Does It Feel or Does It Understand? Communication Style and Task-Contingent Mind Attribution to Conversational AI

It is unclear what consumers infer from the social-oriented (SCS) and task-oriented (TCS) communication styles of conversational AI in e-commerce, or why these associations vary across tasks. Drawing on mind perception research, we examine affective (AH) and cognitive (CH) human-likeness, defined as attributions of human-like affective experience and cognitive understanding. Study 1 (N = 400 and 500) shows that AH and CH are distinguishable from warmth, competence, emotional and cognitive intelligence, experience, agency, and each other, and jointly explain additional variance in preference. Study 2, a three-wave survey of 415 users, measured SCS, TCS, and perceived task demands at Wave 1, AH and CH at Wave 2, and preference at Wave 3; interaction tests are therefore single-occasion. SCS is positively associated with AH and CH, and TCS with CH; both are positively associated with preference. The SCS–AH association is stronger under higher perceived social demands, and the TCS–CH association under higher perceived analytical demands. Analytical demands appeared to attenuate the SCS–AH association more than social demands attenuated the TCS–AH association, but this contrast rests on a single uncorrected test. These findings suggest that the associations of communication style with AH and CH vary across task demands.