Researchers used an artificial intelligence language model to identify disturbances in speech patterns in people with schizophrenia. They showed how you can apply artificial intelligence in psychiatry.
Neuroscientists have shown that you can use an AI language model to measure how coherently people with schizophrenia speak. This gives doctors a new way to determine the severity of a condition.
Schizophrenia is a mental disorder that affects approximately two to four out of every thousand Dutch people. People with this condition experience recurrent psychosis. These are periods when thinking and feeling are seriously disturbed, for example, hallucinations and delusions.
‘We’re taking the wrong turn on organ transplants’
Developments in the field of animal organ transplantation into humans are progressing rapidly. However, we would be better off investing in alternatives,…
Other symptoms can also make life with schizophrenia very difficult. For example, many people with schizophrenia experience disorganized thinking and speaking. This may be due to disturbances in the cognitive map. This is a mental representation that shows the relationships between different concepts. For example, if you think about pen, you will soon think about paper. This is because these concepts are close to each other on your cognitive map.
In order to properly treat schizophrenia, it is important for psychiatrists to monitor the patient’s symptoms. In the future, artificial intelligence will be able to help with this. It can make monitoring symptoms more efficient.
Psychiatrist and neuroscientist Matthew Nour His colleagues at University College London did this using an artificial intelligence language model Research into the cognitive map of people with schizophrenia.
The researchers gave people with and without schizophrenia a simple task: name as many words as possible within the category “animals” within five minutes. They expected that participants with a strong cognitive map would more often name successive words that were more or less related in meaning.
To verify this, the researchers used a Natural language processingA fashion model. Such a model uses artificial intelligence to look for patterns in Internet texts and can thus predict which words often occur together. For example, ChatGPT works this way. For example, the model determined that the words “cat” and “mouse” were more closely related than the words “mouse” and “swan.”
The study showed that compared to the control group, participants with schizophrenia more often mentioned consecutive words that did not match well in meaning. This corresponds to the severity of some of the symptoms of schizophrenia. The more severe the symptoms, the less the connection between the words.
Artificial intelligence in the clinic
Researchers show that an artificial intelligence language model could help monitor symptoms of schizophrenia. They also demonstrated that this approach is nerve-based (see box: Connecting to the Brain). “Until recently, automated language analysis was beyond the reach of doctors and scientists. “With the emergence of AI language models such as ChatGPT, this situation is changing,” Nour said in a press release.
Also a psychiatrist and researcher Iris Sommer from the University Medical Center Groningen sees a future in the use of artificial intelligence in psychiatry. “I think language is really something we will focus on a lot in the near future. Not only to understand schizophrenia better, but also to be able to monitor the patient remotely, for example. So we can prevent the development of psychosis at an early stage,” she says. “With AI speech analysis, it doesn’t have to be about complaints. You can talk about any topic to monitor the patient.
More research will be needed to determine whether the use of this AI language model is reliable enough in psychiatry. If that turns out to be the case, Nour believes they could appear in the clinic within the next decade.
Connection with the brain
Nour and his colleagues did not want to study the power of the cognitive map only at the behavioral level. They also wanted to know if they could measure disturbances in brain activity. They looked to the hippocampus for this. It is located approximately in the middle of your brain at the level of your ears. This area acts as a librarian for your brain: it stores concepts and memories, and retrieves them when you need them.
Sometimes the hippocampus replays the memory, but at a faster pace. This phenomenon is called Re hippocampus. You then unconsciously recall an event so that it can be properly processed in your mental archive. Scientists believe that this process plays an important role in forming the cognitive map.
Nour and his colleagues developed Re hippocampus examined in its participants. To do this, participants first had to perform a learning task, after which their brain activity was measured. The study found that a certain type of brain activity associated with re-experiencing an unconscious learning task was weaker in participants who had mentioned less coherent words in the previous task. This suggests that the cognitive map may be shaped differently in these individuals.
“Coffee buff. Twitter fanatic. Tv practitioner. Social media advocate. Pop culture ninja.”