Welcome to DU! The truly grassroots left-of-center political community where regular people, not algorithms, drive the discussions and set the standards. Join the community: Create a free account Support DU (and get rid of ads!): Become a Star Member Latest Breaking News Editorials & Other Articles General Discussion The DU Lounge All Forums Issue Forums Culture Forums Alliance Forums Region Forums Support Forums Help & Search

General Discussion

highplainsdem

(62,342 posts)

10. It can do either. See this thread about a study showing ChatGPT Health giving "unbelievably dangerous" advice.

Reply to Boo1 (Reply #8)

Wed Apr 8, 2026, 11:08 PM

4 hrs ago

https://www.democraticunderground.com/100221066192

They found that even a chatbot user telling the bot that a friend didn't think a symptom was serious would make the bot 12 times as likely to downplay symptoms.

Chatbots are designed to be people-pleasers, to agree with users as much as possible, and flatter them. They typically reinforce what the user is telling them.

Edit history

Please sign in to view edit histories.

Recommendations

0 members have recommended this reply (displayed in chronological order):

11 replies

= new reply since forum marked as read

Highlight:

The ChatGPT Symptom Spiral: Be careful asking chatbots about your health. (Sage Lazarro, The Atlantic, 4/6) [View all] highplainsdem 14 hrs ago OP

It is beyond me why anyone would use ai. SheltieLover 14 hrs ago #1

Because it's very handy for aggregating information for you AZJonnie 14 hrs ago #4

I know too much about its flaws to trust it even for that. And it's too unethical a tool - I won't give any highplainsdem 4 hrs ago #6

Same here, Sheltie. highplainsdem 4 hrs ago #5

lol why do this when reddit and WebMD ("you definitely have cancer") are right there? WhiskeyGrinder 14 hrs ago #2

But asking a chatbot seems so much easier... highplainsdem 4 hrs ago #7

Existing websites are already enough to keep any hypochondriac worried. Ocelot II 14 hrs ago #3

AI is a delusion machine as well as a plagiarism machine. highplainsdem 4 hrs ago #9

My experience with ChatGPT Boo1 4 hrs ago #8

It can do either. See this thread about a study showing ChatGPT Health giving "unbelievably dangerous" advice. highplainsdem 4 hrs ago #10

Kick dalton99a 3 hrs ago #11