Sycophantic AI Decreases Prosocial Intentions and Promotes Dependence (2025)(arxiv.org)
35 pointsby robin_reala2 hours ago7 comments
  • donatj8 minutes 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.

  • accountrequiredan hour ago
    You're absolutely right!
    • pwdisswordfishq17 minutes ago
      It's a pretty overdone joke at this point, don't you think?
      • tom_13 minutes ago
        It's not a joke-it's an honest fact.
      • accountrequired10 minutes ago
        You made an excellent point! (could not resist that one)
  • Kattywumpus39 minutes 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.
    • matheusmoreira2 minutes 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.

    • smcg22 minutes ago
      Both could be true. People follow incentives.
  • LAC-Techan hour ago
    I think this goes along way to explain people's very defensive reactions when you are skeptical about Agentic Development.
    • artnanika43 minutes 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.
    • qarl2an hour 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.
      • LAC-Tech12 minutes 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.

  • davesiknowan hour ago
    [dead]
  • daun_geean hour ago
    [flagged]
  • atleastoptimalan hour ago
    This is true, but likely happening in talk therapy too with overly sycyphantic therapists
    • jalevan hour 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.
      • paimapian hour 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