February 12, 2007

The School Board was just Practice

Overcoming Bias: More Lying mentions the interesting finding that judges on the Swedish Migration Board are no better than students at recognising the signs of lying.

We have been discussing lying on Overcoming Bias for some time now. People are in general bad at detecting lying, even when they have that as a job - and such "experts" are often utterly convinced that they know what they are doing. It looks like it is possible to get better results in detecting deception and even to train away some of the overconfidence, but that requires "experts" to admit there is a problem.

Overall, the legal system seems to be suffering from serious biases. Lawyers attempt to manipulate jury bias (with mixed effects, since they are bad at predicting how the jurors would vote). Worse, police have a strong confirmation bias making them fix on suspects even when the evidence begins to point in another direction. Once the pressure on a subject builds, False confessions become probable - and of course convinces courts far more than they should.

I was reminded of this today by a story in The Economist about how the Japanese legal system manages to extract confessions from 95% of all people arrested, followed by a 99.9% conviction rate in the courts. Given the number of cases where innocence has been proven in the far more lenient US system, this likely sums up to a horrendously large number of innocents in prison, even if the arrests were better targeted.

It is interesting to note that at least in some experiments groups can be more accurate than individuals in judging whether someone is truthful or lying. By discussing different cues they could make a better estimate than individuals, who often fix on a single cue like eye contact. In particular, groups with more hung judgements were more accurate - they could recognize situations where they could not make a good judgement. As individuals we are less likely to do so. On the other hand, individuals tended to be biased to believe people tell the truth, while groups lacked that bias.

In the case of the Swedish immigration boards it is not hard to imagine many of these biases acting (except that they are unlikely to force people to sign false confessions). However, as Stuart pointed out, there is one group who is better than most at detecting deception: convicted criminals. Maybe we should make serving on the migration board a punishment for repeat offenders?

Posted by Anders3 at February 12, 2007 01:06 PM