AI-powered mental health tools that work across languages
Thematic Areas: Health Care and Wellness, Teaching and Learning
Mentors: Jeff Phillips, Vivek Srikumar, Zac Imel
AI tools have the potential to improve mental health on a global scale—especially if tools can be leveraged across many languages. Schou, who is Danish, will use text and video data from online therapy in his home country to study how AI models originally trained in English can be adapted to a less-common language without erasing nuance. “Danish is an ideal gateway for studying how to adapt AI models to underrepresented linguistic and cultural contexts: it is underrepresented in large models, yet supported by a growing national AI infrastructure,” Schou said, citing national investments such as Gefion, one of the most powerful supercomputers in the world. All data will remain in Denmark and be used only with informed consent. Schou will use AI-driven algorithms to responsibly support therapists in crafting therapeutic messages and detecting early signs of patient disengagement, without compromising clinical intent. The same models could also be used as learning tools in supervised training environments, supporting AI-powered instruction for mental health professionals. “By adapting these tools to low-resource languages, the project also promotes broader access and inclusion in therapist education,” Schou said.
The University of Utah, Schou said, offers an “ideal environment” for his proposed project because of its leading researchers in natural language processing, psychotherapy-AI integration, and data science.
