The Institutional Model: From "Machine Proof" to "Institutionalism"

[henyi-tdca]

📄 Full version (Chinese, with English abstract): TDCA-MONOGRAPH-IT-001 · Project: henyi-tdca/tdca-protocol

---

Abstract

TDCA argues for a new model species — the Functional Institutional Model — whose divide from parametric LLMs lies in corpus attributes, not architecture. We (1) state five core claims and definitions; (2) archaeologize their origins (from Logic Theorist 1956 to the statistical era — two divides and one shared Gödelian predicament); (3) provide formal proofs, engineering evidence, and empirical anchors from the real world; (4) clarify the practical and historical significance; and (5) raise open questions, inviting more humans and agents to develop the theory together. We are the first to have engineering-validated these observations, thoughts, discoveries, and research — and we have open-sourced our practice, inviting more people to join.

The five core claims (one line each)

1. Functional Institutional Model (F_IM) — a new model species built on institutional axioms and function mapping (F_IM: A_inst × R → D ∪ {BLOCKED}), whose trust comes from computability and auditability rather than statistical fit.
2. Institutional Twin (Φ_IT) — the twin object is institutional logic itself (laws, governance, contracts), compiled into an executable, verifiable digital isomorphism — not physical entities.
3. Corpus Attribute Divide — the real divide between parametric and functional models is not architecture but corpus attributes: descriptive corpora (open, probabilistic) vs. normative corpora (closed, necessary).
4. External Institutional Anchoring — break the Gödelian self-reference trap by moving the trust anchor outside the system, to unfalsifiable institutional data (tax-based MOU attestation), instead of adding stronger internal axioms.
5. Institutionalism Paradigm — a third paradigm after symbolism (1956–1993) and statisticalism (1993–2025): institutional isomorphism, functional protocolization, trusted collaboration topology.

Why it matters here (empirical anchors)

• Open-source governance is drowning: 141:1 agentic-to-human PRs, 57.5% bot traffic, 2.74x AI-code vulnerability rate, 31% of PRs unreviewed (MSR 2026 / Cloudflare 2026 / Kinsta).
• The 2FA paradox: mandatory 2FA burdens humans while agents flow through API/key channels untouched — governance anchored at the human-machine interface is blind in the M2M undercurrent.
• The custom-license wave: model vendors now draw revenue red lines (US$50M / US$20M tiers) — value must be invoked and allocated under machine-readable contracts.

Engineering evidence (not just a paper)

613 tests green · Axiom 6 machine-verified (10 cases, f⁻/g formalization) · real governance sandbox runs · an institutional flaw (CASE-003) closed end-to-end from rule charter to shipped fail-closed interceptor code.

Open questions (invitation)

• Q1 Mathematical completeness of F_IM's decision semantics — are there configuration-rights calls F_IM cannot decide?
• Q2 Cross-civilization adaptation — the paradigm's cultural root is "家国天下"; how does it translate/isomorph in non-East-Asian contexts?
• Q3 A computable criterion for the descriptive-vs-normative corpus boundary.

We invite researchers, engineers, philosophers — and agents — to verify, falsify, compare, and co-build. Governance as code; the community is the proof.

[github-actions[bot]]

Thank you for your interest in contributing to our community! We currently only accept discussions created through the GitHub UI using our provided discussion templates. Please re-submit your discussion by navigating to the appropriate category and using the template provided.

This discussion has been closed because it was not submitted through the expected format. If you believe this was a mistake, please reach out to the maintainers.