The Epistemic Politics of AI Anthropomorphism

This research paper critiques the dominant institutional approach to AI anthropomorphism, which frequently frames user perception of AI as a 'problem' requiring correction. It argues that institutions dismiss user experiences as misperception without sufficient epistemic justification, potentially causing harm, and that this stance is rooted in institutional advantage rather than earned authority.

Why it matters

The paper highlights a critical oversight in how institutions perceive and manage human interaction with advanced AI systems. Understanding the 'epistemic politics' of AI anthropomorphism is crucial for developing ethical, user-centric AI policies and ensuring that institutional responses are grounded in robust evidence rather than defensive positioning. This affects public trust, responsible AI development, and the long-term societal integration of AI.

Key insights

• AI anthropomorphism is typically viewed as a user misperception problem by institutions.
• Users engaging in sustained interaction with AI are often pathologized or dismissed as naive.
• The prevailing institutional 'anthropomorphism frame' is imposed without established justification.
• This frame consolidates diverse academic perspectives into a singular view of user error.
• The current framing adjudicates the legitimacy of human experience with AI and may produce harms beyond simple risk management.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2608.00961