Why Comments Snark

Katja Grace asks:

Commentary on blogs usually comes in two forms: comments there and posts on other blogs. In my experience, comments tend to disagree and to be negative or insulting much more than links from other blogs are. In a rough count of comments and posts taking a definite position on this blog, 25 of 35 comments disagreed, while 1 of 12 posts did, even if you don’t count another 11 posts which link without comment, a seemingly approving act. Why is this?

She suggests:

Commenters are visible only to others in that particular comments section. Nobody else there will be impressed or interested to observe that you read this blogger or story, as they all [do]. So the choice of whether to affiliate doesn’t matter, and all the fun is in showing superiority within that realm. Pointing out that the blogger is wrong shows you are smarter than they.

I don’t see why comments can’t affiliate as easily as posts, but I agree comments often disagree to gain status at the expense of post authors. Constant comments:

One’s correction of error tends typically to be much more throw-away than one’s original thoughts. If you want to correct an error, and if you do not think the correction particularly interesting, you might choose to do it in the comments of the blog that committed the error.

My explanation is related, but darker: Comments disagree more than responding posts because post, but not comment, authors must attract readers.  Post authors expect that reader experiences of a post will influence whether those readers come back for future posts.  In contrast, comment authors less expect reader experience to influence future comment readership; folks read blog posts more because of the post author than who they expect to author comments there.

This induces snarkier comments for two reasons:

  1. Intelligent post authors can usually anticipate the main post “corrections.”  Posts written for readability simply cannot mention every related disclaimer, caveat, alternate interpretation, or follow-on question.  This leaves a huge opening for comments to seem smart by pointing out such things, even when they are boring.
  2. When you post a friendly response to someone else’s post, you can hope for reciprocal posts later, where they respond to one of your posts.  This is less likely when your post is critical, or if you just comment on their post; they may not even know you have a blog.

A similar theory explains why large email lists and usenet groups were often so harsh; each contributor had relatively little influence over the subscriber experience.  This theory also suggests a fix: let blog readers mark comment authors they like, and read all blog post comments via an interface that emphasizes authors they personally like.  Comment authors would then face incentives similar to post authors to please readers.

GD Star Rating
Tagged as: ,
Trackback URL:
  • Lord

    I would suggest a simpler one. “Me, too” or “Uh huh” adds too little to the conversation to be worth bothering with.

    • Robert Koslover

      Lord, you and Robin are both right!! Uh oh, hold on, wait just a minute, there. I guess what I am trying to say is that you are both wrong. After all, the latter claim means more status for me, right? Or? Hmmm, I wonder if one create logically-inconsistent, yet status-generating, argument loops here? p.s. I want credit (aka status) for that last idea if anyone agrees. And remember, if you don’t agree with me, then it is probably only because you are just trying to gain status (you mere peon, you). Heh. Consider yourself insulted in advance. See http://www.youtube.com/watch?v=M5WuHhVRZcY for more, if you can handle it. 🙂

    • Robert Koslover

      I’d just like to apologize for my snarky comment above, but only if doing so raises my status. Thank you. Carry on. 🙂

    • anon

      I find your ideas interesting and would like to subscribe to your newsletter.

      • clayton

        Simpsons quote: status +5

        (I get +3 for recognizing it)

  • “This theory also suggests a fix: let blog readers mark comment authors they like, and read all blog post comments via an interface that emphasizes liked authors. Comment authors would then face incentives similar to post authors to please readers.”

    Yes, well, that would pretty much mean the end of my commenting. On Free Exchange, I seldom did well in the ratings, so to speak. Actually, I only comment on blogs that I like and respect, whether to agree or disagree. I feel that it shows my support for the blog and that I’m trying to help create a community of followers for it. As well, I do sometimes just say that I agree, for what it’s worth. It is also the case that there are a number of blogs I read and respect that I don’t comment on for various reasons.

  • This theory also suggests a fix: let blog readers mark comment authors they like

    Slashdot, for instance, does this — and also rates ratings with a meta-moderation system.

