Anthropomorphism in AI raises diverse risks and ways to reduce them
Anthropomorphism in the age of Large Language Models: An overview of potential risks and mitigations
Computers and SocietyComputation and LanguageHuman-Computer Interaction
Summary
People tend to think of AI systems like Large Language Models as if they were human because of how these systems use language. The authors review different studies about why this happens and what problems it can cause, such as misunderstandings about AI’s capabilities or impacts on human decision-making. They organize the risks into categories like knowledge errors, emotional effects, and societal concerns. Finally, the paper discusses ways to reduce these risks through better design, communication, education, and policy-making.
What this means in practice
- •For ai product designers: Improve AI user interfaces by avoiding human-like language that might mislead users about AI capabilities and risks.
- •For policy makers: Develop communication guidelines and governance frameworks that reduce misunderstandings caused by anthropomorphic descriptions of AI.
A survey. It maps existing work.
Authors
Ismael T. Freire, Marceau Nahon, Maud van Lier, Katie Evans, Hélie Bazin, Michele Farisco, Kathinka Evers, Raja Chatila, Mehdi Khamassi
Abstract
Large Language Models (LLMs) and more broadly Artificial Intelligence (AI) systems are often described and understood in human-like terms, a phenomenon known as \emph{anthropomorphism}. This paper provides a synthesis of recent literature on anthropomorphism in AI, covering theoretical frameworks, the role of language in framing AI as human-like, the various risks of anthropomorphizing machines, and strategies to mitigate these issues. After examining why we tend to anthropomorphize AI systems and whether we are right to do so, we highlight the impact of linguistic framing on anthropomorphism. Then, we introduce a conceptual taxonomy of risks associated with AI anthropomorphism. This taxonomy groups twenty-one concerns within five analytical categories: epistemic, affective, human agency, normative, and societal and institutional risks. Finally, we relate these concerns to proposed interventions in design, communication, education, and governance. We argue that a better understanding of AI systems requires concepts and theories grounded in their organization and demonstrated capacities. The linguistic shaping of anthropomorphic perceptions should form part of this scientific effort, since our descriptions influence both how these systems are understood and the roles we allow them to occupy in society.