Welcome back. In our last lesson, we developed a qualitative framework for understanding strategic interactions in a vee, classifying them into recognizable categories like coordination games, the Stag Hunt, and the Prisoner's Dilemma. We saw that the core tension of a situation can be understood by the preference ordering of its outcomes.
Today, we will make this analysis fully quantitative. Our objective is to construct payoff matrices for typical vee dilemmas, moving from abstract preference orderings to concrete numerical models. This is the crucial step of translating a real-world narrative into the formal language of game theory, setting the stage for rigorous analysis. This process of modeling, where we define variables and relationships to represent a system, should be familiar from your work in operations research and simulation.
The Anatomy of a Payoff Matrix
At its heart, a payoff matrix is a table that displays the outcomes for every player for every possible combination of strategies. A well-constructed matrix is the foundation for all subsequent game-theoretic analysis.
The numbers in the matrix represent utility. A common misconception is that utility is simply money or some other tangible resource. In fact, it is much broader. A player's utility function is designed to capture everything that player cares about.
Game Theory (Stanford Encyclopedia of Philosophy/Fall 2011 Edition)
To solidify this crucial concept, let's turn to the Stanford Encyclopedia of Philosophy's entry on Game Theory. It provides a clear and authoritative explanation of utility and payoffs.
Please read the section titled "On Interpreting Payoffs: Morality and Efficiency in Games." Focus on the paragraph that begins "A utility function for a player is supposed to represent...". This passage powerfully argues that payoffs can and should include factors like altruism, social good, or, in our context, emotional security, fairness, and relationship harmony.
This broad definition of utility is what allows us to model the complex, emotionally rich scenarios found in relationships. Feelings of jealousy, compersion, security, or neglect are all valid components of a player's payoff.
How to Build a Matrix from Scratch
Constructing a matrix is a systematic process. We start with a scenario, define the strategic elements, and then assign numerical values to the outcomes. A fantastic guide to this process is provided in the following video, which uses the relatable example of telling "white lies" in a relationship.
The Game Theory of White Lies | How to Construct a Matrix from Scratch
This video by Ashley Hodgson offers a step-by-step walkthrough of building a 2x2 payoff matrix from a narrative scenario. It provides a clear, practical method that we can adapt to our own examples.
Please watch these key segments: First, see how the players and their strategies are defined from the initial scenario here. Next, pay close attention to the process for the first player: ranking the four possible outcomes from most to least preferred, and then assigning numbers that reflect this ranking. Finally, watch as the same process is repeated for the second player to complete the matrix.
The core steps from the video are:
- Define Players and Strategies: Who is involved and what are their choices?
- Establish Preference Ordering: For each player, individually, rank all possible outcomes from best to worst. This is the most critical step, requiring you to step into each player's shoes.
- Assign Ordinal Payoffs: Assign numbers to the outcomes that preserve the preference order. For many simple games, as the video notes, the exact numbers don't matter as long as the order is correct.
- Assemble the Matrix: Populate the cells with the payoff pairs for each outcome.
As you can see, this is essentially a formal exercise in preference elicitation, a key component of the Bayesian Decision Theory you have studied.
Moving to Cardinal Utility
The video correctly states that for finding simple pure-strategy equilibria, ordinal rankings are often sufficient. However, your background in statistics will tell you that the magnitude of differences can matter, especially when uncertainty is involved. For a more robust model, we need cardinal utility, where the numbers reflect the intensity of preferences.
The standard method for this is the von Neumann-Morgenstern (VNM) utility function. The core idea is to measure preference intensity by examining choices between "lotteries" (risky prospects).
Game Theory (Stanford Encyclopedia of Philosophy/Fall 2011 Edition)
The same SEP article provides an excellent, non-technical overview of how cardinal utility is derived.
Please read the section on "Uncertainty, Risk and Sequential Equilibria". Focus on the part that starts at this paragraph and continue until the end of the next long paragraph. The key insight is that by asking an agent to choose between a guaranteed intermediate prize and a lottery over the best and worst prizes, we can precisely measure the cardinal value of that intermediate prize.
While we won't be constructing full VNM utility functions in this course (in keeping with your preference for breadth), it's important to recognize this underlying principle. When we assign payoffs like 10, 5, and 1, we are implicitly making a statement about preference intensity: the jump from 1 to 5 is less significant than the jump from 5 to 10. This will become more important when we discuss mixed strategies and expected payoffs later on.
Modeling the Vee: A 3-Player Matrix
Now let's apply this to our primary subject: the three-player vee. A two-dimensional matrix is insufficient here. We need a way to represent the choices of the Hinge, Arm 1, and Arm 2 simultaneously. This is typically done by creating "layers" of matrices, where each layer corresponds to a choice by the third player.
