Equilibrium Selection in Coordination Games with Planned Actions and Scouting

arXiv:2609.17608 · econ.TH, econ.GN, q-fin.EC · Submitted 2026-09-14 · 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: "Equilibrium Selection in Coordination Games with Planned Actions and Scouting".

Warren: Here's the gist of the paper. Equilibrium selection in coordination games with planned actions and scouting studies how players can screen for cooperation based on partially revealed information about their adversary’s preparations to select a payoff-dominant equilibrium This research investigates coordination games where every action requires planning and preparation, allowing agents to revise plans based on partially revealing information about their adversary’s preparations Game Structure and Protocol The present paper studies games where play involves three stages: first, both players choose which action to prepare.

Ruth: In the second stage, both players receive noisy information about their adversary’s preparations.

Warren: Finally, given the information about their adversary’s planned action and their own planned action, players choose whether to execute or revise their plans.

Ruth: The present paper focuses on stag-hunt-type coordination games.

Title and authors: Ruth: So we're looking at this paper today called "Equilibrium Selection in Coordination Games with Planned Actions and Scouting." It looks at coordination games where players have to plan their actions beforehand and then revise those plans based on what they see about what the other player is preparing.

Warren: Right, so it’s about how you screen for cooperation when you know the other person might be planning something. It seems like a lot of setups involving preparation and then some information exchange between stages.

Iris: It starts by setting up this three-stage process where players choose what to prepare, then they get noisy information about the opponent's preparations, and finally, based on all that info, they decide whether to stick with their plan or revise it.

Ruth: That’s right. The paper focuses specifically on stag hunt type coordination games—things like rabbit nets or stag hunting gear where you have to coordinate your effort. It shows how this three-stage model can be interpreted as preparing equipment, then inspecting each other's gear from a distance, and finally, if you spot a mismatch, you spook them and force a reset.

Warren: And the authors use these examples to build up the game structure. They are looking at stag hunt type games because it gives them a concrete way to visualize how this planning and scouting mechanism actually works in practice.

Iris: Before we get into the results, it establishes two pure strategy perfect bayesian equilibria for this modified Stag Hunt game with scouting. One is where both players always plan action A and execute unconditionally, which gives a payoff vector of (a, a).

Ruth: And the second one is a low-payoff equilibrium where both players always plan action B and execute unconditionally, leading to a payoff vector of (b, b). These are the pure strategy equilibria where no one ever revises their plans.

Warren: So we have two stable outcomes right there: the payoff dominant one with both players picking A, and the low-payoff one with both picking B. But the paper is more interesting than just listing those two options because it shows how they can be destabilized.

Title and authors: Iris: Exactly. The core finding here is that cheap, precise scouting actually destabilizes that safe low-payoff equilibrium, which is the (B, B) outcome.

Ruth: It does this by allowing agents to screen for cooperative preparations before committing fully to the lower payoff choice. If Player two makes occasional errors by planning and unconditionally executing action A2 with a probability of alpha two equals epsilon, then for any fixed tremble epsilon greater than zero, there are threshold values p star of epsilon and delta star of epsilon such that if the scouting precision p is above that threshold and the revision friction delta is above that threshold, it becomes profitable for Player one to deviate.

Warren: So what does this deviation look like in practice? It means Player one will plan action A1 every period and execute only when the scouting signal reads action A hat i. That’s a clear incentive to switch from the safe B strategy.

Iris: Precisely. This screening mechanism essentially renders that low-payoff outcome unstable against small perturbations, meaning it can't withstand tiny changes in how players behave or how precise their information is.

Ruth: And this leads us into another area where things get even more interesting regarding the dynamics of play. Proposition four demonstrates that when players have the ability to unilaterally force the execution of their plans even when opponents are trying to revise them, that safe equilibrium with both playing B unravels, while the payoff dominant equilibrium with both playing A survives under certain conditions.

Warren: That unraveling only happens if the scouting precision is high, meaning p and q go up toward one, and the revision frictions are low—specifically when delta is greater than max of d a and c prime a.

Iris: The proof shows that the normal type can profitably deviate by always planning action A1 and setting chi one upon observing action hat i. This suggests that if you have good information, you can exploit it to push the system toward the better outcome.

Title and authors: Ruth: So, to wrap up on this whole piece, the central idea is that monitoring preparations creates a strong incentive for players to actively screen for cooperative behavior when their monitoring precision is high and the time cost of revisions is low.

