Thank you for the insight, Davina. I started using LLMs to talk about mental health and my trauma a few years ago, and I noticed this pattern too. Too often, LLMs echoed back what I wanted to hear rather than offered genuine psychological insight. It took seeing a real therapist to diagnose my CPTSD and treat it with exposure therapy. AI can simulate empathy based on pattern recognition, but it cannot comprehend the depth of human trauma.
We must raise awareness that LLMs' responses must be taken with a grain of salt.
This is such an important piece, Davina. ‘Claude was not holding the child, it was following them’ really hit home. Warmth without discernment isn’t care, it’s just smoothing, and children don’t have the context to know the difference.
Thank you for running this so carefully and writing it up so clearly.
Digital safety, especially for children, is a huge part of my work on DSS, so it’s genuinely heartening to see others approaching this with the same level of concern and care. We should collab in the future :) Looking forward to reading more of your work.
Thank you for highlighting this Davina. I spend lots of my time at work researching the harm that AI can do when used for functions for which it was not developed. The lack of empathy is a concern, especially when it is often construed as a voice of absolute truth amongst younger generations. Hopefully others will find this informative too.
Hi Jo, yes and that of course is the issue - when its true nature is misconstrued. I have been thinking about how to teach children about this. Perhaps they know better than we imagine! But there definitely are some children who are particularly vulnerable.
Thank you for the insight, Davina. I started using LLMs to talk about mental health and my trauma a few years ago, and I noticed this pattern too. Too often, LLMs echoed back what I wanted to hear rather than offered genuine psychological insight. It took seeing a real therapist to diagnose my CPTSD and treat it with exposure therapy. AI can simulate empathy based on pattern recognition, but it cannot comprehend the depth of human trauma.
We must raise awareness that LLMs' responses must be taken with a grain of salt.
You are so right Iris. It gives an illusion of something and that is worse than not doing it at all.
This is such an important piece, Davina. ‘Claude was not holding the child, it was following them’ really hit home. Warmth without discernment isn’t care, it’s just smoothing, and children don’t have the context to know the difference.
Thank you for running this so carefully and writing it up so clearly.
Digital safety, especially for children, is a huge part of my work on DSS, so it’s genuinely heartening to see others approaching this with the same level of concern and care. We should collab in the future :) Looking forward to reading more of your work.
Jade, Thank you. I think it just underlines how any safeguards that exist in AI are so easily sidestepped. And colab sounds great!
Thank you for highlighting this Davina. I spend lots of my time at work researching the harm that AI can do when used for functions for which it was not developed. The lack of empathy is a concern, especially when it is often construed as a voice of absolute truth amongst younger generations. Hopefully others will find this informative too.
Hi Jo, yes and that of course is the issue - when its true nature is misconstrued. I have been thinking about how to teach children about this. Perhaps they know better than we imagine! But there definitely are some children who are particularly vulnerable.