Recognising unfamiliar faces a risky business

Identification: The public are surprisingly bad at recognising unfamiliar faces, something that has implications for security, according to new research.

Dr Rob Jenkins of the University of Glasgow cited a study in which there was a 30 per cent error rate where people were shown a picture and then asked to pick that person out from a group of 10 pictures. Another study showed a 50 per cent error rate when staff were asked to validate photo credit cards, where a picture is used to compare with a person.

Dr Jenkins told the science festival that the reliance we put on matching faces and photos was "fundamentally flawed" and raised questions about the common practice of visually matching a person with a passport photo, something we rely on to guarantee identity.

"Images of different people can be more similar than two images of the same person," he said. "Humans can't reliably match unfamiliar faces."

Dr Jenkins said our ability to make an accurate match improved greatly when the face was familiar, for example that of a family member or celebrity. He has used this to create a new type of image, which used many different pictures of a person to create a single "averaged" face.

Dr Jenkins has run tests to assess how well people performed at matching faces when using averaged face images. One average created of Paul McCartney using pictures from the 1960s through today produced a readily identifiable image, he said.

© 2007 The Irish Times

This article appears in the print edition of the Irish Times