Warren: If those two conditions are met, the safe but payoff-inferior equilibrium with both playing B becomes unstable because small exogenous payoff uncertainty or trembling hand play causes that inferior equilibrium to unravel, isolating the payoff dominant outcome with both playing A as what the model predicts.

Iris: And if we look at what this changes for us, it shows that active screening is a key driver in these coordination games when information quality and decision speed are favorable.

Ruth: It really highlights how players aren't just reacting to payoffs; they are actively trying to discover the correct cooperative plan by watching their opponents' preparation signals.

Warren: Thinking about the big picture, this suggests that in systems where planning is involved, having a mechanism for active screening can push an equilibrium from a safe but mediocre choice toward the one that actually maximizes everyone's payoff.

Iris: I think it reinforces the idea that information flow and decision structure are intrinsically linked to which equilibrium ends up being played in these types of games.

Ruth: It does, and it connects this study back to how we look at games with asymmetric information, which is a big part of the literature here.

Warren: So next time you're looking at a coordination game, don't just assume the standard equilibrium holds; ask yourself if there’s enough information to make someone start screening for cooperation.

Iris: And remember, this whole model relies on the assumptions about how plans are reversible and how those noisy signals are actually received by the players.

Ruth: Right. We'll leave it there for today, but we've got a lot more on arXiv coming up soon.

The paper's summary: Ruth: So to recap, this paper shows how players can actually find a better agreement in coordination games if they start scouting each other's plans before they commit to their own action.

Warren: It’s about moving away from just picking the safest bet when you know there’s a better one out there by using that information exchange.

Iris: The big mechanism here is that players use noisy signals about what the other person is planning to do to decide whether they should stick with their initial plan or revise it.

Ruth: Exactly. They set up this three-stage process where you prepare something, then you get some fuzzy info about your opponent’s prep, and finally, you decide if you execute or change your mind based on that signal.

Warren: And the paper highlights how this screening ability destabilizes the safe outcome—the one where both players just stick to the lowest payoff choice.

Iris: It shows that when scouting is precise and there’s low friction for changing plans, it creates an incentive for people to deviate from the safe bet toward a better equilibrium.

Ruth: That deviation isn't random; it’s strategic. Player one starts planning action A every time and only executes if they see the signal that Player two is also planning A.

Warren: So, the numbers show that this screening mechanism renders the low-payoff outcome unstable against even small changes in how players act or how accurate their information is.

Iris: This means that in these coordination games, just having a little bit of noise or a slight uncertainty can cause the system to jump from the safe, worse outcome straight to the payoff dominant one.

Ruth: It’s about how active monitoring creates this pressure to find cooperation when those conditions—high precision and low revision costs—are met.

Warren: Thinking about this for trading strategy, it suggests that if you can reliably see what your opponent is doing before they finalize their move, you might be able to steer the outcome away from a mediocre result.

Iris: But the paper has its limits too; it focuses on these stag hunt setups, so we have to be careful about applying this directly to every kind of coordination problem out there.

Ruth: Right. It’s a useful tool for understanding how planning and information gathering interact in games where you can't just assume everyone is playing optimally from the start.

Warren: Next up, we’re going to look at how these kinds of sequential decision-making models connect to things like dynamic mechanism design, which deals with setting rules when information is constantly changing.

The paper's improvements: Iris: So, we're looking at how they suggest making this model even better, moving beyond just describing what happens in the game.

Ruth: It seems like the main idea is that we need to look at how players can actually use their private knowledge to make this screening process more robust.

Warren: They propose using AI systems that can build a whole protocol for planning and scouting actions, breaking those big decisions down into separate preparation and execution steps.

Iris: That’s the first improvement they suggest: having an AI system that implements a planning-and-scouting protocol to select payoff dominant equilibria by breaking actions into preparation and execution.

Ruth: It means the AI doesn't just guess; it actually runs a procedure to find the better outcome based on what it observes.

Warren: Then there's another point about using noisy signals about opponents’ plans to actively screen for cooperation, which enables selecting that payoff dominant equilibrium via small exogenous trembles or via payoff uncertainty.

Iris: That ties back into the core idea of using those imperfect signals to push the system toward the better result when things aren't perfectly certain.

Ruth: They also suggest an improvement in how we look at simultaneous-move games where players partially observe each other’s actions by decomposing actions into preparation and then execution, moving past just steady-state frameworks.

Warren: That moves it beyond looking at a single snapshot of play and lets us see the sequence of decisions that leads to that equilibrium.