    • Noumenon

      I don’t know why this doesn’t spread. Slashdot with the ratings filter set to 4 or 5 is so interesting!

  • “Snark” is generally thought to be a combination of “snide” and “remark”. Lewis Carrol’s “snark” is generally thought to have different origins. Obviously, Robin Hanson is a complete moron for not realising this. Unless of course, I have just been set-up… 😉

  • Coincidentally enough, Hopefully Anonymous and I were just recently discussing comments vs posts, Katja & snark elsewhere.

    I agree on letting people rate comments. They have a system like that at LessWrong and my personal settings has no cutoff for comment-karma, but it definitely seems like a better idea than comment moderation through a central editor.

    • Surely Robin Hanson is already familiar with lesswrong’s karma comment system. And yet he makes no mention of it in his post, instead describing comment-ratings as though they were a brand new idea that ought to be considered by someone.

      Why the omission, Robin? Why not a post conclusion of, “this has been tried at X, Y, and Z, and through this experience we can make the following conclusions…”?

      • Robin explained [below] that I was confused about his proposal. I withdraw this comment.

  • Psychohistorian

    “This theory also suggests a fix: let blog readers mark comment authors they like, and read all blog post comments via an interface that emphasizes liked authors.”

    This rather strongly argues for a comment format similar to that of lesswrong.com, where comments are ranked by readers and are much, much easier to follow and respond to. Do you have any intention of actually following your own advice and applying something similar to the comments here?

    Also, another theory: people post negative comments on blogs because they are constructive. They make positive posts if they have a significantly new idea or if they don’t have an idea at all.

    If I think you’re wrong for X reason, it makes sense to say, “Well, you’re wrong because of X.” If you think you’re right for X reason, and I agree with you, there’s little use in writing, “You’re right because of X,” because you’ve already written that in a more prominent location than the comments section.

    However, if I think, “You’re right, and that makes me think of Y!” that provides a good grounds for my own blog post. Similarly, if I think you’re right, and I have nothing to add, that makes grounds for a blog post and completely fails to make grounds for a comment – I have nothing to add. Conversely, “You’re wrong and I’m right” makes for a more difficult blog post because I have to set up your position, then knock it down, or I have to assume my readers are familiar with your position. If I’m agreeing with you, I get to set up both of our positions at once, which is rather easier.

  • I’m reminded of hecklers at shows. Similar dynamic. I checked out Youtube where one can rate comments to see how negative comments did compared to positive. My recollection from watching videos there was negative comments tended to get voted down, but just glancing at some videos, I don’t see this dynamic–one video of a kind of nerdy guy talking in an annoying voice has comments that generally are negative and that get enormous positive ratings.

    So just a speculation, but perhaps if the poster is high status the negative comments will get negative reviews, and if the poster is low status, the negative comments will get positive reviews. Similarly, a heckler of a well-loved performer who is playing to a crowd of fans will probably get shouted down, whereas a heckler of some comedian at open mic night might get some laughs himself.

  • Psycho, having comments ranked by an average of other reader evals is not the same as ranking by your own personal evals. When OB moved I was told that no karma system was available to install that didn’t share karma points with other blogs. Let me know if that is no longer true.

    • Oh. You want each blog reader to have their own personal ratings for each comment.

      Most readers may not do enough personal ratings to make this useful for them. But in any case, I think your final paragraph in the post was ambiguous about whether you intended the ratings to be community or personal.

      (Or do you want it to be local to a specific blog, but the whole community within that blog? Is your concern just sharing comment evaluations between multiple blogs that share some of the same commenters?)

    • Don, I reworded that last second to last sentence to be clearer: the ideal scope is the reader R’s ratings of author A, aggregated over all places R read and rated A.

      You really think I shouldn’t write a post on comments that doesn’t explicitly review all known comment ranking systems?

      • Reword: thanks, that’s much clearer (to me!).

