Seeing is believing: how (X)AI-augmented advice guides human evidence accumulation in perceptual decision-making

(Explainable) Artificial Intelligence or (X)AI advice can affect human cognitive processes during decision-making in various ways. In perceptual decision-making, there is a risk of users accepting (X)AI advice without further consideration, to save time. We draw conceptually upon evidence accumulation models and information foraging theory to answer whether AI and XAI encourage shallow reliance on advice or facilitate more efficient verification of advised choices. In a laboratory experiment, we investigate the evidence accumulation process of decision-makers during an anomaly-detection task in manufacturing, using eye-tracking technology. Both decision accuracy and total decision time increase from human-only to AI and further to XAI support; in both cases the difference is statistically significant for XAI advice and directional but not significant for AI advice. This pattern favors verification over shallow reliance: advice is checked against the stimulus rather than used to save time. When looking at eye-tracking metrics, we surprisingly find that people do not get distracted by the XAI advice and primarily focus on the original task-based stimulus, while their individual fixations become shorter. Overall, these findings suggest that XAI acts as a verification scaffold: in advice-augmented perceptual decision-making, explanations guide where to check rather than supplanting inspection of the primary evidence.