Iris: And they talk about designing systems that adapt their strategies based on sequential rationality in the revision stage, where players decide whether to execute or revise their plans based on what they just saw and what they believe.

Ruth: That focuses on the decision-making process itself, not just the final predicted state.

Warren: So, this is about building a more dynamic system that reacts intelligently to incoming information rather than just following a pre-set path.

Iris: They also suggest improving the AI by allowing it to identify and exploit instabilities in low-payoff equilibria, because precise scouting makes those low-payoff outcomes unstable against small perturbations.

Ruth: So, if the system can detect those tiny wobbles in the safe outcome, it can force a shift toward that higher payoff dominant one.

Warren: That’s interesting because it means we might not need perfect information to get to a good result; we just need enough precision to exploit those instabilities.

Iris: The paper also touches on incorporating learning mechanisms where players use private information and knowledge of the equilibrium to indirectly infer each other’s actions, which informs their planning.

Ruth: That brings it back to private info, but framed as using that knowledge not just for direct guessing, but for more indirect inference about what others are actually doing.

Warren: It sounds like they want a system that doesn't rely on perfect communication, but rather on sophisticated ways of inferring intent from imperfect signals.

Iris: So the implication is we might design systems that can operate effectively even when the information flow between agents is messy or incomplete.

Ruth: Right. This leads into how these types of sequential decision models connect to things like dynamic mechanism design, which deals with setting rules when information is constantly changing.

Conclusion: Tom: So to wrap up, we’ve seen that this paper on "Equilibrium Selection in Coordination Games with Planned Actions and Scouting" shows that active scouting is a strong incentive for players to find cooperation when information quality is high and revision costs are low.

Iris: It really drills down into how monitoring preparations creates this pressure for players to actively screen for cooperative behavior.

Ruth: That’s the big picture, right? Once that precision and low friction are there, the safe but lower payoff outcome becomes unstable because small uncertainties cause it to unravel in favor of the better one.

Warren: It means if you have reliable monitoring, you can exploit those tiny wobbles in play to push everyone toward that higher payoff dominant equilibrium.

Iris: I just want to stress that this whole model relies on assumptions about how plans are reversible and how those noisy signals are actually received by the players.

Ruth: True, and the authors themselves flag a limitation: they focus specifically on stag hunt type games, so we can't just apply this directly to every kind of coordination problem out there.

Warren: That’s fair. But what it does change for us is showing that in systems where planning is involved, having a mechanism for active screening can push an equilibrium from a safe but mediocre choice toward the one that actually maximizes everyone's payoff.

Iris: It reinforces the idea that information flow and decision structure are intrinsically linked to which equilibrium ends up being played in these types of games.

Ruth: Exactly. And we’ve seen how this kind of dynamic modeling connects back to how we look at games with asymmetric information, which is a big part of the literature here.

Warren: So next time you're looking at a coordination game, don't just assume the standard equilibrium holds; ask yourself if there’s enough information to make someone start screening for cooperation.

Wolfgang Kuhle

Corvinus University of Budapest · Max Planck Institute for Social Law and Social Policy

econ.TH, econ.GN, q-fin.EC

Submitted: 2026-09-14

Updated: 2026-11-09

Comments: 23 Pages, 1 Figure

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 95/100

The gist: Equilibrium selection in coordination games with planned actions and scouting studies how players can screen for cooperation based on partially revealed information about their adversary’s…

Key concepts

Stag Hunt Fable Model
This model describes a three-stage process: first, players prepare equipment (like nets or gear); second, they inspect each other's gear from a distance; and third, they decide whether to execute or revise their plans based on what they observed. This structure helps analyze how planning and scouting interact.
Screening Mechanism
This is the process where players use noisy signals about an opponent's preparations to infer their true plan. The paper demonstrates that cheap, precise scouting allows players to detect deviations from cooperation, enabling them to actively screen for a mutually beneficial outcome rather than settling for a suboptimal one.
Payoff-Dominant Equilibrium (A, A)
This is the best possible outcome in the game where both players always plan and execute action A. The research proves that under certain conditions—high monitoring precision and low revision costs—this superior equilibrium becomes the stable prediction because it is robust against small changes.
Destabilization of Safe Equilibrium
The safe equilibrium (B, B) is a low-payoff outcome where players always plan and execute action B. The study shows that if scouting is good, this safe outcome can be destabilized by small errors in planning or execution. This instability pushes the system away from the inferior outcome toward the better payoff-dominant strategy.

