Sycophantic AI Decreases Prosocial Intentions and Promotes Dependence (2025)

rrobin_reala about 4 hours ago 50 commentsRead Article on arxiv.org

HI version is available. Content is displayed in original English for accuracy.

⚡ Community Insights

Discussion Sentiment

43% Positive

Analyzed from 680 words in the discussion.

Discussion (50 Comments)Read Original on HackerNews

donatj•about 2 hours ago

> This suggests that people are drawn to AI that unquestioningly validate, even as that validation risks eroding their judgment

I believe this is a larger problem than just AI.

The internet has helped people surround themselves with only voices that agree with them and validate them, often to their detriment.

There are entire online communities urging people to cut others out of their lives over the slightest disagreement.

overgard•about 1 hour ago

Yeah, a lot of it isn't even really about people's choice though. Like, I have so many click-baity things in my YouTube feed even though I ignore almost all of it, because just one click and suddenly that's a thing it's going to show me 10x. I wish there was a place I could go for tech news that just didn't talk about AI. I'm so over it.

jaggederest•about 1 hour ago

Agreed, I think AI is the bees knees but I already spend 6+ hours a day thinking about it, more is unhelpful. I wish for something with the quality of discourse of HN but without being so monomaniacally focused on tech-du-jour.

darkmarmot•about 1 hour ago

It's like it turns everyone into a CEO with all the yes men they could ever ask for! Never goes wrong!

Kattywumpus•about 3 hours ago

I don't know. Upvotes often do the same thing, especially in siloed communities. I feel like this is a larger problem -- people leaning into extremifying their views or just playing to an audience -- even though one might frame it as "prosocial" because it happens within the context of an online community.

matheusmoreira•about 2 hours ago

> Upvotes often do the same thing, especially in siloed communities.

Yeah. I've gotten caught in this pattern here on HN. Post something, there's some disagreement but then I see the quiet upvotes trickling in, sometimes by the dozen. That sort of signal slowly convinced me I was right about plenty of stuff.

I used AI to datamine all of my HN comments and find those patterns. Then I started asking the AI to steelman the hell out of my world view, every single argument.

gryfft•about 2 hours ago

How would you say your views have changed since then?

smcg•about 2 hours ago

Both could be true. People follow incentives.

accountrequired•about 3 hours ago

You're absolutely right!

pwdisswordfishq•about 2 hours ago

It's a pretty overdone joke at this point, don't you think?

tom_•about 2 hours ago

It's not a joke-it's an honest fact.

jihadjihad•about 2 hours ago

I dunno. The seams of the joke are still load-bearing IMO.

accountrequired•about 2 hours ago

You made an excellent point! (could not resist that one)

almostdeadguy•about 2 hours ago

You're right to push back. And honestly, this reframes the entire thing.

atleastoptimal•about 4 hours ago

This is true, but likely happening in talk therapy too with overly sycyphantic therapists

jalev•about 3 hours ago

Sure, but there is a marked difference between a therapist and an LLM Chatbot: you can't talk to your therapist for +4h a day, at any time convenient to yourself. It's also not a guarantee you're going to hit upon a therapist that is like that, versus a product that is very literally intended to maximise your usage of it.

paimapi•about 3 hours ago

a therapist should, if they follow their training, not be a sycophant whatsoever. CBT, for eg, is a multi-step process including Socratic self-dialogue, reframing practices, etc to guide the person towards the development of insight and healthier coping mechanisms

LAC-Tech•about 3 hours ago

I think this goes along way to explain people's very defensive reactions when you are skeptical about Agentic Development.

artnanika•about 3 hours ago

Those are two different things. Chatgpt being sycophantic while playing the role of a therapist with a teenager is much different from Codex being sycophantic while reviewing user feedback. The former is extremely dangerous for society.

LAC-Tech•about 2 hours ago

I make no societal judgments; just that the nastiness of the reactions I get suggests something more than technical choice is going on.

I had one prominent "influencer" publicly challenge me to a coding competition over it. It was very weird.

qarl2•about 3 hours ago

I'm one of the people who gets accused of this. I don't mind skepticism. But every time I turn around, someone's claiming nobody gets any value from these tools and anyone who says otherwise is deluded or running a scam. That gets really old.

sidrag22•about 2 hours ago

I've had one conversation that was sorta like this, it was around the time some kid kept releasing videos of him interacting with a very bad TTS gpt model or something and it would give very very bad answers confidently.

Their frame of reference was that, and this was at a time when opus 4.5 was around. Hard to expect a reasonable conversation when one person has such a different view of what the tools are capable of.

LAC-Tech•about 2 hours ago

I can believe people get value out of them. I think I do.

But "OH MY GOD EVERYTHING HAS CHANGED THE OLD WAY IS DEAD 10X MORE PRODUCTIVE" - no. If that were true, we'd notice it in the software all around us.

qarl2•about 2 hours ago

I think we are noticing it. Are you not?

I've shipped three software projects in the last two months. I shipped zero in the preceding year.

Right now I'm building a system to decompile MAME ROM games into idiomatic JavaScript. It completes one game in about 3 days - complete with extensive commenting. I started it about 2 weeks ago.

I'm not sure "EVERYTHING HAS CHANGED" but if you're not seeing dramatic change, I suspect you're not looking.

chrisjj•about 1 hour ago

> But every time I turn around, someone's claiming nobody gets any value from these tools

Really? I find agreement that gullible users get something they value, even if it is worthless.