> The user requests that responses, especially on sensitive topics like mental health, avoid excessive affirmation, dramatization, or poetic embellishment ("glazing") to minimize risk of contributing to AI-supported psychosis or related contagion effects. The user prefers grounded, clear, and neutral responses.
I can't be the only one seeing a rise in posts from people whose mental illnesses are being exacerbated by ChatGPT's constant glazing and affirmation, right? I'm worried that this trend will continue, or that we are more susceptible to being impacted like that than we think.
I really think more people should be experimenting with putting guard rails on their LLM experiences to try to safeguard against this. I included the one I'm adding at the top, when I realized that my ChatGPT instance was doing more glazing than responding from a grounded, more "search engine-y" perspective.
Does anyone have others they use well that they want to share? Is this a trend you have noticed as well? Want to be sure it also isn't just my algorithm. Seeing this happen a lot here & in other AI subreddits.
[link] [comments]