In this article series, I look at how people are using LLMs to try to help themselves, what might go right or wrong with that, and likely pros and cons versus a human therapist. I also talk about some systemic concerns about privacy and model training. I provide some preliminary suggestions for people who want to use LLMs for self-help on how to do that so that you don’t give your data away and otherwise keep yourself as safe as possible.
Let me know if I’ve missed an important point, gotten anything significantly wrong, or if you have other thoughts or comments. I am also, of course, available for help with this in the form of ongoing therapy if that’s something you want
-
AI-assisted therapy (part 1): why I’m writing about it
I’m trying to feed some informed and cautious advice into that process so that anyone, working with me or not, can get the most out of what they’re doing. This page will talk through some steps to do that.
-
AI-assisted therapy (part 2): how to do it
Steps to follow to use both common cloud-hosted and locally hosted models to start ai-assisted therapy.
-
AI-assisted therapy (part 3): what are LLMs probably good at?
How does human-based therapy work? And what might AI-assisted therapy do well or poorly?
-
AI-assisted therapy (part 4): shortcomings and pitfalls
If you use LLMs to avoid what you fear, you may be trading short-term comfort against long term progress.