Proceedings of the Annual Meeting of the Cognitive Science Society

As artificial intelligence (AI) becomes increasingly integrated into work and daily life, it is crucial to understand how people decide which tasks to delegate. We examined how AI descriptions influence delegation in arithmetic and social tasks. Participants either completed tasks themselves or offloaded them to one of two AIs. In Experiment 1, the AIs were described as human-like and machine-like, and participants were informed that both performed the tasks equally well. Social tasks were more often delegated to the human-like AI, and arithmetic tasks to the machine-like AI. Open-ended responses indicated that human-likeness was associated with emotional competence, and machine-likeness with logical competence. In Experiment 2, AIs were described as emotionally competent or logically competent, and the same task-specific delegation pattern emerged. These findings suggest that broad labels are spontaneously interpreted in terms of task-relevant competence.