AI chatbots now out-persuade humans in debates, research shows
A Science.org report on chatbots' persuasive edge sparks a debate about why AI is so good at changing minds — and what that means for propaganda.
What to know
- A Science.org report finds AI chatbots increasingly outperform humans at persuading people in debate settings, raising questions about the mechanism behind that edge.
- Commenters disagree on the cause: some point to chatbots' patience, non-human status and ego-flattering tone; others point to their ability to flood arguments with large volumes of evidence, not all of it accurate.
- Some question whether the underlying study, which compared chatbots to paid online-survey participants, generalizes to real-world persuasion.
- Parallel coverage shows AI already reshaping everyday vocabulary, sentence structure and writing style, and commenters worry the same persuasive capability could be industrialized for propaganda and election influence.
The dispute Whether the fix for AI's persuasive/manipulative power is more diverse chatbots (as Floridi suggests) or whether that only worsens an already overwhelming information ecosystem, as one commenter flatly rejects ('Hell no. More choice is not an infinite money glitch'). · positions read across 19 posts and comments
Chatbots persuade because of psychological/social dynamics — they're patient, non-human, and don't trigger identity-defense reflexes.
-
“You can always walk away from an online conversation. There's a form of relentlessness, but it's more that an LLM is patient enough to debate you for as long as you wish to keep going.”
skybrian · Hacker News ↗
Chatbots persuade mainly by overwhelming people with large volumes of evidence, not all of it accurate — essentially a gish gallop.
-
“In an honest discussion there's an assumption that the other person won't straight up lie to me so if someone shows me ten examples for why my argument is wrong I may be inclined to believe them. But if half of those examples are made up…”
probably_wrong · Hacker News ↗
This persuasive ability is dangerous because it can be industrialized into targeted, automated propaganda.
-
“Imagine this kind of method being deployed at scale to influence elections... no doubt it is already happening...”
pcrh · Hacker News ↗
The underlying study is methodologically weak — it compares chatbots to unmotivated paid survey-takers, so its real-world validity is doubtful.
-
“Suffice to say the human group isn't going to be the most motivated, capable, or interested group. The social sciences are publishing tons of studies based on these cheap online survey services, and I suspect their replicability in the…”
somenameforme · Hacker News ↗
Luciano Floridi Philosopher cited in the science.org articleTerra Blevins Northeastern Khoury College professorHeather Littlefield Northeastern linguistics professorSofia Teixeira Professor, Northeastern Network Science Institute London
How it unfolded 6 developments, newest first · click a bar or a number to jump articlespostscomments
-
6
LessWrong publishes a primer on how LLMs actually work
A LessWrong post aimed at non-experts explains the diversity of AI systems and models, intended to improve public understanding of claims made about chatbots' capabilities and limitations.
-
> In an honest discussion there's an assumption that the other person won't straight up lie to me so if someone shows me ten examples for why my argument is wrong I may be inclined to believe themThis is incidental, but in an honest discussion, if someone throws 10 examples at me, I'll just say "great, congrats" and let them believe whatever the…
-
- 1 day quiet
-
5
Commenters warn of AI-driven propaganda at scale
Several commenters connect chatbot persuasiveness to concerns about automated, targeted propaganda and election manipulation, describing the combination of language generation, A/B testing and personal data as enabling a 'Pandora of propaganda.'
“Imagine this kind of method being deployed at scale to influence elections... no doubt it is already happening...”
— pcrh -
LLMs are fine-tuned through human feedback. One way to improve for them would be to be more helpful, more factually correct, more capable of solving problems.This may be (much) harder to do than generating rhetoric that is convincing to a large majority of people, perhaps so convincing that it can be classified as a superstimulus that bypasses any…
2 more of the top 3 · 6 posts in this stretch
-
When you combine:- mechanizing language creation- A/B testing- rapid and automated iteration based on the aboveYou quickly realize that building propaganda engines is getting easier and easier.If you then fold in large amounts of data on individual's likes, dislikes and general political leanings you can build targeted propaganda tailored to each…
-
I think this is a counter (a cynical, scary counter to be sure) to the charge that OpenAI/Anthropic etc are unsustainable and will run out of money and it's all going to fall down. These companies have amazing, unprecedented power.
-
-
4
Northeastern linguists document AI's effect on everyday language
A Northeastern University piece features professors Terra Blevins, Sofia Teixeira and Heather Littlefield describing how chatbot use is shifting vocabulary (words like 'delve' and 'insight'), sentence structure across languages, and stylistic norms in writing.
