The concept of “neurodiversity” was proposed in the late 1990s, suggesting that developmental disorders, such as autism spectrum disorder, appear as a result of normal fluctuations in neural activity and should be regarded as individuality. With the rapid increase in the number of children diagnosed with developmental disorders, a neurodivergent society is becoming desirable. Overseas companies are rapidly growing by hiring people with developmental disorders with focus on their personal characteristics while the employment of people with intellectual disabilities / mental disorders in Japan is only 6.2% (Ministry of Health, Labor and Welfare, 2015). One reason for this is that the difficulties encountered by people with developmental disorders and their uniqueness are not visible, both to them and those around them, leading to slow progress in rational consideration and the establishment of social environments.

In this project, we aim to evaluate developmental disorders as cognitive individuality more quantitatively by integrating a constructive AI-assisted approach and analytical approach targeting persons with developmental disorders. This will help visualize characteristics that have almost been hidden from both the individuals with developmental disorders and those around them. We will design a computational model using AI that mimics the neural circuitry of the brain and reproduce the brain state in which developmental disorders occur, in order to understand cognitive individuality as a system. Under the tojisha-kenkyu project, we will attempt to understand the mechanisms behind developmental disorders by objectively observing and analyzing the problems of individuals with developmental disorders.

In the AI research, we will use neural circuit models that replicate the principle of so-called “predictive coding” in our brains. Our brains can predict and act on unknown situations by combining the limited information we receive from our sensory organs with the experience and knowledge we acquired from past events. The predictive coding principle has been proposed as a promising hypothesis to make this possible. By constructing a model of the brain on a computer and varying its internal parameters, a range of conditions can be created to simulate organic or functional brain diseases. Accordingly, we can constitutively investigate the mechanism of developmental disorders by learning and generating different cognitive behaviors while varying the conditions.

On the other hand, in the tojisha-kenkyu project, we will analyze various cognitive behavioral problems experienced by people with developmental disorders in daily life from the perspective of an insider observer and extract common mechanisms behind each problem. By comparing the results with models obtained from our AI research, we will uncover factors and parameters that explain cognitive individualities. As a result, we will elucidate cognitive individuality by revealing the principle of individuality generation as a system and develop fundamental technologies for neurofeedback and biofeedback (a technique for estimating the state of the individual in real time based on measured neural activity and biological information and communicating it to the individual) aimed at the conscious control of individuality.

By revealing cognitive individuality and understanding its mechanisms, we can expect to provide new guidelines for support and rehabilitation in the educational field, for employment and training in the employment field, and for diagnoses and treatments in the medical field. It will allow the design of principle-based support methods for diversity issues among individuals that could not be addressed by conventional case-based support. In particular, the application of neurofeedback and biofeedback technologies to revealed cognitive individuality will eventually enable the conscious control of cognitive individuality. This will be a useful technology, not only for people with developmental disorders, but for those with regular development and will contribute immensely to the establishment of a neurodivergent society.

This project has integrated two approaches to explore cognitive individuality: AI research, which investigates individuality from a constructivist approach, and Tojisha-kenkyu, which analytically examines it from a first-person perspective.

Tojisha-kenkyu has involved a new hypothesis-testing type of research based on the predictive coding theory. We conducted a thematic analysis of narrative data obtained from Tojisha-kenkyu meetings and reviewed existing studies on autism spectrum disorder to formulate a hypothesis and develop new questionnaires for later studies. To expand the community and further refine the hypothesis, we established the Tojisha-kenkyu Episode Bank (https://token.episodebank.com) and the Introductory Lecture on Tojisha-kenkyu (https://www.utokyo-ext.co.jp/ids/top). The former enables users to provide and search their own episodes, whereas the latter enables them to learn how to conduct first-person studies. These initiatives have already been implemented in many companies and shown to improve psychological safety and inclusion in the workplace.

In AI research, we have developed neural network models that imitate the functions of the human brain and demonstrated that diverse cognitive individualities can be explained by modifications of specific parameters in the predictive coding processing. Our findings suggest that an imbalance in the precision of predictive and sensory signals can lead to learning difficulties or overlearning for a particular experience, resulting in unstructured or biased internal models. This result implies that hypersensitivity and hyposensitivity, which have bipolar characteristics, are inextricably linked due to modulations of a specific function in the brain. We have also designed and conducted drawing and composition tasks with children and adults to test the hypothesis obtained in the model study. Our results demonstrated that parameter fluctuations as described above cause the individuality appearing in these tasks. Our further experiments showed that variations in the predictive encoding processing can explain cognitive individuality in various cognitive behaviors such as language ability, speech, and tactile and visual perception.

The above findings from Tojisha-kenkyu and AI research have enabled us to develop prototype systems that manifest cognitive individuality. Users can visualize their cognitive individuality by using smartphone applications, virtual reality, and robots. We believe that these systems will contribute to the improvement of inclusion and the realization of a neurodiverse society, in the same way as Tojisha-kenkyu at companies.