Ask HN: Add flag for AI-generated articles Should HN add the ability to flag articles as AI-generated? This doesn't have to act as a regular flag, i.e., it won't de-rank the article; it could just show up as an indicator, allowing others (like myself) who don't like reading AI-generated text, to skip it. Open questions: 1. Why is the regular voting system not enough? 2. Should HN change in response to the gen AI era? It has been successful not changing fundamentals. |
Voting systems can be gamed and as HN becomes bigger and bigger it'll start to attract unsavory audiences who have an agenda.
We don't have a similar rule yet about article content but my sense is that the community mostly doesn't want to read it—or, to put it more conservatively, discounts it. This is why we see so many "just show me the prompt" responses, along with others like this: https://news.ycombinator.com/genai-pushback. I built that list so I have something to send to users who email about why their genai articles got flagged.
It's a fascinating arms race right now: the AIs are training on the humans but the human hivemind is also training on the AIs. Readers are developing allergic sensitivities to language that sounds like an LLM produced it. The AIs will adapt to this, but the humans will adapt in turn. Where it ends up is anyone's guess. I have an optimistic view, but I've already been wrong about this so many times that I have low confidence in it.
The current picture is that there is an emerging class distinction between writing (and writers) that use genai vs. writing that does not. As soon as the "this sounds like an LLM" allergy kicks in, the writing instantly gets relegated to a low-status bucket in the reader's mind. That doesn't mean it won't still get looked at - but it is now under a stigma.
(I was rather pleased with the originality of this until I remembered pg had come up with "writes and write-nots" in https://paulgraham.com/writes.html. Oh well, it's the point that matters.)
This has the happy flipside that anyone who would like readers to classify their article as high-status rather than low-status can apply the judo move of simply writing it themselves.
Now I need to add the disclaimer that none of this is a dismissal of LLM technology per se. We rely on it heavily, and there's no question that it's useful. The question is how to use it (pg again: https://x.com/paulg/status/2058871512451412457) and whether one should use it on writing that one publishes to other humans.
To turn to OP's questions:
> Should HN add the ability to flag articles as AI-generated? [...] it could just show up as an indicator
Flagging-as-just-an-indicator would be tagging, which we've always resisted adding to HN, but I wouldn't rule it out.
What I do think we'll (finally) add is a "please give a reason why you flagged this post" step, and "because I think it's genai" will be one choice among several (spam, offtopic, mean, etc.)
> Why is the regular voting system not enough?
The regular voting system is never enough. https://hn.algolia.com/?dateRange=all&page=0&prefix=false&so...
> Should HN change in response to the gen AI era?
To this I am tempted to reply with https://news.ycombinator.com/item?id=48887149 in an homage to https://news.ycombinator.com/item?id=3742902.
Does that mean that an article being AI-generated is a flaggable offense? Should we be flagging suspected AI-generated articles already, or should we wait for the flagging system to support reasons first?
I don't know for sure yet. But you're right that that would follow from what I said.
What I do know is that users should flag articles that they think don't belong on HN. But this should rely on their reading of https://news.ycombinator.com/newsguidelines.html and not just be based on a like/dislike reaction. That is, users should flag articles that they think break the site guidelines, not just ones they dislike.
I can see a grim future (present?) where "AI generated" turns into a slur, warranted or otherwise, in a world where the difference between human trained to talk like an AI and AI masquerading as human becomes increasingly difficult to discern, and some hidden cabal passes judgement.
That is wholly different from taking a stance on HN being a place for humans to comment on articles.
For example, I think you said "the politics of articles" before I added the thing about a "class distinction". Intriguing overlap!
I don't quite follow what you mean about grim future but if you wouldn't mind reading the edited version of my post, I can respond to anything that isn't addressed there.
Personally that is the future I hope to see. Hence my continuous protestations at the lazy excuses for articles that get posted these days, and the lazier excuses given by their authors for avoiding using their brain.
