There's a lot of concern and outrage toward so-called "AI therapists".
(Which is 100% totally warranted!)
Software shouldn't be sold as something that it literally can never be -- a human sitting down with another human, where one of those humans is appropriately trained and licensed. Oh, and they are a fellow human being with true lived experience, empathy, etc.
We acknowledge that AI cannot possibly handle a back-and-forth conversation with a human and provide actual clinical (and human) insights, assessments, and judgement.
And yet, for some, a contradiction of sorts gets introduced...
Rather than directly participating in the conversation itself, if AI is handed a full text transcript of a real therapy session conversation, they're okay with the AI making clinical observations and assessments from that transcript, with the AI making the initial judgement call as to what to include in the progress note.
How can we almost universally reject the first scenario, but get into debates on the second scenario?
Don't the shortcomings of AI that prevent it from being an actual therapist also prevent it from doing a good job as a therapist processing what happens within a therapy session? (And based solely off of a text transcript, which has its own set of shortcomings!)
