Trust in AI: A Decision-Theoretic framework for Human–AI Interactions in the Digital Society

Trust in artificial intelligence (AI) is crucial for ensuring productive human–AI interactions across domains ranging from medicine to finance. As AI systems become central in the digital society, understanding how people decide when to rely on algorithmic advice becomes increasingly important. This paper proposes that a decision-theoretic perspective is needed to explain how well-calibrated trust in AI can be achieved. We conceptualize trust in AI as a decision to delegate control to an artificial agent, balancing potential rewards and risks of following AI advice versus relying on one’s own judgment. Framed this way, established decision-making theories such as Prospect Theory, Evidence Accumulation Models, and Reinforcement Learning offer valuable insights into the cognitive processes that underlie trust in AI. These theories offer formal, computational models to explain how individuals evaluate AI-generated advice, weigh risks and rewards, form biases, and update trust based on feedback, enabling precise hypotheses about when and why people accept or reject AI advice. By grounding trust in cognitive theory, we offer a roadmap for empirical research that moves beyond surface-level behaviors and toward mechanistic understanding. We conclude that a decision-theoretic approach can enhance psychological insight and guide the responsible design of AI systems within the digital society that are better aligned with human cognition.