It’s okay to have an opinionated website. Not every corner of the internet needs to be a bastion of free speech
Stuff interesting enough to get upvoted (i.e. not slop), but I'm so irritated and triggered by pointless, human, comment threads moaning about "this is AI generated".
Do you believe adding friction to flagging will reduce the quantity of low quality articles?
Or is the flagging of high quality articles a bigger and more pressing problem?
Or is the problem simply too-many-damn-flags?
Just curious.
I really really hope more people take up pen and paper! My last blog post [0] came with proof-of-work attached.
That's not a value-judgement of the true content of an article or piece of media, but a fairly objective facet which impacts the HN participant's experience.
It's kind of like how, back in the day, people really wanted to know the filesize of something before they clicked, to avoid a blind-investment of their dial-up bandwidth for indeterminate minutes of waiting (and opportunity-cost of other things not downloaded) that might be more than they really wanted for whatever-it-was.
But then again, there’s always reddit :)
Single page info-graphics and Awesome-Lists come to mind.
Hell, one or more of PG’s books has one or more baysian generated texts presented as poems.
1st: the presumption that AI generated text is actually unsuccessful, rather than proliferating broadly unchallenged today
2nd: the disposition that negative attitudes towards AI text are unjustified discrirmination, rather than working as a strong latent predictor of low-effort content
3rd: the assumption that human writing is reliably doled benefits, rather than some poor proxy of it (winning the social contest for claiming authenticity)
Operationally, only a very small minority of humans actually successfully identify AI generated text at rates ≥ Pangram. People discriminate against the label of "AI", but mostly fail to vote accurately. It's not uncommon to see bots abusing this gap for their own success -- accusing humans, sympathizing with generated profiles... FUD environment where people routinely get away with dismissing true accusations.
For someone who is mediocre at detection, this would structurally feel like an unhinged, unjustified bias: look at all these good posts, these honest people, getting undermined by discrimination...
> 1st: the presumption that AI generated text is actually unsuccessful, rather than proliferating broadly unchallenged today
I can only talk about HN. If you think this is proliferating broadly on HN itself, I'd like to see such links. Assuming it's "broadly", they should be easy to find.
> 2nd: the disposition that negative attitudes towards AI text are unjustified discrirmination, rather than working as a strong latent predictor of low-effort content
I don't understand this bit.
> 3rd: the assumption that human writing is reliably doled benefits, rather than some poor proxy of it (winning the social contest for claiming authenticity)
Sorry, but I don't understand this either.
> only a very small minority of humans actually successfully identify AI generated text at rates ≥ Pangram.
That's not what we seem to be seeing. I do agree that there's a wide spectrum and a lot of wrong guesses.
> It's not uncommon to see bots abusing this gap for their own success
I'd like to see specific links of this on HN itself. If it's not uncommon, they should be easy to find.
I don't think this is true, at least not right now, and in a way I'm actually thankful for it.
The frantic rush to chase the only potentially profitable use case for LLMs found so far (writing code) and the resulting focus on coding RLHF means models are actively becoming worse at sounding like humans.
This is my favorite example, and it's already relatively outdated: https://progress.openai.com/?prompt=10
Before you say I'm just falsely calling them out, it's typical ChatGPT style of either very amicable or Nobel Laureate tone, lots of formatting, with a couple of paragraphs and then a clever one-line punchline at the end. If you look at those commenters their history, it's all like that. Either generated or assisted. For older accounts you can see the steep increase of it around 2025ish.
Seems like the HN crowd absolutely adores AI comments and the rule banning them is (sadly) unnecessary. Or at least not what 'the people' want.
The OP then replied:
> Not AI. Not sure how I feel getting my writing style called out like that though :D
The HN guidelines[1] include:
Please don't complain about tangential annoyances—e.g. article or website formats, name collisions, or back-button breakage.
Please don't post insinuations about astroturfing, shilling, brigading, foreign agents, and the like. It degrades discussion and is usually mistaken.
