An article by Dani Cooper in ABC Science on 23 March 2016 raises some interesting questions about traditional criminal identification methods.
Our natural tendency is to think that if more witnesses agree on what they have observed, the more accurate it will be. In the case of identifying a suspected criminal the more witnesses that are able to identify a person the stronger the case against that person right?
Well NO actually. According to the research cited by Dani Cooper the more consistency in the witness agreement the more doubt should be raised.
The article identifies the following important points within the research.
Statistical modelling shows probability of a large group of people all agreeing is very small with previous research indicating an error rate of around 47%. This is almost half.
“But new Australian research suggests that unanimity of witnesses should trigger a warning that perhaps police have the wrong person.
In a paper published today in the Proceedings of the Royal Society A, University of Adelaide researcher Professor Derek Abbott and colleagues reveal that rather than being a sure thing, unanimity is actually unreliable.”
Dr Abbott, from the School of Electrical and Electronic Engineering, said if a large number of independent witnesses unanimously identified a suspect to a crime, it was assumed they were right.
However statistical modelling by his team showed the probability of a large number of people all agreeing in these circumstances was small, casting doubt on the veracity of unanimity.
“Paradoxically if many witnesses confirm a certain suspect your doubts should increase rather than decrease,” Dr Abbott said.
He contrasts the identification of an apple among a group of bananas where a high degree of accuracy would be expected. However research shows that a “five-second glimpse” of a suspected criminal fleeing the crime scene would has show the error rate around 47%.
If an error rate of 47% is normal then there must be some doubt cast when there is total agreement after three or four witnesses and increasing doubt as the number of agreeing witnesses increases.
Is this the same in other Areas?
Dr Abbott identifies this “paradox of unanimity” is applicable across many fields. The article cites the following circumstances:
In engineering and computing the paradox suggests where measurements are overwhelmingly consistent it may indicate a systemic failure.
In archaeological profiling, consistent results over the origin of an ancient artefact could also indicate a specific manufacturing process for that artefact.
Interestingly the article indicates that the paradox was “key to uncovering the recent Volkswagen scandal where the company was found to have rigged the carbon emission readings of diesel cars.”
Suspicion was aroused when the emission test results were so consistent.