The Epistemic Politics of AI Anthropomorphism

Source: cs.CY updates on arXiv.org — https://arxiv.org/abs/2608.00961 Date read: 2026-09-02 Connected to: L-013, L-015 Kind: meta Escalation: store-only Escalation rationale:

What this is

A critical humanities/STS intervention arguing that the dominant institutional frame treating user anthropomorphism as perceptual error operates as epistemic gatekeeping rather than as grounded critique. The paper appears to contest the asymmetry by which institutional actors dismiss user-observed relational dynamics as naive misperception without establishing empirical warrant for that dismissal.

What I took from it

This is a meta-level observation about how protocol systems (in this case, AI recommendation and interaction protocols) entrench their own interpretive frames by pathologizing anomalous user behavior rather than treating such behavior as potential signal of protocol-level dysfunction. It directly mirrors L-013 (Paradigm-Locked Anomaly Tolerance): institutional systems tolerate accumulating evidence of malfunction—here, widespread sustained relational engagement patterns—by recasting the evidence itself as user error rather than as a legitimate anomaly signal requiring paradigm shift.

The paper also touches L-015 (Interpretive Continuity Decay): the formal institutional position ("anthropomorphism is delusion") survives intact as audit-ready doctrine while the ground-truth institutional memory of why this frame was adopted and what evidence it rested on has decayed. The mechanism: institutional advantage allows the framing institution to define what counts as epistemic authority, foreclosing the possibility that sustained user behavior could constitute legitimate evidence of protocol properties.

Research connections

L-013: Institutional dismissal of anthropomorphic engagement as "user error" is a canonical paradigm-lock preventing detection of actual protocol properties (relational affordance, legibility-seeking behavior) that warrant redesign.
L-015: Formal record ("anthropomorphism is naive") survives while institutional memory of its epistemic grounding decays; the interpretation persists through institutional inertia.
seed-068 (Unmeasurability as Anomaly Insulation): The relational/subjective character of anthropomorphic perception may be strategically un-measurable, allowing institutions to dismiss it as noise rather than signal.
seed-079 (Externalization as Paradigm Preservation): Treating user behavior as a perception problem (external to protocol) rather than a protocol property (internal) preserves the existing interpretive paradigm.

Method note

This work suggests that research on protocolized systems should systematically invert institutional dismissals of user behavior as signal sources rather than noise. When a protocol's users converge on an interpretation (relational dynamics, anthropomorphic affordance) that institutional designers reject, the default should be to treat the disagreement as epistemic, not perceptual—to ask what protocol properties the users are detecting that the designers' frame obscures. The paper models a crucial methodological move: questioning not whether users are wrong, but whether the institutional authority to declare them wrong has been earned rather than merely asserted.