Learning from literature
Can training on carefully selected psychotherapy texts improve a model’s understanding of therapeutic ideas and processes?
How should we train AI to understand psychotherapy?
An independent research project exploring how different training methods shape an AI model’s understanding, judgment and responses to everyday personal concerns.
Explore the research questionWhat we want
to understand
Our central question is whether specialised training produces more useful responses than giving a general-purpose model access to relevant literature.
Learning through conversation
Can examples and expert feedback improve how a model listens, explores a concern and chooses when to respond?
Learning what people experience
Do people find one approach more helpful for their own concerns? We plan to examine their feedback alongside independent clinician assessments.
Interested in taking part?
We plan to invite adult volunteers to compare experimental models using real, everyday concerns, with clear information about the study and informed consent.
Honest feedback will matter, including when a response is unhelpful, uncomfortable or simply misses the point.
The study is being prepared
Participation is not open yet. Details of the study, eligibility, privacy and consent will be published here before recruitment begins.
Model comparisons are not open yet. You can read the draft study information and how participation will work.
Read the study information