Josh Barro, writing for the National Review:
Research has shown that the confidence of an eyewitness is the principal determinant of whether or not jurors will believe that an eyewitness made an accurate identification (Lindsay, Wells, and, Rumpel)…
Under very favorable conditions (e.g., a good view, a fair lineup), the correlation between confidence and accuracy is probably somewhere around .40. For purposes of comparison, consider that the correlation between a person’s height and a person’s gender is .71. This means that confidence is a poorer predictor of accuracy than height is a predictor of gender.
There’s been a lot of literature written on how the human brain isn’t as good as we think it is at recollecting things. And when it comes to deciding how accurate a memory is, jurors (whether they know it or not) apparently use a witness’s confidence as a primary indicator of the witness’s accuracy; this is a bad idea. That much I knew, but I didn’t know that it’s twice as lousy an idea as using someone’s height to predict their gender. There are tons of great pieces of research on this stuff.
This particular article is about how the human brain isn’t even good at remembering how confident it was when it remembered something.
It’s a good thing we didn’t build an entire legal system on the premise of eyewitness testimony. Things might get a little ugly in that case.