Secondment: Miguel Vazquez
|
Institution of origin |
Barcelona Supercomputing Center (BSC) |
|---|---|
|
Host institution |
Psychology Lab, Institute of Applied Biosciences (INAB), Centre for Research & Technology, Hellas (CERTH) |
|
Initial objective |
Bridge BSC’s technical AI expertise with CERTH-Psy’s domain leadership in mental health and illness beliefs, exploring AI-assisted support for institutional research and laying groundwork for mental health policy and ethical evaluation systems. |
Results achieved
Because the actual clinical and psychological data was not yet ready for BSC to access, Miquel Vazquez focused on designing a functional prototype to illustrate the type of results achievable with AI currently.
He developed a conversational AI simulation between a “Patient” agent and a “Doctor” agent. The patient was assigned a specific disease (like CLL or HIV) and a “ground truth” psychological profile based on the Common-Sense Model (which describes how patients think about their disease and treatment). The Doctor agent had to converse with the patient naturally and then infer the patient’s psychological profile based solely on the conversation.
The prototype was highly successful (the AI doctor accurately inferred the patient’s beliefs with a Pearson correlation of 0.78), though it also revealed interesting clinical biases, such as the AI tending to overestimate patient distress. Data access remains an ongoing effort, to be continued by other BSC colleagues during future visits to CERTH.
