Machine-Learning to Trust

2025-07-14

AI Summary

This paper investigates how boundedly rational agents sustain trust in infinitely repeated interactions under a machine learning–inspired belief formation mechanism that penalizes model complexity. We develop an infinite-horizon dynamic game in which agents possess only one-period memory and decide whether to trust successors under state-dependent trust costs. Departing from standard equilibrium expectations, we introduce a “coarse-fitting” belief strategy that balances predictive accuracy against model simplicity—formalizing cognitive constraints via a trade-off between prediction error minimization and complexity regularization. Results show that this mechanism substantially shrinks the equilibrium set supporting trust and markedly reduces cooperation sustainability compared to symmetric mixed-strategy Nash equilibria. The key contribution is identifying cognitive simplicity—not mere bounded rationality—as a structural impediment to long-term trust, thereby establishing a novel machine learning–based paradigm for modeling belief formation in behavioral game theory.

Technology Category

• Multiagent Systems: Mechanism Design
• Game Theory and Economic Paradigms: Mechanism Design
• Knowledge Representation and Reasoning: Reasoning with Beliefs

Application Category

• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Can players sustain long-run trust when their equilibrium beliefs are shaped by machine-learning methods that penalize complexity? I study a game in which an infinite sequence of agents with one-period recall decides whether to place trust in their immediate successor. The cost of trusting is state-dependent. Each player's best response is based on a belief about others' behavior, which is a coarse fit of the true population strategy with respect to a partition of relevant contingencies. In equilibrium, this partition minimizes the sum of the mean squared prediction error and a complexity penalty proportional to its size. Relative to symmetric mixed-strategy Nash equilibrium, this solution concept significantly narrows the scope for trust.

Problem

• Study trust sustainability with machine-learning shaped beliefs
• Analyze equilibrium behavior in sequential trust games
• Examine complexity-penalized partitions' impact on trust scope

Innovation

• Machine-learning penalizes complexity in beliefs
• Coarse fit partitions minimize prediction error
• Equilibrium narrows trust scope significantly