Economic Behavior, 73, 608–614. p. 1 It also departs from the sequential application of shortlisting and choosing initiated by Manzini
and Mariotti [2007], which has spawned a large literature (e. , Au and Kawai [2011],
3 For an interesting variant, focus on the parent’s welfare and swap the child with a piece of GenAI.Where the paper says it
CHEREPANOV, V. p. 0 , 2002, Eliaz et al. , 2011]; shortlisting [Cherepanov et al. We conjecture that the general case of n binary relations is much
harder to tackle. Indeed, Apesteguia and Ballester [2010] and Demuynck [2011] show
that the general problem of rationalizing choice behavior is NP-complete, whereas the
main case of polynomial computability (for a specific model of sequential rationalization) in Demuynck [2011, Theorem 3] requires at most two binary relations.Where the paper says it
Where the paper says it
CHEUNG, P. H. AND Y. MASATLIOGLU (2025): “Frame-dependent random utility,” arXiv:2502. p. 2 Our general notion of an ownership rule provides a unified way to study such choice procedures. Manzini and Mariotti
[2018] studies the Dual Random Utility model (D-RUM), where the stochastic choice
function is generated by probability distribution(s) over two linear orders.Where the paper says it
DE CLIPPEL, G. p. 0 , 2011]; shortlisting [Cherepanov et al. , 2013]; aggregation [Ambrus and Rozen, 2015]; and bargaining between two agents [De Clippel and Eliaz, 2012]. Compare Borda ownership with the fallback bargaining solution: both describe procedures to make a choice when there are two agents with potentially different orders. In
the spirit of bargaining, the fallback solution seeks a compromise that takes everyone
into account.Where the paper says it
Where the paper says it
DEMUYNCK, T. p. 2 We conjecture that the general case of n binary relations is much
harder to tackle. Indeed, Apesteguia and Ballester [2010] and Demuynck [2011] show
that the general problem of rationalizing choice behavior is NP-complete, whereas the
main case of polynomial computability (for a specific model of sequential rationalization) in Demuynck [2011, Theorem 3] requires at most two binary relations. We conjecture that the general case of n binary relations is much
harder to tackle. Indeed, Apesteguia and Ballester [2010] and Demuynck [2011] show
that the general problem of rationalizing choice behavior is NP-complete, whereas the
main case of polynomial computability (for a specific model of sequential rationalization) in Demuynck [2011, Theorem 3] requires at most two binary relations.Where the paper says it
Where the paper says it
ELIAZ, K., M. RICHTER, AND A. RUBINSTEIN (2011): “Choosing the two finalists,” Economic Theory, 46, 211–219. p. 1 It also departs from the sequential application of shortlisting and choosing initiated by Manzini
and Mariotti [2007], which has spawned a large literature (e. , Au and Kawai [2011],
3 For an interesting variant, focus on the parent’s welfare and swap the child with a piece of GenAI.Where the paper says it
GARICANO, L. p. 1 When the menu has only low stakes, the child owns the decision and gets to choose
his favorite item; when high stakes are present, the parent overrides her child and the
decision is made using her preferences. 3 Somewhat similarly, in the model of knowledge hierarchies by Garicano [2000], standard problems are solved by subordinates,
and exceptional problems by their supervisors. 3 Somewhat similarly, in the model of knowledge hierarchies by Garicano [2000], standard problems are solved by subordinates,
and exceptional problems by their supervisors. An M-form organization is arranged into distinct, semi-autonomous units,
often called divisions.Where the paper says it
Where the paper says it
HORAN, S.: “A simple model of two-stage choice,” Journal of Economic Theory, 162, 372–406. p. 2 Their menudependent version of the D-RUM can be seen as a stochastic analogue of an ownership
rule. In this paper, we operationalize ownership rules over two binary relations in three
different (nested) choice environments, commonly studied in abstract choice theory.Where the paper says it
HUBER, J., J. W. PAYNE, AND C. PUTO (1982): “Adding asymmetrically dominated alternatives: Violations of regularity and the similarity hypothesis,” Journal of Consumer Research, 9, 90–98.
