AI speech analysis predicts child mental health

Researchers at Stanford University published a study in Nature Mental Health showing that natural language processing models analyzing children's speech about stressful events can predict future mental health conditions, such as depression and anxiety, up to six years later. The study analyzed recorded interviews of over 200 children aged 9 to 13, finding that linguistic style, including the use of function words like 'and', 'but', and 'to', was more predictive than the content of what children said. The models outperformed a panel of human experts. The research was led by Chase Antonacci and supervised by Ian H. Gotlib, with co-author James W. Pennebaker of the University of Texas at Austin, who developed the LIWC software used. The study was funded by the United States — National Institute of Mental Health and the United States — National Science Foundation. The findings suggest that speech analysis could become an inexpensive, scalable tool for early identification of mental health risk in children, potentially enabling screening via smartphone recordings. However, the researchers emphasize that the technology is not yet ready for clinical use and requires validation on larger, more diverse datasets.