There Is No Agency Without Attention

ences of the Artificial with a parable about an ant making her way home across a beach and comments that the complexity of her route is not a manifestation of any complex, goal-seeking behavior within the ant but rather reflects the complexity of the environment. The ant “must adapt [her] course repeatedly to the difficulties [she] encounters and often detour uncrossable barriers.” The question that interests us is not whether the ant is complex, but whether she should be considered an intelligent agent. The ant appears to have a goal to get home, and she exhibits control in her ability to navigate novel and complex terrain. Her control is grounded in a coupling of her perceptions to her actions that keep her from continually bumping into an obstacle. Nevertheless, calling the ant an intelligent agent seems like overzealous anthropomorphism on our part. Why is that? Within the artificial intelligence (AI) community, this question is not new. For decades AI researchers have built agents that are capable of carrying out tasks that require human-level or humanlike intelligence. During this time, questions of how these programs compared in kind to humans have surfaced and led to beneficial interdisciplinary Articles