Chinese scientists are developing built-in artificial intelligence (IA) robots to help treat depression, combining brain signal analysis with large-scale language models, has today revealed an investigator.
© iStock
28/08/2026
by
Lusa
“ ANDwe are working in built-in AI that can express emotions to interact more naturally with people. We developed intervention methods for depression based on large-scale language models,” said Zhiguo Zhang, a professor specialized in Brain-Computer Interfaces, Affective Computing and Medical AI at the University of Shenzhen Hetao, during a conference held at the University of Macau.
Zhang, who leads the Center for Language, Intelligence and Machines of that institution, pointed out that he had a social experience of intervention by AI and concluded that “at least in China privacy concerns were not a significant obstacle.”.
“ People have a more positive perspective on AI,” he added.
As indicated during a symposium on emotional AI, its investigation focuses on brain-computer interfaces and affective computing, with practical applications in clinical problems, such as early detection and intervention in depression.
The scientist pointed out that the construction of AI systems with emotional awareness for the treatment of depression involves the use of an electroencephalogram (EEG), a medical examination that records the electrical activity of the brain, as one of the possible inputs to assess the emotional state of the patient.
Zhang pointed out that an EEG has “less precision” than collecting information about patients through audio and video. Since then, it has expanded its approach to include voice, facial expressions, wearable devices (smartwatches and smartphones) and functional magnetic resonance (fMRI).
The investigator then developed a depression tracking system that combines EEG, functional magnetic resonance imaging (fMRI) and wearable device data, having obtained “very good results”.
To respond to privacy concerns, the team uses a technique called “knowledge distillation”, which allows EEG models to learn from audio and video data without directly collecting sensitive confidential information.
Zhang has revealed that the system is already being used in a local hospital in Shenzhen, Guangdong province (south) for mental health risk screening.
“ We now have a profile of the mental health of residents of a community in Shenzhen,” Zhang said.