All we really need is to add "Don't complain that an article is AI" to the guidelines.
Maybe we need a two-dimensional voting system: good/bad, ai/human. I think the second axis could cut down on meta-discussions over how much of the article was AI-generated.
I imagine the set of articles that are somehow both interesting-enough-to-read but not interesting-enough-to-write is smaller than you'd think.
If there is a great post on a topic and the author used AI when generating it, what’s so bad about that?
agreed.
For me, the issue I have is that a vocal group seems to despise AI-edited content and they can't manage to take their disgust eleswhere.
AI-editing is another tool, just like spellcheck.
I think the era of the blog is simply dead now and that’s mostly ok. Blogspam and corporate blogs had killed quality bogs ages ago even before AI was a thing. The real question is what replaces it.
Oh and of course the $64k question is this: if an AI generated article is indistinguishable from a human written article and it is accurate and interesting, do you care who wrote it? We want to avoid low quality, not AI generation, right?
having said that, i'd still much prefer a norm of including prompt history with the article, or the codebase for that matter, so people can choose for themselves :)
Whether you agree that a con artist is only a criminal if he gets caught, or you think he’s a criminal the whole time, surely you can see why many people might want to know if they are dealing with a friend and not something simulating a friend.
Hacker News adopting such a feature would likely do more harm than good.
https://github.com/alvi-se/ai-ublock-blacklist
That's a good example of the exact problem with such a broad stroke rule.
How do you tell which is the case?
If we don't allow AI help at all, is that perhaps discriminating against those who don't feel comfortable posting with imperfect English?
I agree in principle, but am concerned in implementation... I'm not sure we can be fair without high risk of discrimination
Edit: typo fix
Edit: or am I AI?! And making edits looks more legit.... (To be clear: I'm not, I play by rules)
I have a hard time finding these communities
The issue is complicated by the fact that there can be substantial effort invested in a process outside of the writing itself - and so AI written does not guarantee that the content will not valuable. But I'm inclined to punish it anyway to establish a norm of valuing genuine human communication. I think this norm has always been present but we didn't know until we'd really explored the alternatives.
I spend a LOT of time reading AI generated content because I use AI a lot for various purposes - maybe I'm more sensitive to its voice than some. AI voice always bothers me and its been getting more annoying the more I notice it, but there is a huge difference in reading responses to my own prompts and in reading the response to a prompt I haven't seen, when I don't know how many revisions there were, when I don't know if a human mind reviewed it at all before clicking send.
It becomes an unacceptable distraction because I don't know if I'm investing more time in the content than the author did, when in normal written communication the author would be putting in at least 5x the work.
My entire position: I'm not interested in reading text that sounds like it was written by AI.
The latest codemaxxed models all tend to write in very distinct, instantly recognizable ways unless carefully instructed otherwise (honestly a good thing if you want to avoid wasting time reading AI text). A great example is this submission that's currently #1 on the front page (which is also just a thinly veiled advertisement): https://news.ycombinator.com/item?id=48884853
I think it would be great if AI-generated submissions were outright banned as they fundamentally break the balance of effort that HN was built upon, but as was already stated by another user, YC is heavily invested in AI so there's a conflict of interest there.
And im sure this was designed in order to encourage positive discussion
2. judge content not by its cover and think.
AI writing is not the problem - low effort is the problem. Low effort AI articles are full of tics which are obvious, if you've done a lot of AI writing. To write well with AI you need to spend a good deal of time editing.
If you submit something that's low effort but has a clickbait headline that appeals to HN, you may well make the front page even if the article is lightweight (it does happen!) This is true both for AI and for human written articles.
On the flip side, somebody could spend an enormous amount of effort creating a masterpiece with AI. Penalizing that because of the tool that was used is arbitrary.
I do think the dead giveaways (em dashes, it's not X it's Y, etc.) are annoying to come across repeatedly. A person not bothering to remove these tells feels 'low effort' to me.
