HACKER Q&A
📣 kerkeslager

How can we get people to read submissions before commenting on them?


Over and over, top rated comments on submissions are by people who clearly didn't read the thing they're commenting on. The longer the submission, the more uninformed the commentary.

To make matters worse, I frequently see people who say they come to HN because of the quality comments. There's a Dunning-Kruger effect happening, where people don't read the submissions, so they don't know that people aren't reading the submissions, so people think they're becoming more informed by reading the comments when in fact the comments are just as uninformed as they are. There are exceptions of course, but they are becoming more and more exceptions, when they would ideally be the rule.

I think we can do better.

The commenting guidelines say:

> Please don't comment on whether someone read an article. "Did you even read the article? It mentions that" can be shortened to "The article mentions that."

While it certainly avoids some conflict, I am beginning to see some benefit to this sort of conflict, if it pushes people to actually read things before they opine, instead of just sharing uninformed opinions.

Are there any other ideas for what could be done about this?


  👤 collyw Accepted Answer ✓
A lot of the time I come for the comments as they are often more interesting and informative than the article. You get a variety of viewpoints (as opposed to usually just one viewpoint from the article) and HN users in general seem fairly knowledgable on most of the subjects posted here. It's a couple of levels above Reddit.

> when in fact the comments are just as uninformed as they are

What's your basis for this?

(As a small practical suggestion towards your goal, don't allow paywalled stuff to be linked, or stuff that involves to much nonsense before being able read it).


👤 gitgud
Sometimes the discussion here is more interesting than the article. If there's something uninformed, then it's totally fine to politely call them out on it.

It can also be interesting to see discussions deviate from the exact topic of the article. I enjoy reading personal anecdotes and related information (with sources).

How could you even enforce reading the article before commenting? People could have skimmed it, misinterpreted it, intentionally disagree with it.

Have you seen the comment sections on the rest of the internet? Youtube, Twitter, Reddit.... I think we're relatively informed here


👤 vincentdm
HN could allow the submitter to optionally configure up to 3 multiple choice questions about the article.

A user would need to answer all of them correctly before being able to comment. An incorrect answer would add a delay of ~1h before they can comment (without needing to do the quiz again).

Even though it’s fairly easy to cheat or work around, it might be enough to tilt the incentives towards just reading or at least skimming the article.


👤 Normille
I think you're fighting a losing battle there.

I always browse HN by the https://news.ycombinator.com/newest link, so I get the latest articles at the top. Every time there's a story of note afoot, it shows up numerous times in the submissions. I've seen the same news story submitted up to a dozen times in the space of a few hours.

If the HN site isn't capable of flagging up the fact that an article someone is about to submit has already been submitted by several other people; and if people are too lazy [or too desperate to earn upvotes] to bother to check whether a story has already been submitted, before submitting it themselves; then I think expecting them to hold back on commenting til they've read an article is extremely optimistic.

PS: Hope this reply is relevant. I couldn't be bothered reading what you wrote, before responding. :-)


👤 muzani
I feel like this might be a good thing.

For example, someone recently brought up that the solution model in the article is unfair, but misunderstood it. I quoted the part of the article where the author put in his definition, and the Wikipedia article on how this problem is fixed. It's a quick conversation that solves a common misconception, but someone has to raise their hand and bring up the misconception.

One problem is that a lot of people do read the article, but misunderstand it. The better and more information dense it is, the more people don't get it, even after reading it. So simply asking people to read it first might not solve the problem. HN is a good place to "raise your hand" on difficult reading material.


👤 miguelmota
I don’t really see the issue on hand as a bad thing because it opens up discussions and all sorts of viewpoints. I see a lot of comments on articles that aren’t really connected but it provides interesting anecdotes and stories. People that are looking for informed comments should simply read the articles themselves and form their own opinions. Everyone has opinions, whether informed or uninformed, and it’ll be a hard road trying to put barriers up but I think that would take away some of the essence of HN.

👤 stephenr
The problem is not that people didn't read the submission.

The problem is that people commented without knowing what the fuck they're talking about.

I just replied to a comment on the HAProxy 2.2 release. I haven't read the article yet, what I commented isn't related to changes in 2.2 - I was able to comment in a meaningful way (i.e. it's an opinion but it's an informed opinion) because of previous experience and knowledge of the topic.

As with many problems, better "voting" of comments would likely solve this.

If HN adopted reasoned voting (i.e. not just down votes, which lead to an echo chamber of whatever koolaid is flavour of the week) you could easily have a "Incorrect" or "Hasn't read TFA" downvote 'reason'.

Combine that with better handling of downvoted comments (i.e. don't just fade them to oblivion, due the aforementioned koolaid circle jerk) and I believe people would use the voting system more effectively.

Right now I simply refuse to downvote any comment, because it's just completely broken (and yet "working as intended" according to the PTB).


