A choice-based axiomatization of Nash equilibrium

arXiv:2512.03930 · econ.TH, cs.GT · Submitted 2025-12-03 · Read on arXiv

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: "A choice-based axiomatization of Nash equilibrium".

Warren: Here's the gist of the paper. An axiomatic characterization of Nash equilibrium is provided for games in normal form, showing that the Nash equilibrium correspondence is fully characterized by four simple and intuitive axiomsThe Nash equilibrium correspondence is shown to be fully characterized by four simple and intuitive axiomsThe Nash equilibrium correspondence is shown to be fully characterized by four simple and intuitive axioms>.

Ruth: Axiomatic Characterization The paper presents a new axiomatization of Nash equilibrium based on four simple and intuitive axioms that apply to games in normal formThe present paper contributes to this active research stream by offering a new axiomatization of Nash equilibrium.

Warren: The axiomatization applies to Nash equilibria in pure and mixed strategies alike, to games with strategy sets of any cardinality, and it does not require that players’ preferences have a utility or expected utility representation.

Ruth: The Four Core Axioms The four axioms introduced are Independence of irrelevant strategies (IIS), Merging consistency (MC), Invariance to strictly dominated strategies (ISDS), and Joint optimality (JO)The first axiom is called independence of irrelevant strategies (IIS) and says that if a strategy profile belonging to both G and G′ is a solution to G, then that strategy profile is a solution to the “smaller” game G′ too.

Paper summary: Ruth: So we're looking at this paper, "A choice-based axiomatization of Nash equilibrium," and the main point here is they've given a set of four simple axioms that fully characterize what a Nash equilibrium actually is in normal form games. Warren That sounds like a big claim for something as complex as strategy profiles.

Iris: The authors are Michele Crescenzi, and they’re saying this axiomatization applies to Nash equilibria in both pure and mixed strategies, across any game with any number of players or strategy sets, and it doesn't even require players to have a utility representation.

Ruth: That is pretty broad scope. What they claim is that if you have a solution concept that satisfies these four axioms—Independence of irrelevant strategies, Merging consistency, Invariance to strictly dominated strategies, and Joint optimality—then that concept must be a weak refinement of Nash equilibrium.

Warren: So they're not just defining Nash equilibrium; they’re showing what makes it unique by using these properties. They introduce IIS as independence of irrelevant strategies and MC as merging consistency.

Iris: The paper points out that IIS is related to contraction consistency from abstract choice theory, while MC is the counterpart about expansion consistency, which mirrors something called Sen’s property gamma from one thousand nine hundred seventy-one <ref:2512.03930#pg1>. Ruth That tells us they are drawing on established ideas but framing them in a new way for non-cooperative games.

Warren: And what's interesting is that they show these four axioms are logically independent of each other through specific examples where some concepts satisfy them but not all of the others.

Iris: They also noted that while consistency, which is similar to IIS, isn't necessarily implied by IIS, and merging consistency doesn't imply converse independence of irrelevant strategies. Ruth It’s like they are mapping out the relationships between these different solution concepts in a way that shows they aren't all just variations on the same theme.

Warren: The main result is Theorem one which states that a solution concept satisfies IIS, MC, ISDS and JO if and only if it satisfies phi G equals phi NE G for every game G in their d-closed class <ref:2512.03930#pg1>.

Ruth: So the central claim is that Nash equilibrium correspondence is fully characterized by these four axioms. If you have a solution concept satisfying those four properties on a d-closed class of games, then it has to be a weak refinement of Nash equilibrium.

Warren: The paper says they proved that if you satisfy IIS, ISDS, and JO on this d-closed class, it must be a weak refinement of Nash equilibrium because any solution concept satisfying those three properties has that property.

Iris: They also pointed out the domain of this axiomatization is any d-closed class of games. But they did flag a limitation: if you only look at one non-trivial game, it isn't d-closed, and even mixed strategy extensions of finite games might not be d-closed, meaning a concept satisfying these four axioms might miss some mixed strategy Nash equilibria for a finite game.

Ruth: It’s important to remember that the set of games they are talking about needs to be quite rich for this full characterization to hold. That's where the requirement of a d-closed class comes in, which restricts how many things you can do with the concept.

Warren: They also compared their axioms with other concepts like consistency and converse consistency, showing that IIS and consistency are logically independent of each other. Iris It seems they are trying to show this is a fresh way to approach the problem because previous work focused on just one or two of these axioms in isolation.

Ruth: The implication for us is that we have a new lens through which to view solution concepts—not just as specific algorithms, but as satisfying certain intuitive properties about how players make choices. Warren So, if you’re thinking about designing a game where you want to guarantee some sort of equilibrium property, these four axioms give you a solid checklist.

