We read facts to suit ourselves

Two fabricated studies arrive together. One supports the death penalty. One opposes it. Supporters and opponents examine both.

The supporters spot serious weaknesses in the anti-death-penalty research. The opponents find serious weaknesses in the pro-death-penalty research. Each side can explain its criticism intelligently. That is what makes the result useful and slightly horrible.

The asymmetry lay in how that criticism was applied. Friendly evidence received a quick welcome. Hostile evidence got the full forensic service.

This is easy to recognise in somebody else. A five-star review of the appliance we want seems detailed and authentic. The one-star review comes from a fool who ignored the instructions. Our favourite player attempts an ambitious pass; the rival carelessly loses possession. Intelligence may worsen the problem because clever people can produce first-class reasons for using two standards.

Debates about method therefore become strangely energetic after the conclusion appears. Sample size, funding, definitions and missing data all deserve scrutiny. The suspicious part is discovering those standards only when the result annoys us, then forgetting them when the next paper agrees.

Before another evidence review, participants score the sample, method and source without seeing the conclusion. The pro-death-penalty study gets six out of ten. Then the conclusion is revealed.

One supporter notices a sampling issue that was visible from the start. An opponent suddenly thinks the sample may be adequate after all. Both objections go into the notes beside the original scores.

The chair calls for a vote. The first hand goes up.