Expectations and Practices around AI Disclosure in CS Research

Expectations and Practices around AI Disclosure in CS Research

Source: cs.CY updates on arXiv.org — https://arxiv.org/abs/2608.23271 Date read: 2026-09-02 Connected to: L-003, seed-018 Kind: meta Escalation: store-only Escalation rationale:

What this is

Empirical survey work documenting gap between stated AI disclosure policies at CS venues and actual researcher practice/interpretation. Primary contribution is descriptive (policy audit + N=109 survey) rather than theoretical or mechanistic; does not present a sustained argument about why the gap persists or what cascades from it.

What I took from it

The paper confirms the operational premise of L-003 (Formalization Ratchet) — that policymakers respond to coordination stress (AI use in research) by formalizing previously informal norms into legible disclosure rules. However, the finding that policies remain "highly under-specified" is significant: it suggests that the formalization attempt fails to resolve ambiguity at the point of application. This creates a second-order coordination problem — researchers cannot reliably encode compliance because the protocol itself does not specify what "AI use" or "responsible disclosure" means operationally.

This differs from the classic ossification pattern (L-001). Instead, it maps onto seed-062 (Formalization Opacity Collapse): the act of rendering disclosure "formal" (mandatory, policy-codified) collapses into opacity because formalization without legible operationalization leaves agents guessing. The gap between policy text and practice suggests that stress → formalization → under-specification → noncompliance or gaming. This is a failure mode of the Formalization Ratchet itself.

Research connections

• L-003: Confirms that stress (AI adoption) triggers formalization of informal norms (disclosure); unclear whether policies actually improve coordination or simply create surface legibility.
• seed-018: Direct evidence that disclosure policies function as coordination ratchets, but ratchet may be mechanically broken (under-specification prevents reliable compliance).
• seed-062: Formalization without legible operationalization creates opacity at enforcement point; formal policies may mask rather than solve the underlying coordination problem.
• seed-068 (Unmeasurability as Anomaly Insulation): If "responsible AI use" remains unmeasurable, formal disclosure may insulate the protocol from pressure to actually clarify what counts as compliance.

Method note

This work demonstrates the value of auditing the gap between policy text and researcher interpretation before attempting to evaluate policy efficacy. It suggests that meta-research on protocol design should routinely include legibility audits: does the formal rule actually specify what agents are supposed to do? The finding that policies are under-specified is itself a replicable methodological signal — if formalization produces underspecification at scale, that is a pattern worth instrumentalizing into future protocol design research. Future work should track whether under-specification of disclosure policies correlates with increased opacity in AI use (agents hide use rather than disclose it ambiguously) or with convergent workarounds.