KALAI, G., A. RUBINSTEIN, AND R. SPIEGLER (2002): “Rationalizing choice functions by multiple rationales,” Econometrica, 70, 2481–2488. p. 0 [2002], it has been widely known that
the presence of multiple binary relations may explain choice behavior that departs from
classical rationality. This multiplicity may refer to selves, individuals and, more generally, distinct criteria or rationales by which a choice may be made.Where the paper says it
LUCE, R. D. AND H. RAIFFA (1957): Games and Decisions: Introduction and Critical Survey, New York: Wiley. p. 5 some alternative from A
∗
is available. Primacy generalizes the revisitation of the dinner
example from Luce and Raiffa [1957] in Kalai et al.Where the paper says it
MANZINI, P. p. 0 Financial support under PRIN 20222Z3CR7
“Nudging under limited attention” is acknowledged. 1These include: sequential applications [Manzini and Mariotti, 2007, Horan, 2016]; multiple maximizations [Kalai et al. , asymmetric). This last assumption is
weaker than usual, but for instance it features prominently in the model of choice by
sequential application of multiple criteria studied by Manzini and Mariotti [2007].Where the paper says it
Where the paper says it
——— (2012): “Categorize then choose: Boundedly rational choice and welfare,” Journal of the European Economic Association, 10, 1141–1165. p. 1 A child and a parent have their own preferences about the child’s welfare. When the menu has only low stakes, the child owns the decision and gets to choose
his favorite item; when high stakes are present, the parent overrides her child and the
decision is made using her preferences.Where the paper says it
MASATLIOGLU, Y. p. 2 Masatlioglu et al. [2012], Manzini and Mariotti [2012], Horan [2016]). *We thank Franz Dietrich, Federico Echenique, Sean Horan, Marco Mariotti, Alessandro Pavan, John
Quah, and seminar audiences for their useful comments. Financial support under PRIN 20222Z3CR7
“Nudging under limited attention” is acknowledged.Where the paper says it
Where the paper says it
PINA POLO, F. p. 0 For a prominent example, consider how
joint leadership by the Roman consuls worked. Among many provisions to promote
collegiality (collegialitas) and parity (par potestas), the consuls were assigned2 distinct
spheres of responsibility (provinciæ); see Pina Polo [2011]. Among many provisions to promote
collegiality (collegialitas) and parity (par potestas), the consuls were assigned2 distinct
spheres of responsibility (provinciæ); see Pina Polo [2011]. Intuitively, each consul
was responsible for handling the choice problems within his remit or, in our jargon,
owned the decisions for his provincia.Where the paper says it
Where the paper says it
RUBINSTEIN, A.: Lecture Notes in Microeconomic Theory, Princeton: Prince30 ton University Press, 2 ed. p. 2 [2012], Manzini and Mariotti [2012], Horan [2016]). Our objective is to study the observable implications of menu-dependent ownership
in the setup of abstract choice theory.Where the paper says it
SZPILRAJN, E.: “Sur l’extension de l’ordre partiel,” Fundamenta Mathematicae, 16, 386–389.
TYSON, C. J.: “Cognitive constraints, contraction consistency, and the satisficing criterion,” Journal of Economic Theory, 138, 51–70. p. 6 Our goal is to characterize the choice functions generated by a monotone ownership
rule. It is well known that contraction consistency (or Sen’s condition α) is a necessary
condition for the rationalizability of a choice function.Where the paper says it
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Implement a choice function that respects single ownership by first calling on an ownership rule omega to select a preference, and then choosing as if maximizing that preference, as defined in Definition 1:
c(A) = max(A,ω(A)) for any A ∈ P(X).
-
Develop systems capable of handling joint ownership where the choice correspondence is defined by the union of maximizers from two preferences, as described in Definition 2:
c(A) = [≻i∈ω(A) max(A,≻i) for any A ∈ P(X).
-
Design AI models that operate under a minimal choice environment by satisfying the Weak-alpha condition:
For all menus B ⊂ B′ ⊆ A ⊂ A′, if c(A′) ∈ A and c(A) ≠ c(A′), then c(B′) ∈ B implies c(B) = c(B′).
-
Create decision systems for standard choice environments that adhere to the simple ownership rule by satisfying Pairwise Consistency (PC):
If c(x,y) = x and c(y,z) = y, then c(x,z) = x.
-
Build complex choice systems for rich environments utilizing Borda ownership by satisfying a set of seven axioms:
PC, ATT, EFF, EI, NBI, SI and EXP.
-
Integrate a mechanism to verify the rationalizability of a choice correspondence under Borda ownership by checking satisfaction of the required axioms as outlined in Theorem 3.
Abstract
When there are multiple agents (or rationales), an ownership rule assigns the choice for each menu to one of them. We study choice under dual ownership, where decision control alternates between two preferences. The main contribution is the behavioral characterization of a rule that endogenously decides ownership to maximize consensus. We also axiomatize two simpler exogenous rules. Identification results are provided for each rule.