The following video demonstrates how to work with this structure to find solutions, which implicitly shows how it's set up.
Nash Equilibrium with 3 Players
This video, also by Ashley Hodgson, explains how to set up and analyze a 3-player game.
Watch the entire video. Focus on: How Player 3's choices act as a "selector" between two different game matrices for Players 1 and 2 at the start. How each player's best response is found by holding the other two players' strategies constant. For example, Player 3's decision process involves comparing payoffs across the two matrices.
Example Dilemma: The Disclosure Decision
Let's construct a matrix for a common vee dilemma: disclosure of a new potential partner.
- Players: Hinge (H), Arm 1 (A1), Arm 2 (A2).
- Situation: H has gone on a promising first date with a new person (New Person, or NP) and must decide how and when to tell A1 and A2. This is a sensitive issue involving feelings of security, trust, and compersion/jealousy.
- Strategies:
- Hinge:
Disclose Immediatelyvs.Wait to See if it's Serious. - Arm 1:
React Supportivelyvs.React with Insecurity. - Arm 2:
React Supportivelyvs.React with Insecurity.
- Hinge:
We can now build the 8-outcome, 3-player payoff matrix. Let's think through the preference ordering for a few key outcomes, assigning payoffs on a 0-10 scale.
Scenario 1: Hinge Discloses Immediately (Top Matrix)
| Arm 1 / Arm 2 | React Supportively (A2) | React with Insecurity (A2) |
|---|---|---|
| Supportive (A1) | (H: 10, A1: 8, A2: 8) | (H: 4, A1: 9, A2: 2) |
| Insecure (A1) | (H: 4, A1: 2, A2: 9) | (H: 1, A1: 3, A2: 3) |
- (10, 8, 8): The ideal outcome. H feels honest and validated. Both arms feel secure, trusted, and compersive. High utility for all.
- (1, 3, 3): A disastrous outcome. H's disclosure is met with a wall of insecurity, causing conflict and stress. Everyone is miserable.
- (4, 9, 2): A mixed outcome. H is stressed by A2's reaction. A1 feels comparatively "better" and more secure (a slight temptation payoff), while A2 feels threatened and unhappy.
Scenario 2: Hinge Waits (Bottom Matrix)
This matrix would be filled out similarly. For example, if H waits and the relationship with NP becomes serious before disclosure, a supportive reaction from the arms might still yield lower payoffs for them than in the "Disclose Immediately" case, because they may feel they were kept in the dark.
This exercise demonstrates how the abstract structure becomes a concrete model of a specific relationship dynamic, ready for analysis.
Advanced Construction: Parametric Payoffs
Your background in OR and statistical modeling opens the door to a more powerful approach. Instead of fixed numbers, we can define payoffs as functions of parameters. This allows us to build a general model and see how outcomes change as underlying conditions shift.
A fascinating paper on the "human mating game" does exactly this, modeling cooperation (faithfulness) and defection (cuckoldry). This directly connects to your stated interests and professional experience.
Evolution of Co-operation When the Strategies are Hidden - Nature
This research paper models mating strategies using a game-theoretic approach with parametric payoffs. It's an excellent example of advanced matrix construction.
First, read the Methods section. Focus on the paragraph that defines the parameters of the model. Note how terms like L (genetic loss), gc (cost of guarding), and BB (punishment for adultery) are established. Next, examine Table 1, the payoff matrix. Observe that each cell is not a number but an equation derived from those parameters. For example, the payoff for a monogamous, non-guarding, non-policing individual (M+G-P-) playing against a polygamous, guarding, policing individual (M-G+P+) is 1 - (L(1-α)) - pc + BP. This is a sophisticated application of the same principles we've discussed.
This parametric approach allows for powerful sensitivity analysis. One could ask, "How much must the cost of punishment (BB) increase to make defection an unstable strategy?" This moves from analyzing a single game to analyzing a whole family of games, which is the essence of robust modeling.
Conclusion
Today we have transitioned from qualitative game classification to quantitative model building. You now have a complete toolkit for constructing payoff matrices, the foundational element of game-theoretic analysis.
Key Takeaways:
- The Process is Key: Constructing a matrix is a systematic process of defining players, strategies, and then eliciting and assigning utility values to outcomes.
- Utility is Broad: Payoffs can and should capture everything a player values, including complex emotional and psychological states.
- 3-Player Matrices Model the Vee: By using a "layered" matrix structure, we can accurately represent the strategic interdependencies among a hinge and two partners.
- Parametric Models Offer Generality: For advanced analysis, payoffs can be represented as functions of key variables, allowing for a more general and powerful model.
In our next lesson, we will put these matrices to work. Having built the game, we will now learn how to solve it by analyzing the best-response dynamics in a three-player vee game to identify potential stable outcomes, or Nash equilibria.