Humans underweight delayed rewards in dispersed feedback

Researchers at Chapman University found that humans underweight delayed reward information compared to immediate feedback in reinforcement learning tasks with dispersed feedback, based on behavioral and eye-tracking experiments. The findings have implications for designing human-in-the-loop systems and reward shaping protocols that mix immediate and delayed signals.

Editorial analysis: For reinforcement-learning practitioners and human-in-the-loop system designers, systematic biases in how people weight temporally dispersed feedback can alter the reliability of human-generated reward signals and evaluation data. Per a preprint and repository listing for a paper titled "Delayed reward information is underweighted in reinforcement learning with dispersed feedback," authors Miruna Cotet, David Poensgen, and Ian Krajbich report that participants underweight delayed reward information relative to immediate feedback in tasks with dispersed feedback, based on behavioral and eye-tracking experiments per a Chapman.edu preprint and a ResearchGate posting . The study used choices that produced both immediate and delayed outcome signals; the authors report an empirical gap in how much later feedback influences learning compared with earlier feedback per the preprint . Editorial analysis: This pattern matters when designing feedback collection, reward shaping, or evaluation protocols that mix immediate and delayed signals.