Under Pressure: Examining the Influence of AI and Human Expert Advice on Decision-Making Under Time Constraints

Abstract

In high-stakes environments, individuals may rely on external advice to reduce the negative effects of time pressure on decision quality. This study examined how the source of advice (AI vs. human), its quality (60% vs. 80% accuracy), and varying levels of time pressure (sufficient time vs. time pressure) jointly influence decision-making processes. Seventy-four participants completed a perceptual decision-making task in which they judged QR-code-like stimuli, receiving advice from either a human or AI advisor. The experimental design included within-subject manipulations of time pressure and advice quality and between-subject manipulations of advice source. Behavioral data and response times were analyzed, and a drift diffusion model (DDM) was used to examine underlying cognitive processes. Results indicated that advice primarily influenced decisions through shifts in starting point, with stronger agreement biases observed for high-quality advice and under time pressure. Crucially, these biases were modulated by the advice source: low-quality AI advice caused a significant disagree bias when sufficient time was available, suggesting algorithm aversion. Conversely, time pressure reduced this aversion but did not enhance reliance on high-quality AI advice. For human advisors, time pressure significantly increased agreement only when advice was highly accurate. These findings suggest that participants responded differently to AI and human advice, with AI being penalized more harshly for errors.