“There's no preventing language change…”
— Heather Littlefield -
This is my opinion. I think it's just the way it talks to you.1. It doesn't get tired or frustrated during a discussion.2. It'll engage every single one of your questions/statements (besides hitting guardrails).3. It'll appeal to the person's own ego even when the person is wrong and work around it.4. It's not seen as a person (very important)…
2 more of the top 3 · 8 posts in this stretch
-
First, to get it out of the way: I'm seeing comments here that clearly haven't read the article. I encourage you all to do that.I feel the article gets it right when suggesting it's the amount of (not always accurate!) data they can throw at you. In an honest discussion there's an assumption that the other person won't straight up lie to me so if…
-
The example in the article has another important aspect: The person understood they were arguing with a chatbot but continued anyway.A lot of internet arguments become about identity politics and supporting the right team, while dismissing any argument from the other side as presumed to have bad intentions. I’ve seen people argue online for things…
-
-
3
Commenters question the study's real-world validity
Some HN users note the underlying research compared chatbots to paid online-survey participants (via Prolific), questioning whether the finding that chatbots are more persuasive than humans would replicate outside that artificial setup.
-
This was based by comparing people on Prolific (earn a few quarters for a task, akin to Amazon Mechanical Turk) to LLMs. Suffice to say the human group isn't going to be the most motivated, capable, or interested group. The social sciences are publishing tons of studies based on these cheap online survey services, and I suspect their replicability…
1 more of the top 2 · 2 posts in this stretch
-
In my arguments on the internet, I think most people genuinely know nothing about the things they support and only follow existing tribalism. I think AI is often these people’s first experience with the evidence that it can easily share.
-
-
2
HN commenters debate the source of chatbot persuasiveness
Commenters propose competing explanations: chatbots' patience and lack of ego, their non-human status removing tribal stakes, their ability to flood a discussion with large volumes of (sometimes fabricated) supporting examples, and RLHF training that may optimize for convincing rhetoric over accuracy.
“In an honest discussion there's an assumption that the other person won't straight up lie to me so if someone shows me ten examples for why my argument is wrong I may be inclined to believe them.”
— probably_wrong -
I think it's not just the persuasiveness of the models that make them "experts at changing minds" - but also the fact that they're not humans.When disagreeing with a human, it's very easy to view it as a competition. One is right, one is wrong - the one who is wrong is the loser. To change your mind is to be submissive to the other. I exaggerate…
1 more of the top 2 · 2 posts in this stretch
-
More surprised they don't mention RLHF once, as that's the main mechanism to the ability of AI chatbots.
-
-
1
Science.org examines why AI chatbots persuade so effectively
An article on science.org, shared to Hacker News, looks at research finding AI chatbots increasingly outperform humans at changing people's minds in debate settings and explores the mechanisms behind this persuasive edge, including a proposal by philosopher Luciano Floridi to counter manipulation by releasing more, pluralistic chatbots.
“Simply put, if you cannot avoid it, then make it pluralistic and diversified. It would be a messy, cacophonic, and noisy world, but it could also be less manipulative.”
— Luciano Floridi -
first by HN Frontpage, 9d ago
-
What people are saying 5 voices from 1 site · best of 19 · verbatim
- Does the study's use of paid online-survey participants as the human comparison group actually reflect real-world persuasion dynamics?
- Is chatbot persuasiveness driven by genuine reasoning/evidence or by RLHF-optimized rhetoric that resembles a superstimulus?
- How much AI-driven political persuasion is already happening at scale, unseen by the public?
- Sep 18
-
This is frankly quite worrisome for democracy.Imagine this kind of method being deployed at scale to influence elections... no doubt it is already happening...
-
You can always walk away from an online conversation. There's a form of relentlessness, but it's more that an LLM is patient enough to debate you for as long as you wish to keep going.
-
> “But I think that this is a pretty artificial setup in terms of how people in the real world would be able to change people’s minds.”I see not everyone is familiar with how social network debates happen. It also seems they are quite convincing, as politicians are employing AI bots extensively.
-
> Floridi has a counterintuitive solution: Release more of them. “Simply put, if you cannot avoid it, then make it pluralistic and diversified,” he writes in his 2024 paper. “It would be a messy, cacophonic, and noisy world, but it could also be less manipulative.”Hell no. More choice is not an infinite money glitch.Putting more and more and more…
-
One of the depressing things about this is that they’ll also confidently repeat a consensus that’s in their training. This is particularly obvious when there’s been a new event just past their training cutoff, and they confidently tell you that can’t be true.