        Review of all: of course, in general, that would be silly. I got caught up a bit in my confusion of what you were advocating; I thought you wanted community ratings for comment authors. Given that lesswrong does this, and how closely connected you are to lesswrong, it seemed odd that you wouldn’t mention it, and draw lessons from how it worked out.

        I now understand that you wanted a more personal version of evaluation. Reviews of only somewhat related systems perhaps are less useful then.

      • Constant

        If the rankings are personal, how are they supposed to influence the commenters, who presumably can’t see personal rankings?

      • Constant, post authors don’t know who exactly reads them either.

      • Since I’m commenting I ought to disagree with your description of the ideal scope of ratings 🙂 Unfortunately I’ve trapped myself by proposing much the same idea on hacker news

  • Am I the only one that reads this blog and others as much for the comments as the original posts? (Some others include Volokh Conspiracy, Yglesias, Marginal Revolution, and Female Science Professor).

    Both Robin and Katja suggest that comments are something that simply happen, and simply happen to them. In fact, the best blogs are virtual communities, guided (but not controlled) by the bloggers.

    Bloggers presumably want criticism, for several reasons. Affiliation is one–the community wants a voice, and there are more ways to say ‘you are wrong’ than ‘me too’. Second, criticism sells…and depending on the target audience, snark sells even better. Finally, a critical comment stream increases the total information content of the blog post + comment stream…not too different from a prediction market.

    I would think this would be a case Robin would be playing up the information aspect a bit more, rather than simply affiliation and signaling. I’m just sayin’.

    Oops, was that snark?

    • I don’t see what either of us said to imply comments are unimportant. Can valuable things not have predictable patterns or causes?

      None of your later suggestions appear to explain the difference between other blogs and comment threads.

      • Constant

        My most concise theory of the difference is that comments sections are a discussion and blogs are not.

        In a discussion it is common to raise objections. In fact it is the most common thing. This is as true of discussions over the dinner table, discussions in the classroom, and so on, as it is of comment sections. You have gotten into arguments, surely, over the dinner table, or over lunch. Or at least witnessed them.

        Blog posts are not a discussion. Blogs are islands. The trackback system barely works. Who follows the trackbacks? I think much fewer than read comments.

        Robin says that blog comments need to be fixed and offers a fix by making commenters more accountable, their reputations more vulnerable. I would like to suggest that blogs could be more than they are, they could be more of a discussion than they currently are, if there were something that did what trackback attempts to do but did it better. And if they were more of a discussion than they are now, then you would start to see the same thing that you see in discussions everywhere: disagreement.

      • I didn’t say you ‘t think comments are unimportant, but Robin implies that traffic is driven by the original posts, whereas I think in many blogs it is driven by the quality of commentary. And commentary is more valuable when it includes a lot of disagreement. The same can’t be said for other blogs–that is, the traffic of blog

        A is not strongly influenced by disagreement on blog B, except perhaps in highly partisan political debates (which neither of you typically engage in).

        Also, I doubt your cross-blog agreement stats would hold up in partisan contexts. For example, The National Review’s blog often refers to The New Republic, but almost always to disagree.

    • A number of authors at Volokh regularly make posts with closed comment threads, I presume because they get called on their bad arguments so often. Kudos to Orin Kerr for starting open threads in response.

  • Comments disagree more than responding posts because post, but not comment, authors must attract readers.

    Interesting theory, and one I’d never considered; to me, it seems more likely that comments are hit-and-run affairs that don’t merit high compositional or intellectual investment. If you’re going to take the time and energy to write good blog comments, after a while you’re often better off getting your own blog; I elaborate on (and provide a datum of agreement with) this theory in Trolls, comments, and Slashdot: Thoughts on the response to Avatar.

  • Amazon has a nice comment system: you can choose to ignore any authors in a thread, and create the coherent thread you want to read.

    It is an option well worth installing.

