Disconfirming Probe Plan¶
Protocol — instantiates Abductive Explanation Selection
Specifies the observations or tests most likely to overturn the current best explanation.
The Disconfirming Probe Plan takes the current best explanation as given and asks the opposite of the usual question: not "what more would confirm this?" but "what single observation, if we could make it, would most likely prove this wrong?" Its defining move is designing for disconfirmation — selecting the probe with the highest power to overturn the leading explanation, pre-committing the result that would force a revision, and setting the action the explanation may license only once it has survived. It does not generate candidates or rank a field; it attacks the front-runner directly, on the principle that a story able to survive its own hardest test has earned more trust than one merely decorated with supporting facts.
Example¶
An investigative reporter's leading explanation for a city's sudden budget shortfall is that a named official diverted funds. The temptation is to gather more anecdotes that fit. The disconfirming probe plan does the reverse: it asks what record, if it exists, would make the diversion story false — an audited transfer log showing the money went to a legitimately contracted vendor. That single document has more overturn-power than a dozen corroborating quotes, so the reporter pulls it first. The plan pre-commits the revision: if the log shows lawful disbursement, the diversion explanation is retired. It also fixes the action boundary — publication of an accusation requires the probe to survive plus independent corroboration; abductive plausibility alone does not license naming someone. The log comes back ambiguous, the probe neither kills nor confirms, and the plan's pre-set threshold keeps the story out of print until a sharper test resolves it.
How it works¶
- Fix the target. Name the single explanation to be attacked; the plan does not compare a field.
- Ask the falsifying question. "What would we expect to see if this were false, or if the strongest rival were true?"
- Choose maximum overturn-power per cost. Rank candidate probes by how decisively a result would kill the explanation, not by how easy they are to run.
- Pre-register the trigger and the boundary. State the result that forces revision and the action the explanation may support only after it survives.
Tuning parameters¶
- Overturn-power versus cost — the sharpest test versus the affordable one. Sharper probes buy more information but may be expensive, slow, or intrusive.
- Single decisive probe versus a battery — one killing test versus several converging ones. A battery is robust to a flawed instrument but dilutes focus.
- Pre-registration strictness — how firmly the disconfirming result is committed in advance, guarding against re-interpreting a bad outcome as a pass.
- Action threshold — how much a survived probe unlocks, from "keep investigating" to "act irreversibly."
When it helps, and when it misleads¶
Its strength is that it is the direct antidote to the confirmation spiral and the self-sealing story: a probe designed to fail is worth more than ten that were bound to pass, because only the former can actually change your mind.[1] It converts a comfortable explanation into one that has stood a real risk of refutation.
Its failure mode is the pseudo-probe: a test dressed as rigorous that could not realistically have overturned anything, so passing it certifies nothing. The classic misuse is cherry-picking a survivable check precisely because it is survivable, then presenting survival as vindication. The guarding discipline is to judge every probe by the question "what result would kill the explanation?" — if the honest answer is "none," it is not a test, and the plan should discard it.
How it implements the components¶
best_so_far_explanation— taken as the fixed target the plan is built to attack.discriminating_evidence_plan— the core output: the observation or test chosen for maximum overturn-power.revision_trigger— the pre-registered result that forces the explanation to be reopened.action_confidence_boundary— the rule setting what a survived probe does and does not license.
It does not implement candidate_explanation_set generation (Anomaly-to-Hypothesis Workshop) or the explanatory_fit_criteria ranking of a field (Inference-to-Best-Explanation Matrix) — it consumes a chosen best explanation rather than producing or ranking one — nor the base_rate_context weighting (Differential Diagnosis Workup).
Related¶
- Instantiates: Abductive Explanation Selection — the archetype's falsification-first probe design.
- Consumes: Inference-to-Best-Explanation Matrix — supplies the current best explanation the plan is built to attack.
- Sibling mechanisms: Inference-to-Best-Explanation Matrix · Differential Diagnosis Workup · Explanatory Case Memo · Anomaly-to-Hypothesis Workshop · Abduction Log · Forensic Scenario Reconstruction · Model-Debugging Hypothesis Loop
Editorial Notes¶
Form Classification¶
Form family: Representation, Specification & Plan
Rationale: Disconfirming Probe Plan operates as a non-executable information artifact that externalizes static or prospective structure because it specifies the observations or tests most likely to overturn the current best explanation.
Independent corroboration: The frozen evidence defines Disconfirming Probe Plan as 'Specifies the observations or tests most likely to overturn the current best explanation', so its operative form is Representation, Specification & Plan.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Philosophy
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Universal
Rationale: Philosophy of science cohered severe testing: a probe corroborates an explanation only when it had a real chance to expose its failure.
Related originating lineages:
- Statistics & Experimental Design — Error-statistical design operationalizes severity through tests with known opportunities to detect departure.
Review resolution: Both current reviews place disconfirming_probe_plan primarily in philosophy; the reconciled classification retains only lineages that materially shaped the mechanism and keeps breadth of origin separate from reach.
Review outcome: Reconciled after independent review; high confidence.
References¶
[1] Mayo, D. G. Error and the Growth of Experimental Knowledge. University of Chicago Press (1996). Explains that severe probes are likely to expose a specified error and that easy-to-pass tests provide weak evidence. registry ↩