lsanderson | Testing the Thinkers

August 5, 2007
Re:Framing
Testing Testers, Finding Flaws
By DENISE CARUSO

SOME problems are particularly tough nuts to crack. From cancer to computer viruses, no matter how much time and money we spend, they seem to defy all attempts to solve them.

Two computer science researchers at Keele University in England say they believe that more progress can be made by shifting our focus from the problems themselves to the people who strive to solve them. The researchers, Gordon Rugg and Joanne Hyde of Keele’s Knowledge Modelling Group, have come up with a process they call Verifier that is designed to seek out mistakes in existing research on difficult problems.

By applying the scientific method to knowledge itself, Verifier has proved adept at exposing gaps in logic that can result from expert biases and mistakes, gaps that can invisibly skew their research results. More

Autism

SUE GERRARD, a visiting researcher at Keele who has a background in psychology and animal behavior, used Verifier to follow the development of the autism model from the original articles written by Leo Kanner and Hans Asperger, who first identified the disorder, to the formulation of the latest diagnostic criteria.

Autism is characterized by varying degrees of impairment in three sets of behavior: communication skills, social interactions and restricted or repetitive patterns.

“What struck me is that the people researching autism are mainly psychiatrists who aren’t familiar with neuropsychology,” which studies the mechanisms of brain function, Ms. Gerrard said.

“They’ve tried to classify autism as a kind of unitary condition, as though they are saying, ‘You have red-spot syndrome,’ ” she said. “That could be anything — measles, mumps, scarlet fever — but that’s exactly what’s happening with autism.”

As an example, Ms. Gerrard notes that the automatic repetition of words known as echolalia is common in autism. “But which of the three categories of behavior does it fit into?” she said. “You don’t know just by looking at a child whether it’s classic repetitive behavior, if it’s a problem with their hearing and they can’t respond appropriately, or if they just want to annoy the person they’re speaking to.

“Someone trying to diagnose would say the category doesn’t matter,” she said. But she thinks Verifier shows that it does matter. “Because all the behaviors have been classified as interaction problems, different causes are being overlooked,” she said.