Biased Error Attribution in Multi-Agent Human–AI Systems Under Delayed Feedback

Human decision-making is often influenced by cognitive biases, especially under uncertainty and risk. Prior works have studied biases in single-step decision environments with immediate outcomes or in interaction with a single AI agent. However, less attention has been given to (1) cascaded settings with delayed outcomes, where decision at one step influences future system states and (2) to interaction with multiple AI agents. To study this, we designed a naval battle strategy game where participants interact with two AI advisors. We find asymmetric responses to positive and negative outcomes, with significant changes after negative outcomes. Participants failed to identify the actions responsible for failures and misattributed responsibility across AI agents, leading them to modify decisions steps weakly related to the true cause of poor performance. We refer to this behavior as bias in error attribution. Our findings show how cognitive biases can emerge in human-multi AI teams in cascaded settings.