Conformity towards humans versus AI in different task domains: the type of task matters

Generative AI (GenAI) excels at subjective tasks, challenging traditional views of machine intelligence. This study investigated if public perception reflects this reality by measuring conformity to AI or a humans. An online experiment (N=245) measured conformity across six tasks on a subjective-objective spectrum (e.g., morality, knowledge, counting). In a counterbalanced design, participants received competing suggestions from an "AI" and a "human peer group." We then assessed whether conformity was moderated by prior GenAI experience and participants' Perceived Task Objectivity (PTO) ratings. Conformity behaviour followed traditional stereotypes. A preference for human advice in aesthetics, morality and opinion tasks, while AI was trusted more in counting and knowledge tasks. Prior GenAI experience did not increase trust in AI's subjective abilities, and PTO was not a reliable predictor of conformity. Despite GenAI's creative proficiency, public trust remains anchored in a traditional human-machine dichotomy, revealing a gap between AI's evolution and user perception. The task's type strongly influenced conformity, but PTO's limited predictive power suggests the subjective-objective dichotomy is insufficient to explain the nuances of human-AI trust.