AI Automatons: AI Systems Intended to Imitate Humans

There is a proliferation of AI systems designed to mimic people’s behavior, work, abilities, likenesses, or humanness—systems we dub AI automatons. Individuals, groups, or generic humans are simulated to produce creative work in their styles, respond to surveys in their places, probe how they would use a new system before deployment, provide users with assistance and companionship, and anticipate their possible future behavior and interactions with others, just to name a few applications. However, the research, design, deployment, and availability of such AI systems have prompted growing concerns about a wide range of possible legal, psychological, social, and other types of harms. In this paper, we seek 1) to facilitate productive discussions about whether, when, and how to design and deploy such systems, and 2) to help chart the current landscape of existing and prospective AI automatons. To do so, we tease apart determinant design axes and considerations to aid reflections and deliberations about whether and how design choices along these axes could mitigate—or instead exacerbate—harms that the development and use of AI automatons might give rise to. Through a synthesis of related literature and extensive examples of existing AI systems intended to mimic humans, we developed a conceptual framework that foregrounds key axes of design variations and provides analytical scaffolding to foster greater recognition of a) the design choices available to developers and researchers, as well as of b) the possible ethical implications these design choices might have.