Skip to content

Context-Condition Testing

A testing process — instantiates Premise–Action Decoupling

Tests whether the conditions a convention requires before participation may count as acceptance are actually met here — and, when one fails, names the layer where the coupling breaks.

Every rule that turns engagement into assent — silence means consent, continuing means you agree, appearing means you concede jurisdiction — comes with fine print: a set of background conditions that must hold for the inference to be legitimate. Context-Condition Testing makes that fine print explicit and checks it against the actual situation. It does not ask what a message says or what act it performs; it asks whether the conventional and authority conditions that would license reading participation as acceptance are genuinely present here. Its defining move is to treat the ratification rule as a claim with preconditions and to test them one by one — and, when a precondition fails, to pinpoint which layer (the wording, the notice, the authority, the channel) the failure sits in, so the fix is aimed rather than diffuse.

Example

A small business notices its software subscription has quietly renewed at nearly double last year's rate. The vendor's position rests on a negative-option convention: you were notified of the change, you did not cancel, therefore you accepted. Context-Condition Testing puts that inference on trial by its own required conditions. The convention that non-response equals acceptance is only valid where a few things hold: clear advance notice, delivered through a channel the customer actually reads, with enough time and an easy path to decline.

Tested against the facts, the conditions come apart. Notice existed — but it was a line inside a mid-month usage email, not a dedicated renewal notice; the window to act was days, not weeks; and cancellation required a phone call during business hours. Two of the three preconditions fail. The finding is not "the vendor lied" but something more precise: the silence-equals-acceptance inference is not licensed, and it fails specifically at the notice layer. That localization is what tells the business exactly what to demand — a proper renewal notice and a clean opt-out window — rather than a vague grievance about being overcharged.

How it works

  • Name the ratification inference in play — state the exact rule the other side is relying on ("non-response within N days = acceptance of the new terms").
  • Surface its required conditions — from the governing convention, statute, or institutional custom, enumerate what must be true for that inference to hold: adequate notice, competent and entitled sender, a real chance to decline, absence of coercion.
  • Test each condition against the facts — mark each present, absent, or degraded, with what would have to change to satisfy it.
  • Localize the failure to a layer — a failed condition belongs to a specific layer (semantic wording, notice/disclosure, authority, response channel); the layer is the address for repair.

The distinguishing discipline is that it audits the inference's context, not the message. A premise can be perfectly parsed and its force correctly named and still fail here, because the convention that would bind it doesn't actually apply.

Tuning parameters

  • Condition strictness — how demanding a bar each precondition must clear. A strict bar protects the receiver but voids couplings that were arguably fair; a loose bar upholds more couplings but lets thin notice pass as real notice.
  • Authority scope — how broadly you credit the sender's entitlement to make the inference stick (a court, a landlord, a platform, a stranger). Getting this wrong either over-submits to weak authority or ignores real authority.
  • Layer resolution — how finely you localize the failure. Coarse localization ("notice was bad") is quick; fine localization ("notice existed but through the wrong channel") makes the repair precise but takes more digging.
  • Evidentiary threshold — how much proof a condition needs before you call it met versus degraded, trading investigative cost against confidence.

When it helps, and when it misleads

Its strength is that it defeats the most common power move in this archetype — "by participating you already agreed" — without denying that participation happened. It concedes the surface fact and attacks the inference, on the convention's own terms, and it turns a diffuse sense of unfairness into a named, addressable defect at a specific layer.

It misleads when the conditions are read too generously or too harshly. Over-strict testing voids couplings that were substantively fair, breeding a reflex that no participation ever binds; over-lax testing rubber-stamps a coupling because a token of each condition technically existed. It can also be run backwards — assembling a case that the conditions "were met" to defend a coupling already imposed, rather than testing them in good faith. And a clean pass here is not endorsement: conditions holding means the inference is licensed, not that the underlying premise is true. The discipline that guards it is to fix the required conditions from the governing convention before looking at the facts, so the bar is not bent to the desired verdict.[1]

How it implements the components

Context-Condition Testing realizes the conditions-and-context components — the layer that decides whether a ratifying inference is even entitled to run:

  • authority_and_convention_context — its subject matter: it reconstructs the convention, custom, or authority the coupling leans on and states what that backdrop actually requires.
  • ratification_inference_rule — it names the specific "participation ⇒ acceptance" rule and tests it as a defeasible inference with preconditions rather than a settled fact.
  • layer_specific_repair_plan — by localizing which precondition fails and at which layer, it produces the address a targeted repair aims at (fix the notice, not the whole contract).

It does not extract the literal premise (embedded_premise — Literal Content Parsing) or classify the act's force (engagement_assent_coupling, response_coupling_rule — Force Classification); and it produces the repair plan but does not execute or log it (consequence_record), which the response-path and record mechanisms own.

Notes

The test is diagnostic, not corrective. Finding that a coupling is unlicensed does not, by itself, undo a ratification already recorded against someone; it produces the layer-specific repair plan, but that repair still has to be enacted and its outcome logged — work that belongs to the response-path and record mechanisms, not here. Keeping the diagnosis separate from the remedy is what lets the same finding support anything from a quiet fix to a formal challenge.

References

[1] J. L. Austin's felicity conditions — the background conditions (appropriate authority, correct procedure, sincerity, uptake) under which a performative act actually comes off — are the direct ancestor of this test: a ratifying inference, like a christening or a verdict, only binds when its conditions are met. ↩