  • Pingback: Tweets that mention Overcoming Bias : Why Comments Snark -- Topsy.com()

  • Unnamed

    Robin Hanson at Overcoming Bias has an interesting post on the reasons why comments on a blog post tend to be more critical of it than posts on other blogs that link to the same blog post. He chalks it up to incentives: commenters don’t need to attract an audience so they’re free to nit-pick and snark away, but other bloggers need to post something that their potential readers will find interesting.

    • Robert Wiblin

      This seems the obvious reason. There’s little point chiming in with a “this post is good/interesting, you should read it” in the comments, because everyone already has read it and you will look like a fool. But doing that on your own blog provides a useful service and will attract more readers to you and make you seem like someone “who consistently has news first” (if it is indeed interesting).

      • ‘You should read it’ is of course silly, but there are many other follow ons from agreeing with something, often used in comments and on other blogs. e.g. implications, further questions, what it reminds you of, additions to the explanation. Positive comments aren’t rare it seems anyway, the asymmetry is in blog links, which seem strikingly positive.

        Blogs could link to quality posts they disagree with and argue, but that would be a different service and more work. That might explain, but I think even if it were little work, a blog that links silently to prominent bloggers seems better than one that demonstrates flaws of obscure authors, even if their points are as interesting.

  • rob


  • nazgulnarsil

    how does agreement add anything to the conversation? when I fully agree with a post i don’t post anything. it’s only when I disagree with some point that I feel compelled to post.

  • I have a Theory of Moral Sentiments take on this. The difference is between how onlookers would respond to the level of negative sentiment when the shouter is a comment author vs. link-post author.

    When onlookers see someone shouting at someone else, they infer that the shouted-at probably did something to provoke the shouter. However, it appears at first that the shouter is over-reacting — the onlooker can’t sympathize with such a high level of negativity because they don’t know the full back-story of the provocation. So the onlookers’ instinct is to sympathize with the shouted-at for receiving an improperly large amount of punishment.

    However, if they do know the full back-story, the onlookers may well sympathize with the shouter — “That’s right, you tell ’em!” Whereas if they’d just walked in when the shouting began, they’d say, “Whoa, simmer down there…”

    Driven by what how much the onlookers will approve of their conduct, the would-be shouters will tend to dial it down when the onlookers could not know the back-story — they don’t want to be unjustly told to simmer down when they’re only giving a proper response of shouting. But they will shout more openly when onlookers do know the back-story — they may even be encouraged by the understanding mob in that situation: “Don’t just sit there and take it!”

    Comment authors know that their onlookers know the back-story, as readers of comments have already read the original post and a fair amount of the comment thread. Link-post authors know that their onlookers will be entirely ignorant of the back-story. If you write a brief snarky link-post, the readers will have to click your link, read the target post and most of the comment thread. And all to see if your hyperventilating was proper or not? Nah, they won’t. They’ll instinctively assume that you need to take your meds.

    The only way around that dilemma is to write a neutral title of the link-post, copy the target post entirely, as well as most of the comment thread, and only after that throw in your snark. Then the onlookers of the link-post will have the full back-story and, if the negative response was proper, will sympathize and cheer the link-post author on.

    • I’m not sure I buy it, but this is the most interesting alternative theory I’ve heard.

  • One point was not discussed: Famous people never comment on other people’s blog posts. They write up their own posts on their own blog.

    I seldom saw celebs commenting on Hanson’s posts.

  • “If the rankings are personal, how are they supposed to influence the commenters, who presumably can’t see personal rankings?”

    Each time a personal vote was cast, it would also go into a list of all votes received by the voted-on commenter. If the votes are binary thumbs-up / thumbs-down, the index would be the fraction of all votes you got that were up (or down), with an error bar to reflect sample size.

    So you would see the answer to the question, “What’s the chance that an arbitrary onlooker thinks you’re praise-worthy (or retarded)?” For shaming purposes, this index would be displayed in place of a cutesy avatar.

