Strategic Scenario Analysis: Players, Actions, Information, and Payoffs

The Four Pillars of a Game

Any strategic situation, no matter how complex it seems, can be broken down into four essential components. Think of this as the "PAIP" framework: Players, Actions, Information, and Payoffs.

To ground our understanding in formal definitions, which I know you appreciate, let's start with a concise video that outlines the core ingredients of a strategic game.

Building on that video and our previous work, let's define the four pillars in more detail. A well-defined game specifies:

1. Players: These are the strategic decision-makers in the game. We denote the set of players by $N=\{1,2,...,n\}$. A player can be an individual, a company, a country, or even a fictional player like "Nature" or "Chance," which we use to model uncertainty and random events (as we saw briefly when discussing mixed strategies).

2. Actions (or Strategies): For each player $i$, there is a set of available actions or moves, $A_i$. In the simultaneous-move games we're focusing on for this modeling exercise, an "action" is the same as a "pure strategy." A strategy profile is a list of actions, one for each player, e.g., $(a_1,a_2,...,a_n)$ where $a_i\in A_i$.

3. Information: This component describes what each player knows when they make their decision. This is a subtle but critical element. We need to specify:

   • Knowledge of the Game: Do all players know the PAIP components? In most of our analyses, we assume the structure of the game is common knowledge—everyone knows it, everyone knows that everyone knows it, and so on.
   • Knowledge of Past Actions: Do players know the actions taken by others who moved before them?
     • Perfect Information: Every player knows the full history of the game up to the point of their decision. (e.g., Chess).
     • Imperfect Information: At least one player does not know the full history. This happens in simultaneous-move games where players choose their actions without knowing what their opponents are choosing. We use information sets to model this uncertainty, grouping together decision nodes that a player cannot distinguish between.
   • Knowledge of Payoffs: Do players know each other's payoff functions?
     • Complete Information: Every player's payoff function is common knowledge.
     • Incomplete Information: At least one player is uncertain about another player's payoff function (i.e., their "type"). We explored this in our module on Bayesian games.

4. Payoffs: For each possible outcome (i.e., for each strategy profile), each player $i$ receives a payoff, represented by a utility function $u_i$. This function maps a strategy profile to a real number: $u_i(a_1,a_2,...,a_n)\in \mathbb{R}$. These are typically von Neumann-Morgenstern utilities, meaning players aim to maximize their expected payoff. Critically, a player's payoff depends not just on their own action, but on the actions of all other players.

Guided Deconstruction: The Grade Game

The best way to learn this framework is to apply it. We'll use an excellent example from a Yale University course, where Professor Ben Polak starts with a simple paragraph of text and systematically builds a game-theoretic model from it.

First, let's watch how the scenario is introduced and how the initial components are identified.

Let's use our PAIP framework to deconstruct what we just saw.

• Players: There are two players in each interaction: "Me" (the student making the choice) and "My Pair" (the randomly assigned partner). So, $N=\{\text{Me, Pair}\}$.
• Actions: Each player has the same set of two available actions: choose the letter 'Alpha' or choose the letter 'Beta'. So, $A_{Me}=A_{Pair}=\{\text{Alpha, Beta}\}$.
• Information: The game involves a "simultaneous move" (students choose without knowing their pair's choice), so it's a game of imperfect information. For now, let's assume it's a game of complete information, where both players understand the rules and potential outcomes perfectly.

Now for the most important and nuanced step: the payoffs. Notice that what Prof. Polak created was an outcome matrix (showing grades like A, B+, C), not a payoff matrix. As he points out, this is not yet a game.

This is where the art of modeling comes in. We must make an assumption about what the players care about. Let's start with the simplest assumption: players are purely self-interested and only care about their own grade.

By assigning numerical values (A=3, B+=1, B-=0, C=-1), we get the following payoff matrix, which you should recognize as the Prisoner's Dilemma:

The Power of Payoff Assumptions

What if players aren't "evil gits"? What if they have social preferences, as we discussed in our module on behavioral game theory? This is where modeling connects directly to psychology and behavior. A different assumption about motivations creates a completely different game.

This new game, which is a version of the Stag Hunt or Assurance Game, has a completely different strategic dynamic. This demonstrates that the most critical, and often most difficult, part of deconstructing a real-world scenario is accurately modeling the players' payoffs.

From Deconstruction to Analysis

The purpose of building a model is to analyze it. By explicitly defining the PAIP components, we can predict behavior. The final part of the Yale lecture shows this beautifully by analyzing a game between an "evil git" and an "indignant angel." The key lesson is that by modeling your opponent's payoffs (putting yourself in their shoes), you can predict their likely actions and determine your own best response.

Practice Problem: A Duel of Honor

Now it's your turn to apply the PAIP framework. Consider the following scenario, which is a classic setup in game theory.

Take a moment to deconstruct this scenario yourself before reading on.

1. Players: Who are they?
2. Actions: What can each player do?
3. Information: What do they know when they act?
4. Payoffs: What utility does each player get for each outcome? Try to construct the payoff matrix from the story.

Solution:

• Players: Gentleman 1 and Gentleman 2.
• Actions: For both players, $A_i=\{\text{Shoot straight, Duck}\}$.
• Information: They act simultaneously, so it's a game of imperfect information.
• Payoffs: The story translates to the following matrix (where G1 is the row player):

This shows how a narrative can be systematically translated into a formal game model, ready for analysis.

A Final Example: The Emergency Maneuver

Your interest in physics and maths might make you appreciate this final example, which involves the dynamics of two cars avoiding an obstacle.

This example reinforces the same systematic process: a real-world problem is broken down into its constituent strategic parts (Players, Actions, Information, Payoffs) to create a model that can be formally analyzed.

Conclusion

Today, we've focused on the foundational skill of modeling. You learned how to take a messy, real-world situation and deconstruct it into the four essential components of a game using the PAIP framework:

• Players: The decision-makers.
• Actions: Their available choices.
• Information: What they know when they choose.
• Payoffs: Their objectives and valuations of outcomes.

We saw that this process is not always straightforward. The most challenging and most important step is often defining the players' payoffs, which requires making assumptions about their motivations—be they purely self-interested or influenced by complex psychological factors like fairness, guilt, or status.

Mastering this deconstruction process is the first and most critical step in applying game theory. It's the bridge between observing the world and analyzing it strategically.

Preview of the Next Lesson:

Now that we have formalized the process of building a game model, we will apply it to a classic scenario with more than two players. In our next lesson, we will model a public goods game and identify the free-rider problem as an N-player social dilemma. This will expand our modeling skills to situations involving groups, a common feature of many real-world strategic challenges.