When expert recommendation contradicts peer opinion: Relative social influence of valence, group identity and artificial intelligence

Whom do we trust more, the recommendation of an expert or public opinion from a crowd of other users of the site? Does it matter if the expert belongs to our in-group? And, what, if anything, would change if an Artificial Intelligence (AI) system was the recommender rather than a human expert? In order to answer these research questions, we conducted a between-subjects online experiment, informed by MAIN Model (Sundar, 2008), which posits that interface cues signaling different types of sources can influence perceived credibility of content by triggering distinct cognitive heuristics. Participants were assigned to a scenario wherein the expert review contrasted the peer rating about recommending photos for business profiles, with systematic variations in expert review valence (negative vs. positive), expert identity (ingroup vs. outgroup vs. no identity), and agent type (human vs. AI). Results show that positive ratings are more influential on user judgements. However, for negative ratings, human ingroup members generated greater effects than no-identity experts. Moreover, AI systems were as influential as human experts, suggesting the potential for AI to substitute human experts for online recommendations.