Didn’t We Already Know Asking AI For Advice Was Bad?

These are some of the questions I’ve been asked when I’ve talked to friends and family about my researching into asking AI/LLM chatbots for advice. A repository contains not only the paper’s rough draft, but the whole workflow and results for auditing and review. It can be found at GitHub or my personal git repository . This post may not make a lot of sense without reading one or more of those. Why Did You Spend Time On This? Don’t We Already Know This? There appears to be a specific gap in the literature looking at what happens when a regular person just pulls up a chatbot and asks it for advice. Not a model trained for therapy, without support from a mental health clinician, and outside of a structured therapeutic setting. We know it happens; you may have done it yourself. I also wanted to see if if there was a simple way to tweak the prompt to see if that could override the problems noted in other settings and studies. I don’t use LLMs for personal advice, but I do use them. When I…