From Statements to Questions: AI as Epistemic Repair Infrastructure

Author: Rowan Brad Quni-Gudzinas / QNFO Version: 2026-09-26 Format: printable HTML poster (A0 landscape, A0 portrait, 16:9 digital variants) + this markdown source.

Thesis

Human conflict is not always caused by misunderstanding, but it often unfolds through partial understanding, misframing, and failed repair. LLMs can help by converting fixed statements into testable questions, generating adversarial perspectives, and making hidden assumptions visible. Their value is not that they provide truth, but that they expand the space of what participants know they do not know.

Abstract

Human conflict often begins not with a lack of speech, but with speech that has hardened into statements. Statements can conceal the assumptions that make them possible: what one side takes as perception, the other experiences as accusation. This poster proposes a practical framework for using large language models as tools of negative epistemology: not to decide who is right, but to surface what participants do not know they do not know. The framework distinguishes epistemic conflict, interpretive conflict, and structural conflict, arguing that AI is most useful in the first, partially useful in the second, and only indirectly useful in the third. A statement-to-question protocol is introduced for extracting hidden assumptions, generating adversarial perspectives, identifying falsifiers, and deciding whether a situation calls for repair, negotiation, boundary-setting, or enforcement. The poster also addresses dual-use risks: sycophantic validation, manipulative perspective modeling, false reconciliation, and synthetic evidence. The central claim is that LLMs can make epistemic repair cheaper, but only if they are used to interrogate statements rather than decorate them.

Poster sections

1. Magritte framing: what is hidden by what we see / say.
2. Problem: humans default to statements rather than questions.
3. Three-layer model: epistemic, interpretive, structural conflict.
4. AI mechanisms: assumption extraction, perspective generation, unknown-unknown mapping, repair rehearsal.
5. Statement-to-question protocol.
6. Dual-use risks and limits.
7. Takeaway: AI is not an arbiter of truth; it is a cheap instrument for epistemic repair when used adversarially against one's own framing.

S-Q protocol

• State the belief: "They are acting in bad faith."
• Extract hidden assumptions: what evidence distinguishes bad faith from fear, incentives, confusion, or constraint?
• Generate the strongest alternative frame: how might they describe this conflict without lying?
• Identify falsifiers: what observation would change my view?
• Choose repair or boundary: is this epistemic, interpretive, or structural conflict?

Risk box

AI can produce fluent false reconciliation, intensify manipulation by modeling vulnerabilities, degrade verification through synthetic evidence, and validate grievance through sycophancy. Some conflicts require boundary, exit, enforcement, or protection rather than dialogue.

Artifacts

• R2: posters/statements-to-questions-ai-epistemic-repair/ (poster.html, poster-a0-landscape.html, poster-a0-portrait.html, poster-16x9.html, companion.html, manifest.json)
• Repository: QNFO/qnfo-workers -> docs/posters/statements-to-questions-ai-epistemic-repair/

Status note

Template-independent provisional. The poster-session acceptance/session specification was not verifiable in the available QNFO mailbox as of 2026-09-26; final dimensions/template remain conditional on that document.