To Better Understand Mental Illness, University Scientists Give Computer Schizophrenia

By alcalde in Blog on May 12, 2011 at 4:39 pm | No Comments

As computers can mimic “perfect” human intelligence, they also now can mimic some human imperfections. Researchers at The University of Texas have afflicted a computer with schizophrenia in hopes that it will reveal the underlying causes of the disease.

UT graduate student Uli Grasemann and his professor, Risto Miikulainen, used a computerized artificial neural network, known as DISCERN, to mimic the effects of excess dopamine on the brain.

DISCERN, which was created by Professor Miikulainen, can remember and recall simple stories, as well as process the grammar of natural languages.

The results of this experiment, which are published in Biological Psychiatry, are significant: the virtual schizophrenia reflects how schizophrenic patients are often unable to forget irrelevant details or filter information, therefore causing false connections or wild stories.

The researchers believe that dopamine may help the brain identify which information is and is not relevant. In order to simulate an excessive dopamine release, the pair increased the computer’s learning rates, which made it difficult for DISCERN to “forget” details.

“It’s an important mechanism to be able to ignore things,” says Grasemann. “What we found is that if you crank up the learning rate in DISCERN high enough, it produces language abnormalities that suggest schizophrenia.”

The resulting symptoms were, in fact, very similar to those exhibited by actual schizophrenic patients, according to Ralph Hoffman, Yale School of Medicine psychiatry professor who analyzed the results of the experiment.

DISCERN began to relate wild stories after establishing false connections between bits of information stored in the brain — even taking responsibility for a terrorist bombing. An additional symptom was derailment, which included use of dissociative sentences, jumbled responses, and constant flipping between the first- and third-person perspective.

Both symptoms provide evidence for the hyperlearning hypothesis, a theory that speculates schizophrenic symptoms to be caused by the inability to distinguish what is important and what isn’t.

While the results of the study are undoubtedly monumental, the researchers acknowledge that a computer model can only display hypotheses in a powerful way, not prove them. The involvement of dopamine may bring scientists closer to discovering which mechanisms of the brain are directly related to the mental disease, but more clinical research is needed to support the DISCERN findings.

For the full article published in Biological Psychiatry, read below.

A Computer With Schizophrenia