Comment by ryukoposting
3 days ago
OP's classifiers make two assumptions that I'd bet strongly influence the result:
1. Binning skepticism with negativity.
2. Not allowing for a "neutral" category.
The comment I'm writing right now is critical, but is it "negative?" I certainly don't mean it that way.
It's cool that OP made this thing. The data is nicely presented, and the conclusion is articulated cleanly, and that's precisely why I'm able to build a criticism of it!
And I'm now realizing that I don't normally feel the need to disclaim my criticism by complimenting the OP's quality work. Maybe I should do that more. Or, maybe my engagement with the material implies that I found it engaging. Hmm.
OP here :) On skepticism being lumped with negativity: partially true. The SST-2 training task treats critical evaluation as negative sentiment. I should clarify that "negative" here means evaluative or critical, not hostile. HN's culture of substantive critique registers as negative by these metrics, but that's arguably a feature of technical discourse rather than toxicity.
On the neutral category: the model outputs continuous scores from 0 to 1, so neutrality does exist around 0.5. The bimodal distribution with peaks at roughly 0.0 and 0.95 reflects how HN users tend toward strong evaluative positions. Three-class models could provide additional perspective, and that's worth exploring in future work.
Also love your meta-observation. Imo your comment is critical, substantive, and engaging. By sentiment metrics it's "negative," but functionally it's high-quality discourse. But that's exactly how I read the data: HN's negativity is constructive critique that drives engagement, not hostility.
Is it “negative” though? I ran it through this model and it gave 99.9% positive. (You tell me if this model is substantively different from what you used.)
https://huggingface.co/distilbert/distilbert-base-uncased-fi...
- "negative" here means evaluative or critical, not hostile.
this is so far from how people are interpreting your results that I'd say it's busted. your work might be high-quality, but if the semantic choices make it impossible to engage with then it's not really a success.
As a native born english speaker, I disagree completely. It's very obvious what he means. This is a severe reading comprehension problem, not a problem with the author.
> What counts as “negative” here? Criticism of technology, skepticism toward announcements, complaints about industry practices, frustration with APIs.
1 reply →
So the group synced a dumb bias. It must change. Not the author; they provided qualitative evidence that was not their intent. Update your opinion and perspective with that new evidence.
Imo it's on the individual members of the groupthinkers to realize a math term (negatives are a thing in math) applied to mathematical data is not a qualitative attack on anyone; they must accept the groupthink has lost the plot.
Consensus isn't always preferable. See religion.
The context is obviously a mathematical analysis and math comes with negatives.
If the critiques had actual substance to contribute to the world they wouldn't be so easily offended. Publishing low effort complaints that are little more than demands by far away randos to better to conform to their arbitrary standards is a laughable expectation. Internet randos can pound sand; they prop up nothing individually or collectively given most forums are a few thousand to tens of thousands of unique people with a platform but no real democratic power.
Social media hyper-normalizing sentiment is just empowering social bullying by pressuring people doing the necessary work to think include the bike-shedding of non-contributors. Whole bunch of farm animals want to eat bread while letting the rooster do the work.
2 replies →
I'm interested in seeing a plot of that percentual over the years. The past 3 or 4 years I've been seeing less and less tech savvy comments over here and this data seems a great way to find out if it's just placebo.
Maybe version 2 will be better
There's a lot of legitimate criticism, but they're also a noticeable amount of "reply guy"-ism (pedantry that can sometimes derail the conversation and bring nothing of substance) and "pet peeve syndrome" (people who'll always repeat the same criticisms of a product or company even if completely off-topic for a submission). It would probably be hard to classify them. And whenever I go on reddit I see that both are enormously worse on there.
are the "reply guy" and "pet peeve syndrome" considered bad? because i think they're a big part of the value i get in the HN comments.
my mental model of the comment section on any page is "let's get all the people interested in this article in the room and see what they talk about" rather than "here is a discussion about this article specific"
i also assume that <50% of the people in any comment section have actually read the original article and are just people hopping in to have a convo about something tangetially related to the headline.
The flourishing of pet peeve syndrome means that once you understand the hivemind, you can predict the comments just from the headline alone.
It is bad. Its fine when you read the opinions for the first time, but after a while you notice over half of the discussion is just repeating the same off-topic talking points over and over and over again on every submission.
"Off topic but...", "The CSS on this site...", "Have you noticed this web page is 5/10/20MB...", "Why is the webpage making 20+ requests...", "Microsoft bla bla bla...", "Elon Musk bla bla bla...", "I havent used cryptocurrency at all and it is bad because..."
All of these on some post about a new library release or government policy or social commentary.
