The economics of comments
[Disclaimer: this post doesn't attempt to build grand theories about web economics; it's just a point I need to make before writing about the strategies of different search engines concerning user input - hopefully next week]
In an ideal world, every piece of content is just the beginning of a conversation. Umair brilliantly explains how the value lies in the conversation how content providers are just organizing their readers’ discussion.
True. So true.
But you cannot but notice that sometimes, comments are nothing more than noise. Take YouTube. Comments are so bad that you even have a Firefox addon to fix their spelling and punctuation.
Even on news articles, comments can be detrimental if they’re aggressive or plainly moronic. The comment system of the French Libération.fr, for instance, leads to a mediocre and pointless conversation, as Julien Dorra showed (in French).
Some other websites are famous for the abysmal level of their comments. Among the worse comments sections, I’d choose Haaretz.com, USA Today, the Daily Mail and about any other website thinking that the more comments, the better. C’mon, it increases the page impressions, so it must be good!
Marginal value of a comment
The reality is that the more commenters you have, the more likely it is that one of them is a troll. This is just another formulation of Godwin’s law, which states that as a Usenet discussion grows longer, the probability that a comparison with Hitler be drown comes closer to 1.
That’s why I drew the blue curve of the marginal value of a single comment. It decreases as an inverse function of the number of commenters, itself a function of the size of the audience.
Hence the red line: as the audience grows in size, the total value of comments increases more slowly.
Marginal cost of a comment
Now for moderation. I’m assuming that the cost of moderating a single comment remains constant, so that the total cost of moderation increases linearly.
Just look at the curve. At some point, it costs more to moderate comments that to get rid of them. Web managers can push the intersection of the red and yellow line further right, using tricks like a reputation system, a voting procedure, pre-moderation etc.
My point is simply that a larger audience automatically leads to a conversation of lesser value, relative to the number of participants.
Ok, the curves are a bit pedantic, so let’s take an example. If you tell a story to 5 of your friends in your living room, you’ll all have a nice chat until the subject of the conversation changes.
Now, try to tell the same story to 5,000 persons in a shopping mall. Chances are that people will want to react, but you won’t have the time and possibility to chat with them all. The resulting free-for-all conversation is likely to be confusing – and I’m not even speaking of that crazy, half-drunk guy who started talking about his time serving in Afghanistan.
You can take steps to push the point non-conversation further right. For instance, a conference can bring 5,000 people together and keep the discussion under control – but that increases organization costs. Moreover, the value of a conference lies in great part in the informal discussions you have in corridors or bars, as opposed to the plenary sessions – thus fitting into my model 
hi, thanks for the link to the article on the roots of the problem with Liberation comments.
Just to mention that I am in fact the Author — not Xavier !!
In fact the Observatoire des Medias post is a re-post / cross-post — you’ll find the original article here :
http://ils.sont.la/post/commentaires-mediocres-et-sans-interets-5-erreurs-liberation-a-ne-pas-commettre
Sorry Julien!
Link and name are now rectified.
Very accurate analysis, it deserved a comment
[...] takk til Nicholas og Window on the media som lot meg oppdage den geniale [...]
[...] interesting theory on “Windows on the Media” about whether story comments, after some po… My point is simply that a larger audience automatically leads to a conversation of lesser value, [...]
I think your argument (and graph) are basically on target — *if* a site does nothing more than enable an open-ended comment box with articles.
I would argue that as traffic increases, so does the need for intelligent application of technology. The goal is to ensure that the cost of moderation does not scale linearly. Measures that might be considered include:
* tools that alert a site manager to heated discussions that break out (perhaps by recognizing emotional language or a flurry of posts in a short period of time)
* tools that give users control over what comments appear (for instance through ratings and moderation)
* tools that impose structure on the comments.
On the last point, I am currently directing a class of master’s students at the Medill School at Northwestern University … who are developing some new approaches to commenting on news articles. You can read more about what they’re up to at:
http://www.pbs.org/idealab/author/rich_gordon/
and the students’ Web site:
http://www.crunchberry.org
Rich,
Thanks for your message.
Actually, my point was that the return on comments is always diminishing and that the organizer of the conversation can only find ways to decrease the rate at which it’s diminishing. But then I’d need some hard data to prove it
Automated comment moderation looks promising. A researcher on language technologies (who happens to be my dad) told me that some labs produced software able to assess the quality of a comment. They compute the position of commas, the number of words etc. to tell which comments are best written. Not the panacea, but an idea I’d love to see implemented by content companies.