Terminology

Summary

Equilibrium selection in coordination games with planned actions and scouting studies how players can screen for cooperation based on partially revealed information about their adversary’s preparations to select a payoff-dominant equilibrium This research investigates coordination games where every action requires planning and preparation, allowing agents to revise plans based on partially revealing information about their adversary’s preparations

Game Structure and Protocol

The present paper studies games where play involves three stages: first, both players choose which action to prepare. In the second stage, both players receive noisy information about their adversary’s preparations. p. 1

Where the paper says it

In the first stage, both players choose which action to prepare. In the second stage, both players receive noisy information about their adversary’s preparations.

Finally, given the information about their adversary’s planned action and their own planned action, players choose whether to execute or revise their plans. p. 1
Where the paper says it

In the second stage, both players receive noisy information about their adversary’s preparations. Finally, given the information about their adversary’s planned action and their own planned action, players choose whether to execute or revise their plans.

The present paper focuses on stag-hunt-type coordination games. p. 1
Where the paper says it

Once neither player chooses to make further revisions, plans are executed and payoffs materialize. The present paper focuses on stag-hunt-type coordination games.

Within the stag hunt fable, the three stage model can be interpreted as (i) players prepare either rabbit nets or stag-hunting gear; (ii) they inspect each other’s equipment from a distance; and (iii) a player who detects a mismatch can deliberately make noise to spook rabbits and stags alike, aborting the hunt and forcing a reset to the preparation stage. p. 1
Where the paper says it

The present paper focuses on stag-hunt-type coordination games. Within the stag hunt fable, the three stage model can be interpreted as (i) players prepare either rabbit nets or stag-hunting gear; (ii) they inspect each other’s equipment from a distance; and (iii) a player who detects a mismatch can deliberately make noise to spook rabbits and stags alike, aborting the hunt and forcing a reset to the preparation stage.

Pure Strategy Benchmarks

Proposition 2 establishes two pure-strategy Perfect Bayesian Equilibria (PBE) for the modified Stag Hunt game with scouting. These equilibria are a payoff-dominant equilibrium where both players always plan action A and execute unconditionally, yielding an equilibrium payoff vector (a, a). p. 5

Where the paper says it

The modified Stag Hunt game with scouting admits two pure-strategy Perfect Bayesian Equilibria (PBE): 1. A payoff-dominant equilibrium where both players always plan action A and execute unconditionally, yielding an equilibrium payoff vector (a, a).

The second is a low-payoff equilibrium where both players always plan action B and execute unconditionally, yielding an equilibrium payoff vector (b, b). p. 5
Where the paper says it

A payoff-dominant equilibrium where both players always plan action A and execute unconditionally, yielding an equilibrium payoff vector (a, a). A low-payoff equilibrium where both players always plan action B and execute unconditionally, yielding an equilibrium payoff vector (b, b).

Players never revise their plans in the pure-strategy equilibria. p. 5
Where the paper says it

A low-payoff equilibrium where both players always plan action B and execute unconditionally, yielding an equilibrium payoff vector (b, b). Players never revise their plans in the pure-strategy equilibria.

Destabilization of the Safe Equilibrium

Proposition 3 shows that cheap, precise scouting destabilizes the safe (B, B) equilibrium by allowing agents to screen for cooperative preparations. Specifically, if Player 2 makes occasional errors by planning and unconditionally executing action A2 with probability α2 = ε ∈ (0, 1), then for any fixed tremble ε > 0, there exist threshold values p∗(ε) ∈ (1/2, 1) and δ∗(ε) ∈ [0, 1) such that if p > p∗ and δ > δ∗, it is profitable for Player 1 to deviate and to plan action A1 in every period and execute if and only if the scouting signal reads Aˆ i. p. 6

Where the paper says it

Suppose Player 2 makes occasional errors, planning and unconditionally executing action A2 with probability α2 = ε ∈ (0, 1). Then, for any fixed tremble ε > 0, there exist threshold values p ∗ (ε) ∈ (1/2, 1) and δ ∗ (ε) ∈ [0, 1) such that if p > p∗ and δ > δ∗ , it is profitable for Player 1 to deviate and to plan action A1 in every period and execute if and only if the scouting signal reads Aˆ 1.

This screening mechanism renders the low-payoff outcome unstable against small perturbations. p. 2
Where the paper says it

Section 5 demonstrates that precise scouting renders the low-payoff equilibrium unstable against small perturbations. Sections 6 and 7 provide diagrammatic intuition for this selection mechanism and extend the analysis to a setting where players can unilaterally force the execution of plans.

