The most likely good is in the ability to study the model itself to better understand mental disorders and therefore inform treatment, or in the ability to depl

The most likely good is in the ability to study the model itself to better understand mental disorders and therefore inform treatment, or in the ability to deploy something similar to monitor folks who are already clinical patients on an opt-in basis. I'd agree with you that the deployment of the model in almost any context re: observation of the general public is very unlikely to be beneficial.