Do debiasing programmes work?

Teaching people about cognitive biases rarely changes what they do. One family of interventions has produced large effects that lasted months: a single interactive exercise in which people practise catching named biases in their own reasoning and are corrected. The difference between the two is the useful part of this literature.

Why awareness fails#

The standard session presents a list of biases with memorable examples, and people leave able to name them. Recognition in a described case does not produce recognition in your own reasoning, where the bias does not arrive labelled and where the reasoning feels like ordinary thinking. Fischhoff's 1982 taxonomy remains the standard account: he sorted remedies by whether the fault lies with the task, with the judge, or in the mismatch between them, and awareness training addresses only the second and only weakly.

Larrick's review of the field identified what does transfer. Most of it is not training. Incentives change behaviour where the incentive tracks accuracy. Aggregating independent judgements removes individual error without anybody becoming less biased. Considering the opposite, deliberately arguing the other case, is the one cognitive instruction with reasonable support, applied at the moment of decision rather than learned in advance.

The exception with numbers#

Morewedge and colleagues reported two experiments in 2015, with roughly 278 and 269 participants, comparing a single interactive training game against an instructional video on the same content. The game produced large reductions in the biases it targeted, still present at eight weeks in the first experiment and twelve weeks in the second, and outperformed the video in the first.

Three features separate it from a briefing. Participants made judgements themselves rather than reading about judgements. They received immediate personalised feedback on their own errors. And the biases were named and targeted in advance rather than covered as a survey. That is practice with feedback, which is what the rest of the learning literature would predict works, and the outcome measures were bias tests rather than decisions with money or consequences attached.

Red teams, devil's advocates and the rest#

The structured analytic techniques developed in the intelligence community, several of which arrived in business as red teams, pre-mortems and competing hypotheses, were reviewed by Chang, Berdini, Mandel and Tetlock in 2018. They found the psychological rationale weak and evidence of improved accuracy essentially absent. Absence of evidence here means absence of testing rather than a demonstrated failure, which is the fairest reading and still an uncomfortable position for methods in their fourth decade of institutional use.

What this means for an organisation#

Bias awareness training is the most commonly purchased form of judgement development and the least supported, which makes it the first thing to stop doing if the budget is to go somewhere with evidence behind it. The candidates with measurement are the structured debrief, calibration training, and practice with immediate feedback on the person's own errors. The structural options, aggregating independent judgements and reducing inconsistency, do not require anybody to change how they think, which is part of why they work. That is what a noise audit is for.

What this does not establish#

That debiasing is impossible, or that the Morewedge result reaches real decisions. The outcome measures in that work were tests of the trained biases, administered in the same format as the training, which is a friendly setting. No study has shown a debiasing intervention improving consequential professional decisions, and this page makes no claim that one exists.

Key sources

• Morewedge, C. K. et al. (2015). Debiasing decisions: improved decision making with a single training intervention.
• Chang, W. et al. (2018). Restructuring structured analytic techniques in intelligence. Intelligence and National Security, 33(3).

Related SuperSkills research#

• How is judgement trained
• What is a noise audit
• What is a premortem