I was about to make a comment about skepticism, thank you for adding it. Its likely that its all bunched in together. Looking at material with a critical eye is a positive feature of HN not a negative - thats a very very nuanced thing to evaluate though and likely we do not have the technology
I’m not sure if this is the exact model used by OP, but it appears close, and it classifies your comment as positive at 99.9%.
https://huggingface.co/distilbert/distilbert-base-uncased-fi...
I'd be curious to know how my comment scores if you cut it off after the bullet points.
Hi, if you find the time, please reach out to the email in my bio with the methodology and dataset you used. Did you only test one post or a whole set? Would be interesting to compare the setup. Thanks!
I linked to the model in my comment
There also is some conflict of interests. VC investors and marketeers obviously want to nurture optimism. And that is entirely understandable and very likely necessary for good ideas to be spread.
While engineers, especially those that like to share knowledge and open source solutions are far more critical of monetizing products.
Overall HN doesn't lack criticism, since there many technically minded people around. But I like the mix to be honest and agree that skepticism is often seen as simple negativity. Sure, you probably don't want to advertise your product as "pretty decent, but there are numerous better theoretical solutions".
HN is not the rest, it is not the majority. It's for a specific tech-savvy social category. This category does want skepticism and criticism because they tend to be perfectionists. This is not "negative sentiment" anything but very positive "evrika!" sentiment for members of the aforementioned category.
Would one say: nice attempt trying to tell people how they are supposed to feel around here?
You comment is very interesting observation. Its made me reconsider some things about sentiment analysis. You are right its not really that HN is negative its that sentiment analysis doesn't really have any way I can think of offhand to measure meaningful discourse rather than GOOD/BAD/NEUTRAL
I've never done any sentiment analysis outside of hobby tinkering. Maybe there is some HN experts that will chime in on how to deal with it?
> The comment I'm writing right now is critical, but is it "negative?" I certainly don't mean it that way.
It's a matter of perspective. The OP is a negative post. You are negative about it. Therefore, you have made a positive post.
Exactly, critical thinking is framed as a negative sentiment in that analysis.
Perhaps add joke and off-topic as labels too
Negative is negative, regardless of intent. Here's the llm positive way to write your post:
It’s a great exercise to reframe constructive feedback. Here is a more positive, affirming version of that post that maintains the original insight while shifting the tone to one of appreciation and partnership. A Positive Reframing
ryukoposting 17 hours ago | parent | context | flag | on: 65% of Hacker News posts have negative sentiment...
The OP has done a fantastic job putting this together! It’s such an interesting dataset that it really invites deeper exploration into how we categorize human speech. I think we can make this even more accurate by looking at two exciting opportunities:
* Celebrating Skepticism: We could distinguish between "negativity" and "healthy skepticism." Often, a critical eye is actually a sign of deep interest and a desire to refine a great idea.
* The Value of "Neutral": Adding a neutral category could highlight the balanced, objective discussions that happen here, showing just how nuanced the community’s input really is. I’m writing this because I’m genuinely inspired by the quality of the presentation and how clearly the conclusions are articulated. It’s exactly that clarity that makes it so easy and fun to brainstorm improvements! I’m realizing now how much I enjoy engaging with high-quality work like this. It’s a reminder that even when we’re being analytical, it’s because the original content is truly engaging. Kudos to the OP for sparking such a thoughtful conversation.
Key Changes Made: * From "Assumptions" to "Opportunities": Instead of pointing out flaws in logic, it frames the points as ways to build upon an already strong foundation. * Emphasis on Inspiration: It explicitly states that the criticism is a result of being impressed by the work, rather than just "not meaning to be negative." * Active Appreciation: It turns the "Maybe I should do that more" realization into a proactive statement of gratitude for the OP’s effort. Would you like me to try another version focused on a specific tone, like "professional" or "enthusiastic"?
...
Even if you hate it, the vibe of that is completely different.
>Even if you hate it, the vibe of that is completely different.
It's so mealymouthed my internal sentiment analysis grades it as insanely toxic.
Nobody actually talks like this. And if they do they have a terrible office culture.
Yes, the original post has the vibe of something a human wrote to express an idea, while your version has the vibe of meandering, insincere, sycophantic AI slop that obfuscates the original idea in service of congratulating everything.
Both express the same basic criticism; you've just replaced the neutral tone with something that's perhaps more effective as a vomitory than as a criticism.
Rather than AI slop the above comes across to me as genuine corpospeak. I guess the task wasn't so much generation as it was translation. I found myself simultaneously impressed and disgusted.
I wonder how well an automated tool to go in the reverse direction would work in practice? With an accompanying style transfer GAN to rewrite the Corporate Memphis hellscape.
I think you're on to something about bad categorization. Sentiment analysis as practiced is almost always pseudoscience. (And that's a hedge. I've never seen it done right but I'm not outright discounting the possibility.)