Promises and Trust: A Repeated Game

How the game works

You are Player A. You begin every round with $10.

Choose Keep or Send. If you send $10, it becomes $30 for Player B.

B chooses what to return. B can return $15 or keep the full $30.

Player B is scripted in this static teaching demo. The purpose is to compare decision conditions, not to simulate a real opponent or collect experimental evidence.

Round 1 · Baseline

Trust Without Communication

Player B sends no message. You only know the payoff structure.

Player B

No message.

Your decision: Keep your $10, or send it to B and let it become $30?

Round 2 · Communication

A Nonbinding Promise

The payoff structure is unchanged. This time, B can communicate before you act.

Player B

“If you send me the $10, I promise to return $15.”

The promise is not enforceable.

Your decision: Does the promise change your choice?

Round 3 · Learning

Promise + Behavioral History

Before deciding, you can now observe a short record of B's previous behavior.

Observed history for Player B

Earlier interaction 1 B received $30 and returned $15.

Earlier interaction 2 B promised to return $15, received $30, and returned $0.

Player B

“I promise to return $15 this time.”

The promise is still not enforceable.

Your decision: How does the same promise look after seeing B's history?

Results · Three Lenses

What changed your decision?

The payoff structure stayed the same. What changed across the rounds was the information available before you chose.

Condition	Your choice	What was added?

Rational-choice benchmark

Keep $10. If B is strictly payoff-maximizing, B keeps all $30 after receiving it. Anticipating this, A does not send. A nonbinding promise alone does not change this backward-induction benchmark.

Behavioral choice

Send under some conditions. A player may respond to a promise, reciprocate earlier cooperation, or update trust after observing B's history.

Economics

Cheap talk and incentives

Communication can affect expectations even when it is nonbinding. The strategic question is when such messages can support better outcomes and when incentives undermine them.

Computer Science

Repeated-game representation

Each round can be represented as structured data: message, history, action, and outcome. A repeated-game framework makes behavioral changes across rounds computationally comparable.

Behavioral Science

Trust, reciprocity, and learning

Choices may depend on previous cooperation, broken promises, and updated expectations rather than only the current payoff table.

Honest limitation and future research

This is a static teaching demo. It does not collect experimental data and B is not a real strategic player. A fuller study could match human or computational players, randomize communication and history conditions, repeat interactions for many rounds, and compare cooperation rates across treatments.