Iris: And for the one-player games, they simplified things by dropping the merging consistency axiom because joint optimality alone implies that the solution concept is contained within the set of Nash equilibria. Ruth That’s a nice simplification for those simpler cases, showing how different axioms can sometimes be redundant depending on what you're looking at.

Warren: Overall, this paper "A choice-based axiomatization of Nash equilibrium" provides a new foundational structure for understanding Nash equilibrium through these four core axioms.

Ruth: So we're done with this piece. It’s about giving us a systematic way to check if a solution concept is indeed behaving like Nash equilibrium by looking at independence, merging, dominance invariance, and joint optimality.

Conclusion: Ruth: So we've been looking at this paper by Crescenzi on "A choice-based axiomatization of Nash equilibrium." The main thing here is they’re trying to define what Nash equilibrium actually is using just four simple rules.

Warren: Four rules? That sounds like a lot of work for something so fundamental to how markets or any game works. What are these rules they’ve picked up?

Iris: They call them Independence of irrelevant strategies, Merging consistency, Invariance to strictly dominated strategies, and Joint optimality. They apply this framework to games with any number of players and strategy sets without needing a utility representation for the players.

Ruth: That’s a big assumption—that you don't need to know what people *think* about their payoffs just to define an equilibrium concept. What are they actually doing with that in terms of testing these rules?

Warren: They show that if you have a solution concept that passes all four tests on a certain class of games, then it has to be a weak refinement of the true Nash equilibrium correspondence. That’s the core result, proving these four axioms fully characterize it under specific conditions.

Iris: The caveat is that this works only if you’re looking at a d-closed class of games, and they admit that for some situations, like just one game, that class isn't closed. So the scope is quite restricted.

Ruth: And what does this mean practically for us when we're trying to model real-world decisions? Does it give us a better way to build predictive models or something?

Warren: It gives you a formal structure. Instead of just picking one equilibrium concept, you have this checklist of four properties you need to satisfy. It helps define exactly what kind of behavior is truly an equilibrium outcome versus just some random choice.

Iris: The interesting part for identification is how they compare these rules to other concepts like consistency or converse consistency, showing they are logically independent in some ways. That means you can test them separately without them automatically implying each other.

Ruth: So the title itself suggests this isn't just a rehash of old theory, but a new way of building the definition from the ground up using these choice-based axioms.

Warren: Yeah, it’s about moving away from needing complex utility functions and instead focusing on these structural properties of strategy sets. It opens up a different kind of inquiry into equilibrium concepts.

Michele Crescenzi

University of Helsinki and Helsinki Graduate School of Economics

econ.TH, cs.GT

Submitted: 2025-12-03

Updated: 2026-11-15

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 83/100

The gist: An axiomatic characterization of Nash equilibrium is provided for games in normal form, showing that the Nash equilibrium correspondence is fully characterized by four simple and intuitive axiomsThe…

Key concepts

Independence of irrelevant strategies (IIS)
This axiom states that if a strategy profile solves a game G and is also a solution to a smaller game G', then it must also be a solution to G'. It ensures that solutions are not overly dependent on irrelevant strategies in larger games.
Merging consistency (MC)
This axiom requires that if a strategy profile solves two games, G' and G'', it must also solve the larger game formed by merging them. It addresses how solutions behave when considering combined strategic scenarios.
Invariance to strictly dominated strategies (ISDS)
This axiom posits that removing any strictly dominated strategies from a game does not change the set of solutions. If a game G' is formed by removing such strategies from G, then G and G' share the exact same solution set.
Joint optimality (JO)
This axiom requires that if every player chooses their strategy independently to maximize their own preferences without regard for others, the resulting strategy profile is a Nash equilibrium. It links individual maximization to collective equilibrium.

Terminology

Summary

An axiomatic characterization of Nash equilibrium is provided for games in normal form, showing that the Nash equilibrium correspondence is fully characterized by four simple and intuitive axiomsThe Nash equilibrium correspondence is shown to be fully characterized by four simple and intuitive axiomsThe Nash equilibrium correspondence is shown to be fully characterized by four simple and intuitive axioms>. p. 0

Axiomatic Characterization

The paper presents a new axiomatization of Nash equilibrium based on four simple and intuitive axioms that apply to games in normal formThe present paper contributes to this active research stream by offering a new axiomatization of Nash equilibrium. The axiomatization applies to Nash equilibria in pure and mixed strategies alike, to games with strategy sets of any cardinality, and it does not require that players’ preferences have a utility or expected utility representation. p. 0

