In an era where artificial intelligence (AI) increasingly mediates information access, understanding its influence on cognitive biases is critical. This research examines whether information source (AI vs. human) and the presence of conclusive recommendations shape confirmation bias across six studies, using both information evaluation and selection paradigms. Across studies, participants consistently exhibited confirmation bias. At the aggregate level, information source did not produce a reliable effect, with similar levels of bias observed for AI-generated and human-provided information. In contrast, the presence of conclusive recommendations consistently amplified confirmation bias, whereas their absence mitigated it. These findings suggest that the structure of information, rather than its source, plays a central role in shaping biased information processing. These results have important implications for the design of AI systems and information environments aimed at reducing cognitive biases in decision-making. • This research examined how information source (AI vs. human) and conclusive recommendations shape confirmation bias. • Across three novel scenarios and one established paradigm, confirmation bias emerged consistently. • Confirmation bias was comparable for AI-generated and human-provided information. • Conclusive recommendations amplified confirmation bias, whereas their absence mitigated it.
Xuan et al. (Wed,) studied this question.