Welcome back. In our previous lesson, we established a framework for analyzing how partners can use costly actions as credible signals to convey private information, such as their level of commitment. We worked through a detailed analysis of a separating equilibrium, where different types of partners choose different signals, allowing the Hinge to learn their true type with certainty.
However, communication is not always successful. Sometimes, actions are ambiguous and leave the receiver just as uncertain as they were before. This lesson focuses on that ambiguity. Your learning outcome is to distinguish separating from pooling equilibria in vee relationship signaling contexts. We will directly compare the successful information transmission of separating equilibria with the informational failure of pooling equilibria, exploring the conditions that lead to each.
The Great Divide: Revealing vs. Concealing Information
The fundamental difference between these two types of equilibria lies in how much information the signal conveys. Let's start with a clear definition of each, which you can review in the "Signaling game" Wikipedia article.
This section provides concise definitions of the different types of equilibria in signaling games.
Please read the paragraph that begins "The kinds of perfect Bayesian equilibria...". Focus on the core distinction: in a pooling equilibrium, all types choose the same signal, whereas in a separating equilibrium, they choose different signals.
To put it in the context of our vee relationship:
- A separating equilibrium is a success story for communication. The Hinge observes a specific action and can confidently say, "Because you did X, I now know you are a High-Commitment partner." Information is successfully revealed.
- A pooling equilibrium is a story of ambiguity. The Hinge observes an action but must conclude, "You did X, but both High-Commitment and Low-Commitment partners would have done that, so I haven't learned anything new." Information remains concealed.
Why would this happen? A pooling equilibrium typically arises for one of two reasons:
- The signal is too cheap: The action is not costly enough to deter the "Low-Commitment" type from mimicking the "High-Commitment" type. If saying "I'm planning our future" costs nothing, both types might say it to gain short-term favor.
- The alternative is too costly: The outcome of not sending the signal is so bad that all types are compelled to send it, just to avoid the negative consequence.
Last time, we focused on the step-by-step process of verifying a separating equilibrium. Now, let's apply a similar systematic approach to a single scenario and see how it can result in either a separating or a pooling equilibrium, depending on the payoffs.
A Scenario: Builder vs. Tourist
Consider a Hinge in a vee with a relatively new Partner. The Hinge is uncertain about the Partner's long-term intentions.
- Players: The Partner (Sender) and the Hinge (Receiver).
- Partner's Types:
- Builder (probability ): Genuinely wants a deeply integrated, long-term relationship.
- Tourist (probability ): Enjoys the relationship for what it is now, but does not want deeper long-term entanglement.
- Partner's Signals (Actions):
- Propose Integration (I): A costly signal, like suggesting moving in together, meeting family, or making significant joint financial plans.
- Keep Casual (C): Maintain the current level of interaction.
- Hinge's Responses:
- Deepen (D): Invest more time, emotion, and resources into the relationship.
- Maintain (M): Keep the relationship at its current level.
The following video provides a fantastic walkthrough of how to analyze a game just like this one for both separating and pooling outcomes. We will adopt its clear, methodical style.
17. A Quick Review for Pooling and Separating Eqm of Signalling Games (Game Theory Playlist 10)
This video by Selcuk Ozyurt provides a quick and clear review of finding both separating and pooling equilibria in a single signaling game. We'll use this method to analyze our "Builder vs. Tourist" scenario.
First, watch the introduction to see the game structure. Next, follow the analysis of the two potential separating equilibria. Observe how one is confirmed as valid while the other fails because a player has an incentive to deviate. Finally, watch the analysis of the two potential pooling equilibria. Pay close attention to why one fails but the other one holds, and note the discussion about the "free parameter" (q in the video) for beliefs off the equilibrium path.
Now, let's apply this logic to our own game.
Analysis 1: The Separating Equilibrium (Information Revealed)
Let's assume the following payoffs (Hinge, Partner):
- If Partner is a Builder:
- I -> D: (2, 2) (Ideal outcome for both; signal cost is worth it)
- I -> M: (0, 0) (Gesture ignored; signal cost wasted)
- C -> D: (1, 3) (Builder gets deep commitment without paying the signal cost - great for them, but Hinge wouldn't do this)
- C -> M: (0, 1) (Status quo, no progress)
- If Partner is a Tourist:
- I -> D: (-2, -1) (Tourist is unhappily entangled, and paid the cost of the signal)
- I -> M: (0, -2) (Gesture ignored, costly failure)
- C -> D: (-1, 2) (Tourist gets undeserved investment without paying signal cost - their dream, Hinge's nightmare)
- C -> M: (0, 1) (Ideal for Tourist: status quo, no costs)
Proposal: Let's test the intuitive separating strategy: Builder Proposes Integration (I), Tourist Keeps Casual (C).
-
Hinge's Beliefs & Best Response:
- Observing I: The Hinge believes the Partner is a Builder with 100% certainty (
Prob(Builder|I) = 1). The Hinge's best response is to Deepen (payoff 2 > 0). - Observing C: The Hinge believes the Partner is a Tourist (
Prob(Tourist|C) = 1). The Hinge's best response is to Maintain (payoff 0 > -1).
