Does thinking harder make a group smarter?
More numeracy and reflection should make judgments more accurate. On topics tied to group identity, the opposite shows up: the strongest thinkers polarize the most. Kahan and colleagues showed in 2012, in a large study on climate risk, that polarization between political camps was greatest among people with the highest science literacy and numeracy. Kahan found the same pattern in 2013 for the ability to reflect cognitively. In the motivated numeracy experiment by Kahan, Peters, Dawson and Slovic from 2017, the same people read a data table correctly as long as it was labelled a skin cream study, and incorrectly once the exact same table was labelled gun control data and the correct result contradicted their own political side. The most numerate participants rationalised the most skilfully. For a single person, this behaviour even makes sense. Their opinion on the climate does not change the climate, but it can cost them standing in their own group. One caveat belongs here: a large, preregistered replication from 2021 confirmed that people answer in line with their own group on these topics. That the most numerate people distort the most, specifically, did not hold up cleanly in that replication.
Why that is even rational for the individual
For a single person, this behaviour even makes sense. Their opinion on the climate does not change the climate, but it can cost them standing in their own group. One caveat belongs here: a large, preregistered replication from 2021 confirmed that people answer in line with their own group on these topics. That the most numerate people distort the most, specifically, did not hold up cleanly in that replication.
What flips the effect
The effect flips once accuracy gets rewarded instead of belonging. In the forecasting tournaments run by Mellers and Tetlock, one hour of training against common thinking errors consistently improved forecast accuracy, measured by the Brier score, by 6 to 11 percent across four tournament years. Taking part in the tournament itself also lowered attitude polarisation. Meetings and classrooms need formats that do the same: predictions that get scored afterwards, disclosed assumptions and outside feedback.
On record
Wager Dossier 3: The Strongest Opponents, in Their Best Form, status 6 July 2026.
- Kahan and colleagues (2012, 2013): polarization on climate risk and on cognitive reflection was greatest among people with high numeracy and reflection ability.
- Kahan, Peters, Dawson and Slovic (2017): the same data table was read correctly or incorrectly depending on whether it was labelled as neutral or as politically contested.
- Chang, Chen, Mellers and Tetlock (2016): one hour of training consistently improved forecast accuracy in forecasting tournaments by 6 to 11 percent over four years.
- Persson and colleagues (2021), preregistered replication: the group-consistent answering pattern held up, the amplification through high numeracy did not hold up reliably.
This note keeps growing
2026-09-02: Planted from the wager dossier for World Modeling.