Cool! Various other people have also had good experiences with using other LLMs for therapy or coaching, e.g.:
- @KatWoods found GPT to be an excellent therapist
- A couple of people told me they had really good experiences with this character.ai IFS bot
- Several folks have said that Hey Pi is excellent for emotional issues
- I found this character.ai Life Coach pretty good
- I made this Ideal Mother bot to do something called Ideal Parent Figure therapy; character.ai had limited space for customizing it so it's a bit one-note, but I had some good experiences with it.
Huh, that's weird, it works when I'm logged in to the site with my own account but not when I'm logged out, even though it has been set as public. Also doesn't seem to show up in the site search. Did the staff shadowban it because they don't want anyone getting an ideal childhood? :D I'll have to try to figure out what's going on.
2CharacterAI bots don't show as public until some condition is fulfilled, which I don't remember right now.
ChatGPT-3.5 (when prompted the perfect way) outperforms CharacterAI, I assume GPT-4 with the right prompt would as well, you might check those options out. (I haven't tried a therapist specifically, though.)
Apparently there's a filter that shadowbans bots based on some secret arbitrary criteria. Well that's annoying. I guess I'll try making it into a GPT prompt instead.
1The link works for me, even not logged in to character.ai. When I search character.ai for "mother", it isn't listed, but almost all of what does show up is pretty disturbing. The great majority of the characters are very negative types of "mother".
The link works for me, even not logged in
Well that's bizarre, for me it continues to work when I'm logged in but not when I'm logged out.
Thanks for those links. I tried this IFS bot and it's really amazing.
Btw, is there a risk that after undergoing Ideal Mother therapy I will realise how less than ideal a mother I am to my own children and this will ultimately increase my anxiety?
That is a thing that I have seen happen! The normal reply is to remember that ideal parent figures are intentionally unrealistic and portray a standard that isn't even meant to be humanly attainable. (Though also I think the version in the chatbot comes off as a little too over-the-top loving, so it doesn't necessarily quite hit the intended ideal either.) But if you feel like you might react to it this way, it might be a good idea to try working through the anxiety in a less directly triggering way and avoid ideal parent chatbots for now.
It doesn't seem sufficient anymore to have a VPN in order to get an access to Claude. You also need a UK/US -based phone number. If anyone knows how to get around this, please let me know!
update on my current views: a lot of the useful meaning in Coherence Therapy sessions comes from intonation/paralanguage. Unfortunately, language models unfortunately can’t currently perceive this. I don't think current AI systems can do Coherence Therapy nearly as well as a counselor, though I imagine they can still help a lot for issues that can be totally verbalized (which I think isn't that common for people just starting out with therapy).
Coherence Therapy is, IMO, probably part of the "one true therapy". I think the crispest explanation of how to update one's own maladaptive beliefs/expectations/parts/traumas, especially when beliefs get stuck and the usual methods don't work.
Kaj Sotala provides an evidenced-based summary here Book summary: Unlocking the Emotional Brain which I adore.
However people I've recommended this to have told me that it's hard to implement.
It might be possible to simulate Coherence Therapy with LLMs, though— especially Claude 2.0 with its 100k context window.
Here's a demo I captured:
Ultimately I enjoyed my session with Counselor Claude.
To try it yourself:
Let me know in the comments if you come up with improvements for this, or areas where you feel Counselor Claude is lacking.
I am writing a much longer series about how to solve all social drama in your own life. Subscribe to my profile for when I release that (or DM me to review the draft).