It got to the point I'm sitting on reviewed and tested patches for Mesa that I'm too ashamed to submit because of Claude's participation in their making.
AI slop is AI slop.
However, a lot of the academic or technical posts on this site have turned out to include AI hallucinations (a problem that is not just on HN, in fairness). If an article or paper contains nonexistent or obviously, blatantly wrong citations, then I feel very strongly that such content is disrespectful to the HN audience, certainly to me, because I've just wasted my time trying to take the author seriously only to discover that the argument was founded on a hallucination. In a sense, it doesn't matter to me whether the hallucination was due to AI or not -- if a person puts in bogus citations, AI-generated or otherwise, my view is that their article or paper fully deserves to be flagged so that we don't waste other people's time.
Is it purely just a "human supremacist" desire that fuels the motivation to ban or block such articles?
A simple beneficial step that would lead to modest improvements and little downside: partner with Pangram. Either adding it as an automated spam filter, or by simply attaching the detection % to all posts.
Humans? We're not particularly effective at this as a whole...
AI service ? We'd probably have to pay for that AI to detect that AI and well.. Its also not particularly effective
Effectiveness is important, because we dont want real human produced data to be accidentally removed from view, just as much if not more so than having AI gen data being left on the site.
Conspiracy minded responses are low value, and yours is especially so considering HN bans AI comments.
"Submit to my knowledge or else!" is abusive.
People can drive a car without needing an expert copilot. Why should they need a software engineer to use a computer?
Spare the appeals to history as the historical record would show software engineers have unemployed many others. Technology moves on; rotary phone makers and travel agents have a seat for in their support group.
Your self selection and vanity could not be more obvious.
AI could provided assistance with building both cars and software, but you still need a decent level of knowledge in either field to get a good result.
LLMs are mass theft of intellectual property. I didn't need them to "drive" this computer for the last ten years.
Just stokes addiction to a false sense of empowerment and social contribution
Downvote away to hide from ideas and opinions you don't like.
Conservative people are the only people so dysregulated by ideas they don't understand or that do not validate prior experience and embedded biases they just try to hide them from view. Not the kind of people whose judgment I concern myself with in any meaningful way
The massive 1000+ comments/upvotes AI model release threads only show up on HN. Lobsters doesn't accept "business" articles like that. You're quite likely to find reactionary articles that are critical of AI on the front page though. There are several on the front page right now, in addition to the ones derisively tagged vibecoding.
I'm not picking on you - it's practically a universal response, so much so that it must be driven by human hard-wiring. I've written about this so many times that for once I don't even know what to link to. Perhaps https://news.ycombinator.com/item?id=23308098. Or maybe https://news.ycombinator.com/item?id=48427800
Slashdot is still around
So the guidelines are in some sense a red herring.
".. Fault-tolerant and highly available hardware must facilitate low-latency, single-threaded communication with high semantic density in order to achieve multi-dimensional consensus in a safety-critical, heterogeneous, adversarial environment. .."
I am not sure why you think someone saying "not AI trust me bro" carries any merit.
At any rate, like I said, I've given up the war. People enjoy reading that stuff, I'll just be the old man no longer yelling at the clouds.
I think the criticism also signals to the submitter or idk co-author? That the article isn't valued
> Please don't use HN primarily for promotion. It's ok to post your own stuff part of the time, but the primary use of the site should be for curiosity.
Not every person can drive a car.
> but they can't build a car.
Not every person is unable to build a car.
Eg: my father, born 1935, still alive, has built double axle trailers, cars and vehicles that work, medical caravans for St Johns Ambulance, ...
His education was six years of primary school and these were all "side projects" for himself or for his community.
> you still need a decent level of knowledge in either field
and the barrier to picking up sufficient knowledge to safely create software, cars, even ground effect flying boats isn't that high for normal human beings .. although it does require a certain can do attitude.
