Abstract

The increasing use of artificial intelligence (AI) in decision-support systems has reshaped human performance in complex problem-solving, yet its effects on confidence, trust, and accuracy remain unclear. This study investigates how AI generated feedback influences decision accuracy, confidence, and reliance behaviour using 3,400 trial-level observations from 100 participants in a chess-based dataset. Results show that decision accuracy improves significantly after AI feedback (mean increase from 0.43 to 0.82), particularly under accurate feedback conditions. While misleading feedback reduces post-decision confidence, overall confidence shifts remain small (mean = 0.08). Reliance on AI is strongly associated with improved performance (r ˜0.66). Machine-learning models further confirm that trust behaviour and feedback accuracy, rather than confidence, are the strongest predictors of decision outcomes.

Authors

Omolola Mariam Oparinde¹, John Olalere Ogunlola², Oluwakemisola Adewole³, Bisola Kafayat Oyebamiji⁴, Dan Taiye Aremu⁵, Blessing Alice Alao-Olatunji⁶
University of Greater Manchester, United Kingdom^1,2,3,4,6, London Metropolitan University, United Kingdom⁵

Keywords

Artificial Intelligence, Calibration, Confidence, Decision Accuracy, Feedback, Human–AI Interaction