👤 kwhitefoot
> The longer the submission, the more uninformed the commentary.

Well that's hardly news; and it applies to pretty much all text both online and offline. If you really want someone to comment on what you think is the main point of a submission write an executive summary and do an Ask HN instead.

Do the summarizing work once instead of expecting ten thousand people to do it.


👤 krapp
Unless people want to go through the effort to read an article, they won't - and it is more effort to read the article than not, so most people will simply either read what they want into the title, or pile on to the first comment that catches their interest.

I don't believe there is technical solution to this, all feasible solutions are likely social, and have diminishing returns over time. Donwvote poor comments, upvote good comments. Correct misinformed commenters as civilly possible, and know when to disengage from toxic threads and people. The culture has to encourage engagement the way that it does civility, or seriousness.


👤 matt_s
Uninformed opinions have likely been more prevalent on the internet since the first comment box appeared.

People that comment without reading the article or maybe even without clicking into it will likely make comments that just don't make sense. I think in engineering circles the corrector/completionist personality trait is more prevalent.

People that did read the article and spark conversation will get more up arrow clicks. Uninformed comments will float to the bottom most of the time on HN.


👤 anotheryou
I'd think for a system to scale you need good comment ranking like reddit has. Not sure if hn is strictly "top" votes or like reddits "best" which takes in to account that younger comments have been seen less and comments with few votes need the benefit of doubt and a boost to get "reviewed".

That should be able to weed out bad comments that are bad because ppl haven't read the thing. (first child comment will call them out, than people will downvote)


👤 latte
Reading the original article takes effort, in part because it may be hosted on a slow-loading site with a bad UI, have intrusive ads, be long, uninteresting and / or written in a language that's difficult to read.

I wish HN had at least snippet previews for submissions (like in Twitter, WhatsApp or Facebook) so that I could look at them without following a link and decide whether to open them or not based on more information than just the title.


👤 ilaksh
I think we are pretty close to having AI that can automatically generated little pop-quizes that verify that people read the article or at least opened them to look up the answers. Not sure it is 100% there but within a few years there should be at least an imperfect implementation.

👤 venmul
The heading gives them biases, though there is moderation after submission. But may be if the moderator or AI could give a small abstraction of the article posted could benefit more on the corresponding thread.

👤 peruvian
Despite the intellectual aura HN puts off, most people come here to blow off steam or waste time. You're not going to get these people to read the article or formulate complex thoughts. At best they'll type a lot.

👤 jhildings
Having a good headline, as opposed to the clickbaits being popular the last 10 years or so

👤 troelsSteegin
It would be interesting to analyze HN comments at scale to see if there are predictable genres or features of comments that are upvoted. Same for down voting, but it seems easy to predict downvotes. What pockets of the HN readership are rewarding what genres of comment? I am sure this is not a fresh idea...

Readership could choose to reward TLDR summaries of submissions. In general, discussion would then be better informed about the content submission.

Personally, I have found interesting discussion to sometimes be not so much about the content of a submission, but about the topic of the submission. A tedious pattern for content-focused stuff is that "what the author didn't do ..." post; a more engaging one is essentially "yes, and here is related or parallel work..." Elaboration is more interesting than negation for this reader.

Up or down voting is a big signal of relevance for the reader but not the only one. Really what I would like is some agent that can read HN submissions and comments for me, extract topics from submissions and interesting comments and links from discussions, and allow me to browse that digest first. Again, not a fresh idea.


👤 hindsightbias
Since most people select the link before commenting to at least see the page, take a timestamp between that selection and comment time for that IP. Based on the length of the article and average reading time, make them wait.

Speed readers can spend more time ruminating, since they are probably thinking too fast also.


👤 bjourne
My highest upvoted comment this month was for an article I didn't read because it was behind a paywall. So yes, I think there is a problem. The karma system rewards you for being on time more than it rewards you for being correct. It also seem to reward snark a lot.

For example, some time ago there was an article about why dynamic linking was worse than static linking along supported with a set of benchmarks. Many people posted highly upvoted, snarky, derisive comments about how the author was wrong and questioned his benchmarks. No one tried to reproduce his benchmark figures (had they tried to they would have found some bugs in them).

I'm confident that if someone were to post an article about why static linking is worse than dynamic linking it would be met with the same kind of snark. So it's not the subject.


👤 nitwit005
Blocking links to paywalled articles might help, although I'm sure people will disagree with that idea (you can detect most paywalls, as they try to play friendly with Google).

You might try to find a way to soften the influence of people who are badly behaved voters, and causing those comments to float to the top. A lot of review sites seem to suffer from the people who give well thought out ratings being drowned out by a majority who only ever give the minimum and maximum rating. That's made some sites give up and only offer thumbs up and thumbs down, but if you can detect those users, you can always try to reduce the influence they have on the final rating.