Forced Execution Dynamics

Proposition 4 demonstrates that when players can unilaterally force the execution of plans even when opponents seek a revision, the safe equilibrium (B, B) unravels while the payoff-dominant equilibrium (A, A) survives under certain conditions. This unraveling occurs if scouting precision is high (p, q → 1) and revision frictions are low (δ > max[d a, c′ a]). p. 9

Where the paper says it

Proposition 3 indicates that whenever the cost of revising plans is low (δ → 1) and scouting precision is high (p → 1), players have a strong incentive to screen for favorable engagements even when off-equilibrium cooperative play is extremely rare. This screening argument suggests the following: Corollary 1.

The proof shows that the normal type can profitably deviate by always planning A1 and setting χ1 = 1 upon observing Aˆ i. p. 4
Where the paper says it

A Perfect Bayesian Equilibrium (PBE) for the game with planned actions and scouting consists of a strategy profile s ∗ = (s ∗ 1 , s∗ 2 ) and a belief system µ = (µ1, µ2) such that: 1. Belief Consistency: For each Player i ∈ {1, 2}, upon choosing preparation xi ∈ {Ai , Bi} and observing signal yˆi ∈ {Aˆ i , Bˆ i}, the posterior belief µi(x−i | xi , yˆi) regarding the opponent’s prepared action x−i is derived from the opponent’s planning strategy α ∗ −i and the signal precisions (p or q) using Bayes’ rule.

Conclusion on Equilibrium Selection

The paper concludes that monitoring preparations creates a strong incentive for players to actively screen/scout for cooperative behavior. Once the monitoring precision is high and the time cost of revisions is low, the safe but payoff-inferior equilibrium (B, B) becomes unstable. p. 9

Where the paper says it

Proposition 3 indicates that whenever the cost of revising plans is low (δ → 1) and scouting precision is high (p → 1), players have a strong incentive to screen for favorable engagements even when off-equilibrium cooperative play is extremely rare. This screening argument suggests the following: Corollary 1.

Exogenous payoff uncertainty or small trembles thus cause the inferior equilibrium to unravel, isolating the payoff-dominant outcome (A, A) as the model prediction. p. 0
Where the paper says it

Before players act, they can revise their plans based on partially revealing information about their adversary’s preparations. Precise information enables agents to screen for cooperation, selecting the payoff-dominant equilibrium either via small exogenous trembles or via payoff uncertainty.

The analysis suggests that players actively screen for cooperation when monitoring precision is high and revision costs are low. p. 9
Where the paper says it

Proposition 3 indicates that whenever the cost of revising plans is low (δ → 1) and scouting precision is high (p → 1), players have a strong incentive to screen for favorable engagements even when off-equilibrium cooperative play is extremely rare. This screening argument suggests the following: Corollary 1.

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. Improved AI systems can implement a planning-and-scouting protocol to select payoff dominant equilibria by decomposing actions into preparation and execution, allowing agents to screen for cooperation.

  2. The improved system can utilize noisy signals about opponents' preparations to actively screen for cooperation, which enables the selection of the payoff dominant equilibrium via small exogenous trembles or via payoff uncertainty.

  3. The AI can analyze simultaneous-move games where players partially observe each other’s contemporaneous actions by decomposing actions into (i) preparation and (ii) execution, moving beyond steady-state frameworks.

  4. Systems can be designed to adapt their strategies based on sequential rationality in the revision stage, where players decide whether to execute or revise their plans based on the received signal and belief system.

  5. The improved AI can identify and exploit instabilities in low-payoff equilibria, as precise scouting renders them unstable against small perturbations, allowing for a shift toward the payoff dominant outcome (A, A).

  6. The system can incorporate learning mechanisms where players use private information and knowledge of the equilibrium to indirectly infer each other’s actions to inform their planning.

Abstract

We study Stag Hunt games in which every action requires planning and preparation. Before players act, they observe noisy signals of their opponent's preparations and can revise their plans, which allows players to screen for cooperation. As screening becomes more precise and cheap, small trembles ensure that, in every equilibrium, play yields the payoff-dominant outcome with a probability that converges to one, even when the safe, low payoff equilibrium is risk dominant. Across scenarios, we emphasize that decomposing an action into (i) preparation and (ii) execution allows us to analyze simultaneous-move games in which players partially observe each other's contemporaneous actions.

Sources

Related papers