←back to thread

'It cannot provide nuance': UK experts warn AI therapy chatbots are not safe

(www.theguardian.com)
165 points distalx | 8 comments | 10 May 25 15:35 UTC | HN request time: 0.005s | source | bottom
Show context
hy555 ◴[10 May 25 17:02 UTC] No.43947151[source]▶
Throwaway account. My ex partner was involved in a study which said these things were not ok. They were paid not to publish by an undisclosed party. That's how bad it has got.

Edit: the study compared therapist outcomes to AI outcomes to placebo outcomes. Therapists in this field performed slightly better than placebo, which is pretty terrible. The AI outcomes performed much worse than placebo which is very terrible.

replies(9): >>43947337 #>>43947376 #>>43947393 #>>43948478 #>>43948522 #>>43948984 #>>43949480 #>>43949609 #>>43950587 #
1. sorenjan ◴[10 May 25 17:32 UTC] No.43947376[source]▶
What did they use for placebo? Talking to somebody without education, or not talking to anybody at all?
replies(1): >>43947498 #
2. hy555 ◴[10 May 25 17:48 UTC] No.43947498[source]▶
Not talking to anyone at all.
replies(2): >>43947566 #>>43947681 #
3. zargon ◴[10 May 25 17:57 UTC] No.43947566[source]▶
What did they do then? If they didn't do anything, how can it be considered a placebo?
replies(2): >>43947601 #>>43947975 #
4. risyachka ◴[10 May 25 18:01 UTC] No.43947601{3}[source]▶
Does it matter? The point is AI made it worse.
5. trod1234 ◴[10 May 25 18:12 UTC] No.43947681[source]▶
That seems like a very poor control group.
replies(1): >>43947781 #
6. hy555 ◴[10 May 25 18:27 UTC] No.43947781{3}[source]▶
That is one of my concerns.
7. phren0logy ◴[10 May 25 18:56 UTC] No.43947975{3}[source]▶
It's called a "waitlist" control group, and it's not intended to represent placebo. Or at least, it shouldn't be billed that way. It's not an ideal study design, but it's common enough that you could use it to compare one therapy to another based on their results vs a waitlist control. Placebo control for psychotherapy is tricky and more expensive, and can be hard to get the funding to do it properly.
replies(1): >>43948461 #
8. ◴[10 May 25 20:00 UTC] No.43948461{4}[source]▶