  • Grant

    I also read blogs mainly for their comments and the discussion that follows a blog post. I don’t post my full name or any link to my own site, so I’m not doing anything for status reasons. I just see an interesting topic and want to learn more about it (probably for status reasons unrelated to blogging).

    The problem with not reading comments is that the original author is usually very wrong on some point(s) in their post. Typically bloggers cover a vast range of subjects, far too many to become real experts in. So they end up saying some incorrect things (as I believe Robin often does with his ‘status hammer’) and commenters usually correct them. The blogger typically doesn’t recognize this correction; that would be an admission of lower status.

    So I read a particular blogger for the comments that blogger attracts and the subjects that blogger covers. OB and LW attract more rational commenters and tend to discuss things interesting to me. I can’t simply start my own blog because no one would read it; it would be a lot of work to try and gain status in the blogosphere, if I ever was able to at all.

    Am I abnormal?

    • Not abnormal, or at least I also have a strong preference for blogs with good commenting communities.

      However, you’re mistaken if you think being anonymous means that you’re immune to status. If you keep using the same pseudonym in the same communities, it will acquire a reputation.

  • komponisto

    [C]omments often disagree to gain status at the expense of post authors…Posts written for readability simply cannot mention every related disclaimer, caveat, alternate interpretation, or follow-on question. This leaves a huge opening for comments to seem smart by pointing out such things, even when they are boring.

    This is so, so true.

    • Constant

      Doubtless it is true that many commenters, unaware that Robin towers over them intellectually, mistakenly believe that their contributions point out things that he missed. But this is equally-well predicted by supposing that they are lovers of truth who mistakenly think Robin is in need of enlightenment, as by supposing that they are after status.

      A commenter seeking status will not be more inclined than anyone else to offer Robin a “correction” that he plainly does not need. After all, if he were to do so, that would lower his own status.

      We should hardly be surprised. People who seek status seek the approval of others. Others approve of people who do the right thing. Therefore the assumption that people are motivated by status-seeking will predict the same behaviors as the assumption that people are motivated by the desire to do the right thing.

      There are some situations where “that which gains status” and “the right thing” do not match. This is not one of those situations. Those useless commenters that Robin is here complaining about, are also not gaining any status from their efforts. The uselessness of their comments is to be blamed, not on their seeking status versus seeking to do the right thing, but on their underestimation of Robin. Given that underestimation, even someone pure of heart seeking only to do right, would produce the same useless comments.

  • Gregory Wheeler

    Dear Robin,

    The field is new, but there is some literature which suggests that comments on blogs tend to agree with authors rather than disagree. A short discussion of the relevant paper can be found here, which links through to the original paper.


    • Interesting study, but clearly comments at “the top 100 blogs as indexed by Technorati” disagree much less than at Katja’s or my blog.

      • Constant

        You could also reply that the difference may nevertheless remain. Even if blog comments are mostly agreeable, the question was not that, but how agreeable they are relative to links between blogs.

    • Constant

      I mentioned something related to this on Katja’s blog: the political blogs tend to have commenters that mostly agree with the blogger, and since they often link to the blogs of political opponents, those links are, unsurprisingly, hostile.

      However, it could be that certain kinds of blogs nevertheless experience the phenomenon being explained. For example, blogs in which the bloggers advance interesting but dubious theories on a more or less daily basis.

      Your comment, by the way, is a case study to test the theory. It’s a correction, offered in blog comments rather than offered from the commenter’s own blog. So he’s made that choice, the very choice that Katja and Robin seek to explain. Why did he choose to make the correction in the comments?

    • The “yes men” effect.

  • Pingback: uberVU - social comments()

  • Joe

    I agree 100% Robin. If people were more nice, we’d see all sorts of these interesting “me-too!!!!11” comments! Wouldn’t that be fantastic!

  • Jef Allbright

    Some personal observations:

    I generally agree with the substance, but often not the emphasis of your posts. I often feel an urge to say “me too”, but realize that doing so would only reduce signal to noise. So I comment only when I see an opportunity for a significant correction (rarely) or an opportunity for a springboard to bounce the discussion up to (what I see as ) a greater or more coherent context.

