No Certificate, No Categorical Speech Act: A Brouwerian Assertibility Constraint for Public Reason

No Certificate, No Categorical Speech Act: A Brouwerian Assertibility Constraint for Public Reason

Source: cs.CY updates on arXiv.org — https://arxiv.org/abs/2603.03971 Date read: 2026-06-24 Connected to: none Escalation: escalate-to-deep Escalation rationale: This proposes a sustained mechanism for constraining artificial speech authority through epistemic certification — absent from current inventory — and generalizes across high-stakes domains as a governance principle for protocolized systems.

What this is

A normative-technical proposal grounding the governance of AI utterance in constructive logic (Brouwerian intuitionism). The core argument: generative systems conflate confidence with justification, producing "hypersuasive" categorical claims without warrant. The corrective is a three-status interface requiring publicly inspectable certificates of entitlement (proof, audit trail, or delegated authority) before assertion — otherwise systems must return Undetermined rather than risk false authority.

What I took from it

This work identifies a mechanism of authority-slippage in artificial systems: the conversion of probabilistic output into categorical speech acts, which displaces rather than distributes epistemic responsibility. The Brouwerian constraint is not merely a UI pattern but a structural principle — it makes undecidability a first-class speech status rather than a failure mode, and it ties this back to democratic epistemic agency (the human-side dependency for legitimacy).

What's novel: the work treats the certificate itself as the unit of governance, not the model or the output distribution. This suggests that protocolized systems generate legitimacy through publicly contestable warrant structures, not through accuracy or confidence calibration alone. This has implications beyond AI: it's a principle for any automated decision or utterance system operating in domains where epistemic authority matters (medicine, law, public policy).

The paper doesn't argue for abstention or silence; it argues for epistemically honest interfaces that separate what can be warranted from what cannot.

Research connections

• Epistemic authority in artificial systems: How do protocolized systems acquire and maintain standing to make claims in public domains?
• Speech-act boundaries in AI: Where does probabilistic inference end and justified assertion begin? What structures enforce that boundary?
• Governance through constraint architecture: Can normative constraints be embedded as interface requirements rather than post-hoc regulation?

Candidate laws or signals

• CL-2603-1: Artificial systems that assert categorical claims without publicly inspectable warrant structures displace epistemic responsibility rather than distribute it, weakening downstream democratic agency.
• CL-2603-2: Undecidability as a positive status (not a failure state) is structurally necessary in high-stakes protocolized speech — systems that suppress it risk authority capture.
• CL-2603-3: Legitimacy of automated utterance correlates with the contestability and transparency of its warrant structure, not solely with output accuracy.