The language used by people with schizophrenia and people on the autism spectrum is somewhat different from that of healthy people. Scientists using the example of the Polish language show that artificial intelligence can help psychiatrists and psychologists in recognizing the symptoms of schizophrenia, and sometimes - maybe autism.
The researchers focused on statements from standard clinical interviews (diagnostic tests) of adults with schizophrenia and children with autism spectrum disorders. In the case of schizophrenia, the algorithm was highly effective - almost 90%. (so he was wrong once in 10 ratings). And in the case of autism - it was weaker - its effectiveness was about 65%. (So he accurately recognized two out of three people).
The results of the research on this subject appeared in Cognitive Computation ( https://doi.org/10.1007/s12559-021-09834-9 ). The authors are the spouses Dr Aleksander Wawer from the Institute of Computer Science of the Polish Academy of Sciences and Dr. Justyna Sarzyńska-Wawer from the Institute of Psychology of the Polish Academy of Sciences, as well as Dr. Izabela Chojnicka from the Faculty of Psychology of the University of Warsaw and Dr. Łukasz Okruszek from IP PAN.
"Our tool may be of interest to psychiatrists or clinical psychologists. It can help them in the diagnosis or in the analysis of the progress of therapy - especially in people with schizophrenia" - comments Dr. Justyna Sarzyńska-Wawer. And Dr. Aleksander Wawer adds that such research has already been carried out for other languages - even English, but it has not yet been tested in Polish.
The researchers explain that the algorithm was trained on texts collected by a psychiatrist during research on social cognition of people with schizophrenia. These people include they verbally answered general questions (about themselves and their loved ones) as well as abstract questions (e.g. why do people get sick and why do they believe in God). These statements were later written down. The control data was collected in the same way - statements of healthy people from similar demographic groups.
In the case of autism research, the algorithm learned from texts written from partially structured standardized studies involving children, adolescents and adults.
"On these data, we tried to learn artificial intelligence models to recognize which statements are statements of people with a diagnosis" - says Dr. Aleksander Wawer.
And Dr. Sarzyńska-Wawer explains: "In the case of the schizophrenia study, these were not people in the active phase of psychosis. And the differences between healthy and diagnosed people were difficult to detect + with the naked eye +. An experienced psychiatrist, whom we asked to evaluate on the basis of the same texts, whether it was schizophrenia or not, it fared worse than the algorithm. "
It has long been known that the language of people with schizophrenia or autism spectrum disorders is different from that of healthy people. For many years, scientists have distinguished features of language in autism and schizophrenia. Earlier studies have shown, for example, that people on the autism spectrum disorder use fewer emotionally marked words. There is also a difference in the degree of abstraction of statements. "On the other hand, it is known from other studies that in people with schizophrenia there are more affective statements, but mainly negative ones" - says the psychologist.
The diagnostic process of both of these units, however, is based on subjective behavioral indicators. "And we show that this process can be automated, objectified and improved" - says Dr. Wawer. And he adds that in psychiatry there are not as many objective tools as in other areas of science, such as blood tests or brain scans, that would say with certainty that someone has a given disease or not. "So there is a large role of a psychiatrist and a psychologist, their experience. And yet there are many disorders that may be similar to each other. An additional diagnostic tool may be useful here" - the interviewees say.
Scientists hope that someday such a tool could be used remotely for screening. Therefore, the algorithm would help psychiatrists and clinical psychologists - on the basis of the submitted written statements - to choose people who should apply for further specialist tests. Thanks to this, it would be possible to diagnose diseases earlier and start therapy.
"From my perspective, the use of the few-shot method in this research was a novelty," says Dr. Aleksander Wawer. "Neural networks usually require huge bases for training. For example, to learn an algorithm for recognizing road signs, millions of photos are needed. And we do not have millions. cases of written statements of people diagnosed with schizophrenia. We only had 50 such texts. The challenge was to use such methods of training artificial intelligence that require a small database. It was unique from the IT point of view "- he says.
PAP asked scientists if they were not afraid that this tool could be used to analyze the texts of people who would not wish to be diagnosed (even statements by public figures). Dr. Sarzyńska-Wawer emphasizes that the tool will not replace a specialist in the diagnosis. "The algorithm detects only symptoms that may be a sign of schizophrenia. But these symptoms are not synonymous with diagnosis" - he points out. And Dr. Wawer adds: "For now, we've been taking into account the data from clinical interviews. It's still a long way to go to take any text into account."
Scientists announce that when it comes to supporting the diagnosis of schizophrenia, their tool is so effective that they would soon like to make it available to psychiatrists. (PAP)
Author: Ludwika Tomala