Decision Making through Game Theory and Artificial Intelligence
Updated: Aug 17
TL;DR / Key Takeaways
Stop treating the other side like a wall. Every negotiation, pricing call, or partnership deal you're stuck on has a person behind it who's also trying to win. The moment you model them as a static obstacle instead of a decision-maker with their own payoffs, you've started on the wrong trail.
You can't skip to the clever move. Before you eliminate bad strategies or hunt for equilibrium, you have to actually map out what both sides can do, and what each combination costs or pays each of you.
"Should I move first?" isn't about confidence, it's closer to math. Whether you want to make the first offer or wait them out isn't a gut call or a personality thing. Once you've mapped it, the answer is usually obvious.
But none of this works on a grudge. Game theory assumes the other person is rational, chasing their own interests in a clear-eyed way. The second ego, spite, or an old wound enters the room, the predictions fall apart. Know when you're not actually playing a game anymore.
Explained like I am 5
They're playing too! 🙂 When you and a friend are deciding something together, remember — your friend is thinking too! They're not just standing there. They want things, just like you do.
Think first, then play! 🤔 Before you make your move, think: "If I do this... what will they do?" That way you don't get surprised!
Going first or waiting isn't about being scared or brave. 🐢🐇 Sometimes it's just smarter to go first, and sometimes it's smarter to wait, like in a game of tag. It's not about who's brave, it's about what's smart!
But if someone's grumpy, none of this works! 😠 If your friend is really mad or just wants to be difficult, then all the smart thinking doesn't help anymore. Sometimes feelings are just bigger than thinking!
This article describes the core principles behind a Claude Skill that helps me with these types of situations.
The principles behind the skill
Most decision-making advice assumes you're the only actor in the room. Optimize your options, pick the best one, done. That works fine for choosing a job offer. It falls apart the moment another party's choices affect your outcome and yours affect theirs: a negotiation, a pricing move, a market entry, a partnership. For those, I use a different skill: Strategic Decision Making, built on Pfleiderer's game theory framework.
The core discipline is stated as one line, and I make sure it's the first thing the skill says:
"The other party is not an obstacle, they are a decision-maker with their own objectives and incentives. You always analyze the situation from both sides of the table."I put this at the top because it's the single most common failure mode in strategic thinking, and it's not a knowledge gap, people know the other side has incentives. It's that under pressure, everyone reverts to modeling only their own move and treating the other party as a fixed obstacle to route around rather than a party who will react. The skill forces the payoff table to include both sides before any analysis starts, because a strategy that only makes sense from your own perspective isn't a strategy. It's a wish.
Here's a snippet from how the skill builds that table:
"Once you have enough information, summarize it in a simple two-player payoff table, your choices on one axis, their choices on the other, with approximate outcomes for each combination... Confirm the table before proceeding. The analysis is only as good as the payoff structure."This step feels slow, and people want to skip it and jump straight to "so what should I do." I don't let the skill do that. Every downstream move eliminates dominated strategies, checking for a first-mover advantage, finding the Nash Equilibrium is only as good as the table underneath it. If you can't say what the other side's realistic options are and roughly what each combination means for them, you're not doing strategic analysis, but biased, motivated reasoning with extra steps.
The piece I think is most underused, even by people who know some game theory, is the first-mover logic:
"If moving first gives you a better outcome — there's a first-mover advantage here and you should consider committing early. If moving first gives you a worse outcome, there may be a second-mover advantage, and you should consider waiting to react... Moving first only creates advantage if the other party believes you're truly committed."People treat "should I move first or wait" as a personality question, some people are just bold, some are just cautious. It's not a personality question, but a structural one, answerable from the payoff table: work out the other party's best response to each of your possible first moves, then check whether your best first-mover outcome beats what you'd get moving simultaneously. Half the time the honest answer is that waiting is strictly better, and the "boldness" people admire in early movers is really just an information advantage nobody bothered to calculate.
And I built in a check on the biggest weakness of the whole method, right in the coaching principles:
"Rationality is an assumption, not a guarantee. The analysis assumes the other party acts in their best interest. In high-stakes business contexts, that's usually safe. In lower-stakes or emotional contexts, it may not be."Game theory is precise and it's also brittle: the moment the other party is acting on ego, fear, or a grudge instead of their stated incentives, the payoff table stops predicting anything. I'd rather the skill flag that limit explicitly than let someone walk away with false confidence from a clean-looking Nash Equilibrium that assumes a rationality the other side never had.
The table makes sure you're arguing with the actual game instead of the one you assumed you were in.
The anatomy of the skill
The game-theory skill is a structured, interactive coaching protocol designed to guide users through inter-dependent decision scenarios using Professor Paul Pfleiderer's Stanford GSB game theory framework.
Header Metadata & Invocation Constraints
Trigger Conditions: Activates when outcomes depend on an identifiable counterpart's reactions, explicitly looking for triggers like "it depends on what they do" or "should I move first or wait".
Exclusion Boundary: Explicitly forbids invocation when outcomes depend purely on impersonal uncertainty, routing those scenarios to a standard decision tree skill instead.
Persona & Core Discipline
Role Definition: Frames the AI as a Socratic coach who asks focused questions rather than filling in data for the user.
Core Philosophy Quote: > "The other party is not an obstacle — they are a decision-maker with their own objectives and incentives. You always analyze the situation from both sides of the table."
The 5-Phase Analysis Protocol
Phase 0 (Map the Strategic Situation): Uncovers players, choices, and payoff structures, culminating in a mandatory two-player table. Quote: "Confirm the table before proceeding. The analysis is only as good as the payoff structure."
Phase 1 (Eliminate Dominated Strategies): Systematically strips out inferior moves for both parties. Quote: "The first question in any strategic situation is whether any of your options — or theirs — can be ruled out immediately..."
Phase 2 (Cooperation vs. Reneging): Evaluates joint value against defection incentives. Quote: "Before entering a cooperative arrangement, ask: what is the mechanism that keeps both sides honest?"
Phase 3 (First-Mover Advantage): Calculates structural timing advantages versus simultaneous moves. Quote: "Moving first only creates advantage if the other party believes you're truly committed — that you won't change course after they respond."
Phase 4 (Nash Equilibrium): Identifies stable landing points when prior phases leave unresolved choices. Quote: "A Nash Equilibrium is a combination of choices... where neither side wants to change their choice given what the other is doing. It's the stable landing point..."
Coaching Principles & Guardrails
Enforced Analytical Rules: Commands the AI to always construct payoff tables before analyzing, account for non-monetary factors like reputation, and differentiate between one-time vs. repeated interactions.
Rationality Warning Quote: > "Rationality is an assumption, not a guarantee. The analysis assumes the other party acts in their best interest. In high-stakes business contexts, that's usually safe. In lower-stakes or emotional contexts, it may not be."
Want to put this + and the rest of the decision-making skills to work on something you're actually facing? I am giving away the skills for an one-hour discovery interview. Sign up here.
