From Confidence to Catastrophe: A Statistical Framework for Assumption Bias Mitigation in Human-AI Collaborative Decision-Making

Introduces a framework for reducing assumption bias in AI-assisted decision-making, showing how humans often misinterpret probabilistic AI outputs as certainty while AI systems fail to assess contextual completeness. Drawing from clinical case studies and forecasting research, the framework demonstrates how multi-source verification, base-rate analysis, and adversarial review can significantly reduce errors in high-stakes domains ranging from healthcare to autonomous systems.