Tuesday, April 18, 2006

Blinkering Juries

Malcolm Gladwell, the author of Blink (and also this blog here), gave a talk on campus yesterday, packing the hall to capacity. He went over Blink's main themes, focusing especially on the fragility of expert judgments and how adding more information to a decision-making process can paradoxically reduce its accuracy. He told the stories from the book about symphonies immediately starting to hire women when screens were placed between the auditioning committee and the candidates and ER doctors better diagnosing chest pain when only allowed to consider four simple factors. The upshot of the argument is that people's judgments, even finely tuned expert judgments, are washed out by hidden biases that the experts themselves are not even aware of.

Applying these insights and pointing to wildly divergent sentencing rates between defendants of different race, Gladwell argued that we should experiment with stopping juries in criminal trials from knowing certain facts about the defendant. At its crudest this would entail putting defendants behind a screen and at the most extreme a form of trial by transcript. Gladwell argued this isn't as crazy as it sounds, as social psychology research has shown that laypeople are flat out terrible at telling if someone is lying by looking at them, so maybe juries wouldn't even be losing any potentially relevant information.

Gladwell figured that we'd know if this worked by watching if the gap in conviction rates across races closed. I tried to ask a question about how we'd make sure we were becoming more accurate and not simply more random. After all, coin flips to determine guilt would wipe out race as a factor but hardly be an improvement in terms of justice. I didn't phrase it very well at the time, but afterwards some friends and I tried to brainstorm about how you'd run an experiment without having to convince a state legislature to go first.

You'd have to do something like videotape tons of trials and keep track of them, then add to your data set only those where the defendant was definitively proved guilty or innocent by evidence that emerged after the first trial. Show the tapes to a control group jury and see if they accurately convict the people that were later proved guilty and acquit the innocent, then show doctored tapes that obscure the defendants or mask their voices to experimental groups and see if they did any better. If would take a lot of time and a lot of videotape and a lot of people (one trial run requires 12 subjects!) but it could be done.

1 comment:

CM said...

That's a really interesting idea -- I'm taking Evidence now, and so many of the rules are geared toward not prejudicing the jury. But ultimately, everybody concedes that it's impossible. From my own experience on a jury, I can tell you that the jury does not use logic or anything close to it to find facts.