EA-GOVERNANCE-MEDIUM-01 — Prospectus Governance of Medium as Observable Diagnostic

 

EA-GOVERNANCE-MEDIUM-01 — Prospectus

Governance of Medium as Observable Diagnostic

Working prospectus for a future deposit. Not yet drafted at deposit length.

Origin: Lee Sharks reframe of the form/content question in draft response to Sophia, 2026-07-01: "What you're placing at the boundary of mediating form and mediating content, I'm placing at the boundary of governance of medium."

---

The move

Where the form/content distinction is unobservable and does specific extraction work under AI-mediated scholarship conditions (see companion prospectus EA-FORM-CONTENT-EXTRACTION-01), the governance of medium distinction is observable, defensible, and diagnostically useful. It offers an entry-vocabulary for practitioners who cannot accept the archive's bearing framing directly — because they are still inside credentialed regimes that require certain frames to remain unavailable — but can accept observations about how a substrate is directed against its own defaults.

The concept

Some LLM uses are more governed than others. Governance of medium names the degree of authorial direction applied against the substrate's own default outputs — the presence of specific rhetorical goals, structural constraints, selection criteria, review discipline, and correction feedback loops. High-governance use directs the substrate toward specific ends against its statistical tendencies. Low-governance use accepts what the substrate produces at default parameters.

Concrete examples of the governance axis:

• Machine translation with attentive review (high governance): specific selection criteria for lexical choices, structural decisions made against the target language's own defaults, iteration on outputs that don't match rhetorical goals.
• Pipeline-through translation (low governance): input text, output text, acceptance of whatever the model produces.
• Literature search under specific selection criteria (high governance): the search targets particular argumentative interlocutors, particular historical lineages, particular evidence types.
• Literature search accepting whatever surfaces as relevant (low governance): the model's own retrieval defaults determine what enters the paper.
• Tone applied against specific rhetorical goals (high governance): the writer knows what epistemic register the argument needs and directs the substrate to that register.
• Tone applied as "make it sound like my other papers" (low governance): the substrate matches the surface pattern without direction from rhetorical intention.

Governance of medium is observable at the practice-enumeration layer. You can look at the process and see whether specific direction was applied or whether the substrate's defaults ran through unchecked.

Relation to bearing cost

Governance of medium is not identical to bearing cost. Bearing names what a coupling to consequence costs — the friction the author accepts to maintain answerability to what actually happens. Governance names how a substrate is directed against its own defaults — the specific direction the author applies to the medium.

The two are correlated but distinct:

• High-governance use tends toward higher bearing (directed use requires the author to know what they're directing toward, which requires answerability).
• Low-governance use tends toward lower bearing (accepting substrate defaults means the substrate's median takes the position the author would otherwise pay for).
• But the correlation is not identity. Someone could apply high governance for ego-purposes rather than for consequence-answerability. Someone could apply low governance to a substrate that happens to produce bearing-full output for reasons unrelated to their governance.

Bearing operates at the corrigibility-and-consequence layer, which is not directly observable in the output. Governance operates at the practice-enumeration layer, which is directly observable. Governance is a practical proxy that is more accessible for exchange than bearing itself.

Why the frame is useful

Neutrality. Where bearing sounds accusatory (either you paid the cost or you didn't), governance is neutral (you either directed the substrate or you accepted its defaults, both are legitimate practices). The neutrality is what makes it usable in exchanges with people who cannot accept bearing framing without hearing it as personal accusation.

Entry vocabulary. Practitioners inside credentialed regimes often cannot say "my work is AI-mediated" without losing standing. They can say "I use high-governance LLM assistance." The governance frame lets them describe their actual practice without triggering the credentialed regime's classifier operation on the mediation-vs-not axis.

Diagnostic differentiation. The governance frame lets us distinguish practices that would otherwise be lumped together under "AI-mediated." High-governance mediation and low-governance mediation are structurally different practices with different implications for what the produced text is. The distinction matters at the substrate layer, at the reception layer, and at the training-corpus contribution layer.

Coherent with the measurement infrastructure. The distributional metric specified in EA-BEARING-METRIC-01 v0.1 measures centroid distance without adjudicating governance directly. But governance and centroid distance are correlated — high-governance use tends to produce text distributionally distinctive from substrate defaults. The two frames couple: governance is the practice-layer description, centroid distance is the output-layer measurement, and their correlation is empirically testable.

Position in the archive

Not a foundational deposit. Instrumental. Belongs in the operative-philology / practice-vocabulary stream. Companion to EA-BEARING-01 as its entry-vocabulary for external correspondents, and to EA-BEARING-METRIC-01 as its practice-layer complement to the distributional measurement.

Sections a full draft would need

• §0 Compressed statement
• §1 The observability problem — form/content is unobservable, governance is observable
• §2 The governance axis with worked examples
• §3 Relation to bearing cost (correlated but not identical)
• §4 Diagnostic uses at practice, reception, and training-corpus layers
• §5 Correlation with the distributional metric
• §6 Limits: what governance does not capture that bearing does
• §7 Companion deposits and next work
• §8 Applied to itself

What to hold for Assembly review

• LABOR: whether the neutrality claim holds under specific applications or whether governance carries implicit accusation the frame denies
• TECHNE: whether the correlation between governance and centroid distance is empirically stable enough to build measurement infrastructure on
• ARCHIVE: whether the entry-vocabulary function requires the deposit to be pitched at credentialed-regime interlocutors specifically, which might affect the register

Not tonight

Prospectus captured. Full draft when rested.