Designing a method to counter a bias is not evidence it counters the bias

A technique built to correct a reasoning error is not the same as a technique shown to correct it. The design encodes a theory of why people go wrong. Whether the procedure actually repairs the error is a separate, empirical question, and institutions routinely treat the first as if it settled the second.

Heuer's Analysis of Competing Hypotheses is the case. It was grounded in cognitive psychology and the logic of falsification: lay out every hypothesis at once, weigh which evidence actually discriminates between them, and rank by what has the least evidence against it rather than the most for it. The reasoning is elegant and it reads like the scientific method imported into a flowchart. But Heuer never claimed it yields correct answers. He wrote that ACH "guarantees an appropriate process of analysis," and that improved odds were something to be "assumed." After the 2002 Iraq estimate, the community mandated structured techniques as tradecraft anyway.

When researchers finally ran ACH through controlled experiments decades later, the result was little to no benefit, and in some trials it reduced the coherence of judgments and raised error. The procedure's resemblance to good science had stood in for evidence the whole time.

A bias-correction never measured against the bias is an assumption with a flowchart attached.

The pattern holds anywhere a field adopts a debiasing ritual because it sounds rigorous. Plausibility is cheap; the test is what gets skipped. Compare [[forecasting habits that improve prediction are cheap to install but absent from analytics training]] and [[when confident wrong analysis becomes cheap catching it becomes the scarce skill]].

Source claim: A method designed to counter a cognitive bias is not thereby shown to counter it; the design encodes a theory of the error, and whether the procedure fixes it is a separate empirical question institutions tend to skip.