- Observing I: The Hinge believes the Partner is a Builder with 100% certainty (
-
Check for Sender Deviations:
- Builder: In equilibrium, they play I and the Hinge plays D, for a payoff of 2. If they deviated to C, the Hinge would play M, for a payoff of 1. Since 2 > 1, the Builder does not deviate.
- Tourist: In equilibrium, they play C and the Hinge plays M, for a payoff of 1. If they deviated to I, the Hinge would think they're a Builder and play D, for a payoff of -1. Since 1 > -1, the Tourist does not deviate.
Conclusion: This is a stable separating equilibrium. The high cost of integration for the Tourist makes the signal credible, and information is successfully revealed.
Analysis 2: The Pooling Equilibrium (Information Concealed)
Now, let's change the payoffs. Imagine the Tourist doesn't dislike deeper commitment; they're just neutral. And let's make the signal cost lower.
New payoffs for the Tourist:
- I -> D: (-2, 1) (Entanglement is bad for Hinge, but Tourist gets investment and only pays a small signal cost)
- All other payoffs remain the same.
Proposal: Let's test a pooling strategy: Both Builder and Tourist Propose Integration (I).
-
Hinge's Beliefs & Best Response:
- Observing I: Since both types are meant to play I, the Hinge learns nothing. Their beliefs remain the same as their prior beliefs:
Prob(Builder|I) = 0.7,Prob(Tourist|I) = 0.3. The Hinge must calculate their expected utility for responding with D or M.- Expected Utility of Deepen (D):
- Expected Utility of Maintain (M):
- Since 0.8 > 0, the Hinge's best response is to Deepen.
- Observing C (Keep Casual): This is now an off-the-equilibrium-path action. No type is supposed to do this. Bayes' rule doesn't apply. We must specify a belief for the Hinge. Let's assume a pessimistic belief: if the Hinge sees this unexpected casualness, they assume the worst. Belief:
Prob(Tourist|C) = 1. Given this belief, the Hinge's best response is to Maintain (payoff 0 > -1).
- Observing I: Since both types are meant to play I, the Hinge learns nothing. Their beliefs remain the same as their prior beliefs:
-
Check for Sender Deviations:
- Builder: In equilibrium, they play I and the Hinge plays D, for a payoff of 2. If they deviated to C, the Hinge (with our specified off-path belief) would play M, for a payoff of 1. Since 2 > 1, the Builder does not deviate.
- Tourist: In equilibrium, they play I and the Hinge plays D, for a payoff of 1. If they deviated to C, the Hinge would play M, for a payoff of 1. Since the payoff is the same (1 vs 1), the Tourist has no strict incentive to deviate. They are indifferent, which is sufficient for the equilibrium to hold.
Conclusion: This is a stable pooling equilibrium. Because the signal is now "cheap enough" for the Tourist to mimic the Builder, both types send it. The gesture of "Proposing Integration" becomes ambiguous and uninformative. The Hinge is forced to act based on their initial probabilities, and the Tourist successfully "hides in the crowd" with the Builder.
The role of off-path beliefs is the critical, and perhaps most subtle, distinction in analyzing pooling equilibria. As you saw in the Spaniel video from the last lesson, and the Ozyurt video from this one, the analyst must define what the receiver would believe if they saw an action that "shouldn't" happen. The stability of the pooling equilibrium often depends entirely on these assumptions.
You can see these different equilibrium structures visualized clearly in the diagrams from the MIT lecture slides.
These slides offer clear visual representations of the different equilibrium types.
First, quickly review the definitions on slide 4. Then, look at the diagrams on slide 5. Notice how in the pooling equilibrium, both types (tw and ts) take the same action ("beer"), while in the separating equilibrium, they take different actions ("beer" vs. "quiche"). This visually captures the essence of the distinction.
Key Takeaways
This lesson was dedicated to the crucial distinction between revealing and concealing information through strategic action.
- Separating vs. Pooling: In a separating equilibrium, different types send different signals, revealing private information. In a pooling equilibrium, all types send the same signal, concealing that information.
- The Role of Costs and Payoffs: A signal's credibility hinges on its cost. If a signal is too costly for a "low type" to mimic a "high type," a separating equilibrium is possible. If the signal is cheap, or if not signaling is prohibitively costly, a pooling equilibrium is more likely.
- Off-Path Beliefs: Analyzing pooling equilibria requires specifying the receiver's beliefs for actions that are not supposed to occur in equilibrium. The stability of the equilibrium can depend on these specified beliefs, a complication not present when analyzing a simple separating equilibrium.
In our vee context, this distinction is paramount. A gesture intended to signal high commitment might be truly informative (separating) or completely ambiguous (pooling), depending on the perceived costs and benefits for all involved. Understanding this difference is key to interpreting partners' actions accurately.
So far, we have focused on costly signals. But what about communication that is free—simple conversation, or "cheap talk"? Can it ever be credible? Our next lesson will explore the conditions under which costless communication can (and cannot) resolve uncertainty in a vee relationship.