The Four Core Axioms

The four axioms introduced are Independence of irrelevant strategies (IIS), Merging consistency (MC), Invariance to strictly dominated strategies (ISDS), and Joint optimality (JO)The first axiom is called independence of irrelevant strategies (IIS) and says that if a strategy profile belonging to both G and G′ is a solution to G, then that strategy profile is a solution to the “smaller” game G′ too. The second axiom, merging consistency (MC) axiom posits that any strategy profile that solves both G′ and G′′ must solve the “larger” game G as well. p. 1 The third axiom, invariance to strictly dominated strategies (ISDS), postulates that whenever a game G′ can be formed by removing only strictly dominated strategies from another game G, then G and G′ have the very same set of solutions. p. 1 The last axiom is called joint optimality (JO), which says that if every player in a given game has a strategy that maximizes her preferences irrespective of what the other players are going to choose, then the strategy profile consisting of all those preference-maximizing strategies is a solution to the game. p. 1

Logical Relationships and Scope

IIS and MC are adaptations to non-cooperative games of well-known axioms from abstract choice theory, specifically IIS is a contraction consistency property mirroring Sen’s property α (Sen, 1971), while MC is about expansion consistency and is the counterpart of Sen’s property γ (Sen, 1971). The paper shows that IIS is logically independent of consistency principles like consistency and converse consistency. p. 2 Furthermore, the four axioms are shown to be logically independent through examples where solution concepts satisfy some but not all of themThe examples in this subsection show that the four axioms introduced in Section 3. p. 81 are logically independent. p. 2

Characterization Results

Theorem 1 states that a solution concept on a d-closed class of games satisfies IIS, MC, ISDS and JO if and only if it satisfies ϕ(G) = ϕNE(G) for all G in ΓThe central result of this paper is that the Nash equilibrium correspondence is fully characterized by the four axioms introduced in the previous subsection. The proof demonstrates that if a solution concept satisfies IIS, ISDS and JO on a d-closed class of games, then it must be a weak refinement of Nash equilibriumPart (a) of the lemma above implies that any solution concept satisfying IIS, ISDS and JO on a d-closed class of games must be a weak refinement of Nash equilibrium. p. 7

Domain and Limitations

The domain of the axiomatization in Theorem 1 is any d-closed class of gamesThe domain of the axiomatization in Theorem 1 is any d-closed class of games. The paper notes that a class consisting only of one non-trivial game (e. p. 1g., the game G in Example 2) is not d-closed. p. 4 Additionally, the class of mixed extensions of finite games may not be d-closed, meaning a solution concept satisfying the four axioms may not select all mixed strategy Nash equilibria of a given finite game G. p. 1

Other Axioms Comparison

The paper compares its axioms with other concepts from the literature, noting that consistency (CONS) is analogous to IIS but they are logically independentIIS and CONS are logically independent. Converse consistency (COCONS) and converse independence of irrelevant strategies (CIIS) prescribe how to extend solutions from “smaller” to “larger” games, and these are also logically independent of MC. p. 2 The paper concludes by showing that the Nash equilibrium correspondence is fully characterized by the four axioms introduced in Section 3. p. 72The central result of this paper is that the Nash equilibrium correspondence is fully characterized by the four axioms introduced in Section 3. p. 72>. The examples illustrate that consistency does not imply IIS, and MC does not imply CIIS. p. 8

One-Player Game Simplification

In one-player games, i.e., games in which n = 1, the axiomatization in Theorem 1 can be simplified by dropping the MC axiom because the inclusion ϕNE(G) ⊆ ϕ(G) becomes an immediate consequence of JO alone. p. 9 For one-player games, it is possible to drop both IIS and MC for solution concepts defined on a strictly closed class of one-player games. p. 9 The paper presents Lemma 2 which shows that if a solution concept satisfies ISDS, then ϕ(G) ⊆ ϕNE(G) for all G in Γ. p. 7

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

  1. Improve solution concept selection by requiring satisfaction of four simple and intuitive axioms: independence of irrelevant strategies (IIS), merging consistency (MC), invariance to strictly dominated strategies (ISDS), and joint optimality (JO). This characterization allows for a full characterization of the Nash equilibrium correspondence, meaning a solution concept satisfies these axioms if and only if it assigns the same set of strategy profiles as the Nash equilibrium correspondence.

  2. Enable verification that any profile satisfying IIS, ISDS, and JO is a weak refinement of Nash equilibrium. This capability is useful in scenarios where checking every profile against the full definition of a Nash equilibrium is computationally prohibitive on large game structures.

  3. Allow for the derivation that any strict refinement of Nash equilibrium must violate at least one of MC or JO. This provides a structural tool to identify inconsistencies between different solution concepts, helping researchers distinguish between types of solution concepts, such as those selecting strong Nash equilibria versus those selecting pure strategy Nash equilibria.

  4. Simplify the axiomatization for one-player games by dropping the merging consistency axiom (MC). This reduction simplifies the necessary conditions for a solution concept in single-player decision problems, making testing and verification more straightforward when dealing with games where n equals 1.

Sources

Related papers