I found this discussion of overwhelming evidence accidentally while searching for material on another topic and thought it looked interesting.
"The paper shows, for instance, that a police line-up where victims all agree on the identity of an attacker is actually less likely to be correct than one where there is some disagreement between witnesses. Similarly, a set of DNA samples that are all in perfect agreement could be evidence of an underlying systemic error."
The original paper is available. It's not about criminal trials specifically but includes the example of police identification line-ups. Because evidence suggests high rates of identification error in normal practice, unanimity amongst witnesses can, counterintuitively, imply a problem with the evidence. A correctly conducted and unbiased procedure would not produce that result.
The paper uses Bayesian analysis and depends on certain assumptions to be valid. It reminded me of other cases where Bayesian analysis turns up counterintuitive predictions (e.g. a Bayesian analysis where it was shown that a convicted person having a previous conviction can make them more likely to have been wrongly convicted, as prior convictions can cause police and prosecutor bias).
"The paper shows, for instance, that a police line-up where victims all agree on the identity of an attacker is actually less likely to be correct than one where there is some disagreement between witnesses. Similarly, a set of DNA samples that are all in perfect agreement could be evidence of an underlying systemic error."
The original paper is available. It's not about criminal trials specifically but includes the example of police identification line-ups. Because evidence suggests high rates of identification error in normal practice, unanimity amongst witnesses can, counterintuitively, imply a problem with the evidence. A correctly conducted and unbiased procedure would not produce that result.
The paper uses Bayesian analysis and depends on certain assumptions to be valid. It reminded me of other cases where Bayesian analysis turns up counterintuitive predictions (e.g. a Bayesian analysis where it was shown that a convicted person having a previous conviction can make them more likely to have been wrongly convicted, as prior convictions can cause police and prosecutor bias).
via International Skeptics Forum https://ift.tt/2LKPmfe
Aucun commentaire:
Enregistrer un commentaire