falsification-minded calibration

Works

A philosophy of testing and the psychology of why it is hard.

• Popper's Conjectures and Refutations (1963) supplies severity: a test is worth running only if the theory could fail it, and a theory that forbids nothing has said nothing.
• Wason's selection task (1966–68) supplies the empirical finding: given a rule to test, people overwhelmingly turn the confirming card rather than the potentially falsifying one.
• Nickerson's 'Confirmation Bias' (1998) is the standard review — the tendency is robust, general, and largely invisible to those exhibiting it.
• Mayo's Statistical Inference as Severe Testing (2018) supplies the modern formalization, replacing naive falsification with degrees of severity a claim has actually withstood.

Central ideas

Four ideas, and the third is the one that makes the construct usable.

• Severity: evidence supports a claim to the extent the test would probably have found it false had it been false — passing an easy test is not support.
• Confirmation bias: the default search pattern is for instances that fit, so the disconfirming search must be deliberate and structured rather than left to curiosity.
• The disconfirming question: 'what would change my mind, specifically?' — asked before the evidence arrives, because asked afterward it is answered by whatever did not arrive.
• Unfalsifiable is not false: a claim that forbids nothing is outside this construct's reach rather than refuted by it, which is a limit the construct's users routinely forget.

Distinctive vocabulary

Five terms, guarded.

• Severe test: one the claim would probably have failed if false. Not a difficult test.
• Falsifiable: forbidding some observable outcome. Not false, and not scientific-by-definition.
• Confirmation bias: preferential search for and weighting of fitting evidence. Not wishful thinking.
• Barnum effect: acceptance of vague descriptions as personally accurate. The specific failure mode for character and type claims.
• Disconfirmation: evidence that would count against. Not disagreement.

Debates and disagreements

The construct inherits philosophy's arguments and psychology's.

• Naive falsification does not describe science: Kuhn and Lakatos showed anomalies are tolerated for long stretches and rightly so, which is why severity replaced simple refutation as the operative idea.
• Holism: because beliefs face evidence as a web, a failed prediction never identifies which belief to abandon — the Duhem-Quine problem, which bounds what any disconfirming search can conclude.
• Whether confirmation bias is a defect: some argue that within an argumentative division of labour, each party seeking confirming evidence is efficient, and the correction happens between people rather than inside one head.
• Whether the discipline is self-applicable: seeking disconfirmation of one's self-account uses the same faculties that produced the account, which the hermeneutics-of-self construct documented here says are constitutive rather than neutral.

Intellectual relationships

It divides the calibration territory with its sibling.

• Epistemic calibration, documented here, measures confidence against resolved outcomes; this construct disciplines belief where no outcome will resolve — the two are complements and the record cites them for different claim types.
• Fallibilism, also here, is the general disposition both specialize.
• The hermeneutics of self supplies the hardest case: self-interpretation installs the pattern it names, so the disconfirming search has to be aimed at one's own reading rather than only at the world.
• Moral psychology supplies the mechanism at scale — motivated reasoning is confirmation bias with a stake attached.

How it changes this reading

Three placements, all on unresolvable claims where scoring cannot reach.

• For self-inquiry as maintained practice, it supplies the operative discipline: ask what would show the account wrong, and go look, repeatedly, because disconfirming evidence does not present itself.
• For the type-framework review, it supplies the decisive question — what result would have counted against the description — and the Barnum effect as the reason recognition is not evidence.
• For iteration as ground state, it supplies severity's structure: surviving a test earns another round, not a conclusion.
• The guard: unfalsifiable is not false. Several things this record cares about — meanings, commitments, values — forbid no observation, and the construct's job there is to say so rather than to discredit them.

Useful comparisons

Against its neighbours.

• Against scored calibration: same family, different domain — outcomes against no outcomes, measurement against practice.
• Against scepticism: the construct seeks disconfirmation in order to hold beliefs better, not in order to withhold them.
• Against devil's advocacy: a rehearsed opposing case is not a severe test unless it could actually change the conclusion.

Where the ideas meet

What the sources agree on.

• The informative move is the disconfirming one, and it is not the default.
• Support scales with what a test could have caught, not with how much evidence accumulated.
• The bias is invisible from inside, which is why the discipline has to be structural.

Where they part

Where positions part.

• On strictness: refutation as decisive against severity as a matter of degree.
• On whether the bias needs fixing in individuals or is corrected socially.
• On self-application: whether the discipline can be turned inward at all, which the constitutive-interpretation objection puts in doubt.

Limits

What the construct cannot do.

• It cannot reach unfalsifiable claims, and most claims about value are among them.
• It cannot say which belief to revise when a test fails: holism leaves that open.
• It cannot verify its own application: someone can perform a disconfirming search that was designed not to find anything, and only the design reveals it.
• It has no natural stopping point: the demand for further tests can become a way of never concluding.

Criticisms

Standing objections, at strength.

• That naive falsificationism is a poor description of inquiry, and the construct's popular form is the naive version.
• That the demand for falsifiability has been used as a rhetorical weapon against claims outside its scope rather than as a criterion within it.
• That performing disconfirmation is easy and doing it is hard, with no external way to tell them apart.
• That the argumentative-division defence of confirmation bias is stronger than the construct's individualist framing admits.

Common misreadings

Four.

• Unfalsifiable as a verdict: the label places a claim outside the method's reach and does not refute it.
• Token disconfirmation: naming a hypothetical that would change one's mind, chosen because it will not occur.
• Applying it to values: asking what would disconfirm a commitment misunderstands what a commitment is.
• Using it against others only: the construct's whole difficulty is turning it on one's own account, and outward-only use is confirmation bias with better vocabulary.

What remains outside this idea

The boundary.

• That a surviving belief is true: severity earns provisional retention, nothing more.
• That someone searched honestly: the design of a disconfirming search is what reveals its seriousness, and designs are rarely published.
• That an unfalsifiable claim is empty: many of the most important things a person holds forbid no observation.