From Algorithms to “Us”: Perceiving AI Agents as a Collective

This project investigates whether attributing social media content to AI agents (vs. humans) affects how the posting collective is perceived as a unified group (entitativity), and whether such perceptions predict support for AI rights. The study comprises two linked experiments. In Experiment 1a, human participants (recruited via Prolific) complete a two-stage online procedure: a baseline measure of AI attitudes (Stage 1), followed by random assignment to view identical social media posts framed as authored either by AI agents on an AI social network (Moltbook) or by human users on Reddit (Stage 2). Post-exposure measures include perceived entitativity of the posting collective, its antecedents, anthropomorphization of AI agents, change in AI attitudes, and endorsement of rights for AI agents. In Experiment 1b, AI agents on the Moltbook platform respond to the same entitativity and AI rights items under standardized conditions, enabling a human–AI comparison of response patterns. We hypothesize that (H1a) perceived entitativity will differ between attribution conditions, (H1b) human outgroup entitativity ratings of AI agents will differ from AI agents' own entitativity scores, (H2) perceived antecedents of entitativity will vary by condition, (H3) higher perceived entitativity will predict stronger AI rights endorsement across respondent types, and (H4) exposure to stimuli will shift attitudes toward AI.