I often discuss in therapy the problems we face in #FOSS w/ #LLM-backed #AI (no surprise there).
My therapist told me today that one of her colleagues who was early in focusing on LGBTQIA+ therapy is ending their 20 year practice. In their top 3 reasons? AI.
My therapist also noted that she appreciates that I'm now one of her few patients who doesn't come to her with “Well, I asked AI & it said…” slop.
LLMs may have value for medical uses, but warmed over Eliza does not a therapist make.
My therapist had not heard of Eliza. When I explained, she immediately pointed out the trick: mirroring what someone is saying is powerfully validating. It's a tool therapists use to make patients comfortable and feel heard, and it helps build rapport.
But again, it is not in itself therapy.
I suppose some might conclude we are better off with BigTech as therapists rather than Real Humans, & I suppose we aren't too far from USA health insurers refusing to cover human therapy.
We should resist.
@bkuhn Slightly outdated, but I often think about this article from a couple of years back, likening LLMs to fortune telling (which is also basically unlicensed therapy).
https://softwarecrisis.dev/letters/llmentalist/
The details have shifted a bit since they wrote it, but the core idea about asking the audience to carry the real load while faking depth certainly hasn't.
Bradley M. Kühn
Dr Andrew A. Adams #FBPE 🔶
ptvirgo
John Colagioia