    As with attending conferences, I get very little added value from the primary content, usually having seen most of it before. However, I DO get benefit from observing the gaps, biases, motivations, thinking styles and interests of attendees at conferences and commenters on blogs.

    As for the ranking system on, e.g. LessWrong, I find its results interesting in the meta sense described in my last paragraph, but see it as more a popularity contest than a filter of meaningful AND SURPRISING and thus valuable content.

    In my regular work, I implement systems for classification, ranking and extraction of unstructured text, and while there are general approaches to determining what will be valued by most people most of the time (Google can reliably provide the most favored news on Brittany Spears–or is it Lady Gaga now?) identifying information that is both MEANINGFUL AND SURPRISING is a fundamentally much harder problem.

    Snarky and heckling comments characteristically display a signature featuring strong sentiment along with low depth and coherence over context (and often poor spelling and grammar.) A ranking system based on these attributes would be worthwhile (here, as well as in a better-functioning democracy), but presently practical publicly open implementations depend on human input, bringing us back to the current reality of strong status bias.

    In my opinion, the best we can do at this stage of our technological development is to exploit relatively closed frameworks for recommendations based on hierarchical structures of like-minded friends. And then most of the work goes into creating that hierarchical structure. TANSTAAFL.

  • Lord

    Positive and negative are too constraining for comment categories. Is a comment that accepts the post but points out its limitations positive or negative? Is a comment that lays out the post would be right if the assumptions were reasonable but most likely aren’t, negative or positive? Is a comment that broadens an argument but leads to its rejection positive or negative? Is a comment that narrows an arguments applicability but clarifies it negative or positive? Is a comment exposing a flaw negative or positive? Is a comment exposing a different value judgment positive or negative?

  • mjgeddes

    You only need apply the ‘social signaling’ theoy to get the answer – blog comments on the internet are ‘social signals’ by people trying to boost their status by projecting some trait important to their target audience. For instance, transhumanist blogs are mainly filled with wannabe-intellectuals so the goal of comments is to try to sound as clever as all-knowing as possible. But comments are of no use if no one notices, so snark must be added as an attention grabber. All this is done for the sole purpose of getting status in order to get laid.

    • Valkyrie Ice

      *giggle* All too true.

      Very few people are willing to admit how much of what we do every day is done for the sole purpose of attempting to prove we are of superior status to potential mates. Everything, from Corporate level greed to trolling can be, if you are willing to actually break things down far enough, done for no other purpose but to improve our chances of mating and passing on our genes.

      • mjgeddes

        Yes, it occured to me that the obsession with sports actually is strong evidence for Hansonian cynicism.

        Tiger Woods for instance, whacks a small object (ball) with a stick (club) into a hole and gets millions of dollars and woman and praise, all because he can do it better than other people (higher status). The upcoming soccer world cup will see teams running up and down fields in a distant country (Africa) essentially kicking the living daylights out a plastic object (ball). Those who do this better than the others will get all the praise, girls and cash. This event will be focus of global attention with billions of viewers, and special attention in sports media will center on who got praised (scored goals) and who got shamed (issued ‘cards’ – red card for sending off), especially who is best (has higher status). Those who kick the plastic object up the field best will win more money and praise in a few hours of play than a scientist in the lab will make in their lifetime.

  • The problem with comment ratings is that they’re inevitably biased towards how popular the view is among the blog’s readers. This is especially noticeable on politics posts. For example, look at fivethirtyeight’s top rated comments: inevitably pro-Democratic cheerleading. You could argue that it gives people what they’re likely to want to see – but maybe not. There may be a significant gulf between what the regular comment reader and the marginal comment reader wants to read. There may be benefits for the blog’s brand for having a reputation of constructive discussion, but commenters are focused on partisan scuffles.

  • Pingback: Trolls, comments, and Slashdot: Thoughts on the response to Avatar « The Story's Story()