Before you read that article, what was your opinion of fingerprint evidence? [pollid:74]
My first exposure to the idea that fingerprint evidence might not be all that good was in L. Neil Smith’s The Probability Broach, in which the viewpoint character, a policeman, wonders whether all fingerprints really are unique, and also wonders whether the government might disappear people who had identical fingerprints.
(Surprisingly to me, the Salon article mentions that identical snowflakes have been found.)
My second exposure was an article in Lingua Franca, which brought up the more plausible issue that fingerprints from crime scenes are likely to be partial and/or fuzzy.
The article doesn’t actually contain any data saying that fingerprints are reliable. If I had to guess, I’d say that a (non-partial) fingerprint match had an odds ratio of around 10^7, a hefty 22 bits of info. Is there any data to contradict that? Or is this just the “but there’s still a chance, right?” fallacy?
Movement toward taking a statistical approach to the quality of evidence from fingerprint similarity
Before you read that article, what was your opinion of fingerprint evidence? [pollid:74]
My first exposure to the idea that fingerprint evidence might not be all that good was in L. Neil Smith’s The Probability Broach, in which the viewpoint character, a policeman, wonders whether all fingerprints really are unique, and also wonders whether the government might disappear people who had identical fingerprints.
(Surprisingly to me, the Salon article mentions that identical snowflakes have been found.)
My second exposure was an article in Lingua Franca, which brought up the more plausible issue that fingerprints from crime scenes are likely to be partial and/or fuzzy.
Write-in: I believed it was among the more reliable forms of forensic evidence, but didn’t believe the bombastic claims of absolute certainty.
The article doesn’t actually contain any data saying that fingerprints are reliable. If I had to guess, I’d say that a (non-partial) fingerprint match had an odds ratio of around 10^7, a hefty 22 bits of info. Is there any data to contradict that? Or is this just the “but there’s still a chance, right?” fallacy?