They are driving (lol) towards the sale of a computer with a model on chip without a programmer salary and benefits adding to cost and resource use (all the tools of the trade need to be stored and copied around).
This is capitalism, not honorific obligationism! Saving your job is no one else's burden! Disrupt! (YC crowd 15 years ago). Oh how the turn tables.
Move the context of the code to the presentation layer and eliminate state by keeping the hardware focused on computing geometric transforms. Biggidy bam label the data on the display and save both geometry and labels
There's no reason for biz logic context to exist in code.
Depends on your use-case, but most people don't need content used in network nuisances like YC AstroTurf posts. =3
Even scoring posts does provably change peoples behavior, and studies showed it tended to make people more punitive in their conduct with strangers.
At bare minimum, YC should have bot CAPTCHA protection on posts. As unleashing chat bots on users is often disrespectful, and just cows people out of participating in conversations in good faith. =3
Are you implying the front page is botted/manipulated?
Then it was to protect our modems bandwidth (cost and time) while now it's more about protecting our own cognitive bandwidth.
I agree. We've had large and thriving online forums for at least a decade before upvote systems became ubiquitous and things were just fine.
yeah keep chatting with the sycophantic psychosis-inducing LLM chatbot, I think you'll have a better time.
"You are the average of the five people you spend the most time with."
(Subconscious training like when we pick up an accent, though eventually folks might automatically code switch - so that’s hopeful.)
> It's a fascinating arms race right now: the AIs are training on the humans but the human hivemind is also training on the AIs. Readers are developing allergic sensitivities to language that sounds like an LLM produced it.
Humans are training to detect AI content. Humans writing more like AIs is an unrelated (and slower) phenomenon.
If this is true then this statement should be added to the guidelines. Many a time I've seen a submission and thought "My god, this is the most moronic thing I've ever read in my life, nearly every single factual claim in this is wrong" but I don't flag it because "don't post utter drivel" is not in the guidelines.
So either I read your mind, or it is best not to flag unless something is obviously, unequivocally, running afoul of the guidelines.
I keep being annoyed at how arbitrary, and dare I say hypocritical, the entire flagging mechanism and its surrounding politics are.
People love to make vague claims like this based on things we supposedly said. How about you quote what exactly what I literally said so readers can make up their own minds?
Sorry if that is testy (well, it is testy), but I can't count the number of times that disgruntled users have posted falsely to HN threads to vent their residual frustrations with the mods.
> I'm afraid we took vouching privileges away from your account because you vouched for too many comments that were unsubstantive and/or flamebait and/or otherwise broke the site guidelines
> If you want to build up a track record for a while of vouching for good comments only, and then email so we can look at the recent vouches
Wondering how does one tell what is ‘good’ in your mind:
> Btw if you're unsure about a case you can always check with us about it. I know the borderline cases are not always easy to call.
—-
In other words, do as dang himself would do, or get penalised. I am not in the business of mind reading.
This was years ago. I am pretty sure lately I have been flagging/vouching stuff I genuinely believe were OK, though you have all the data in front to smite me with “ah, but on this day you vouched this bad comment! Gotcha!” so in the end it’s always a losing battle.
In any case, my issue is with you saying, I quote, ”users should flag articles that they think don't belong on HN”. Emphasis mine. It is more nuanced than that.
I know you hate the supposedly "noob" comment, as you've put it over the years, and as it is written in the guidelines, that HN is or isn't turning into Reddit, but this aspect of HN makes it indistinguishable to me as a long-time reader here whether you accept it from your perspective as a long-time moderator or not.
Because HN moderation allows people to downvote or flag things simply because they don't like something, it will always be like Reddit. There is no distinguishing feature to separate it otherwise.
One thing that would, would be forcing users to say why they downvoted or flagged something.
Edit: If you spend years saying "no, no, it isn't true" to people, you're just sticking your fingers in your ears. Where there's smoke, there's fire.