[...] . WindowsOnTheMedia : L’é;conomie des commentaires (angl.) Dans un monde idéal chaque morceau de contenu participe à une conversation, mais bien souvent les commentaires ne sont rien d’autre que du bruit, d’autant plus que bien des systèmes de commentaires sont loin de les mettre en valeur. Le problème de la gestion des commentaire explique Nicolas Kayser-Bril, étudiant en économie des médias, c’est que plus votre audience est forte et plus la valeur des commentaires décroit alors que le coût de modération augmente. [...]
[...] Window on Media When comments trigger… more [...]
[...] The economics of comments | Window on the Media "The value lies in the conversation how content providers are just organizing their readers’ discussion. True. So true. But you cannot but notice that sometimes, comments are nothing more than noise. Take YouTube… My point is simply that a larger audience automatically leads to a conversation of lesser value, relative to the number of participants." (via Robin Hamman) (tags: content ugc networks conversation community) [...]
Hello Nicolas,
See below some comments.
“My point was not to say that collaboration is worthless”
> OK. This what we understood when published in vinicius torre freire’s blog.
“(..) more people interact, the more likely it… is going to be corrupt by ‘crazy’ elements (..)”
>You should have efficient filters, team of moderators or make your network to filter itself.
>At same time, the probability to find a breakthrough argument or idea also increases.
>You can not deny or neglect this potential only looking the bad side
“(..)more people in the conversation, the less qualitative the average input of each user”
>This point is not clear for me. You should develop better to consider: 1) marginal contribution may not fall too much 2) cost of moderation may fall if you use more effient tools and process 3) the total contribution curve may not be a parabolic curve with vertex in (0,0), it may be a curve with maximum or minimum in some point, my doubut is about the adopted shape of the curve..despite it has coherent trend, it may not be the only solution or curve possible 4) breakthrough arguments or briliant ideas may appear in a large sample…
Regards, Docontra (”againstall”)
docontra,
Thanks for posting here, and in English!
I agree totally with what you’re saying. But all I wrote to this point concerning comments was all things equal (ceteris paribus as we use it in econ).
The conclusion is that as the number of participants increases, you need to invent new ways of moderating their inputs.
Post moderation, as in a blog, works badly on a large scale. Just look at the BBC ‘have your say’ forums: is the time spent reading everything worth it?
The flagging system may work on a small scale (though I doubt it) but when you have millions of comments on thousand of pages, the probability of having foul language remaining unflagged increases. Imagine a nazi rant on one of your brand’s page (I saw that on an AOL website, couple of years back).
But it doesn’t mean you can’t have a large website running on UGC. Wikipedia does that perfectly well. Then again, I’m sure you’d agree their moderation system is way more subtle than those of other content website (again, I’m thinking of Google).
The example of h.ua (read about it here http://windowonthemedia.com/2008/11/ukraines-highway-citizen-journalism/) also shows that clever systems of community-run content management are efficient.
I just argue that such systems also have a turning point, when they are overwhelmed with contributions. But I’d need to do further research to show that, I know.
OK. I got your point.
But partially agree because you neglect completely the chance to find a great idea, a high level dialectic contender, a positive thinker, i.e., a good contribution.
The other part is that a Shopping Mall, as cited in your example, is different than a blog about media, politics, soccer or even newspaper.
In the case of Brazilian blogger Vinicius, one of his posts were so violent and rough that he multiplied the chances of finding trolls. Even so, he used your model to demonstrate that internet is plenty of violence and is useless.
I told him that every blog that deals with soccer/sports, church/religion and politics will suffer much more attacks than any other.
Hi again, and sorry for the late reply.
When it comes to great ideas popping up in a user-generated area (ie, comment section, forum etc.), I don’t think – as you seem to assume – that the probability of its occurring is constant for each posting.
Let me explain. What does the great commenter want? To have an impact. (Let’s be honest: if you spend time conceiving and formulating a great idea, you want it to be read).
If the great commenter knows that a particular comment section is likely to be trolled – and therefore not read by people looking for quality debate – he’s likely not to post.
In even more clear terms: Would you spend 30 minutes writing a comment under a Youtube video? A Ha’aretz article? A Google SearchWiki?
mmm… probably not. But you do spend time arguing with bloggers, because people who read and answer to comments on a blog are likely to be interested in what you have to say.
[...] wrote a while ago that the marginal value of a user comment declines, ceteris paribus, i.e. if the commenting system remains [...]
[...] agonize over this fact, wonder what it is I’m doing wrong — or I could take heart from this research: My point is simply that a larger audience automatically leads to a conversation of lesser value, [...]
[...] The economics of comments | Window on the Media – My point is simply that a larger audience automatically leads to a conversation of lesser value, relative to the number of participants. [...]
[...] to put comment moderation on over the holidays or just let anarchy prevail. This graph from this article on the economics of blog comments is thought provoking: Iain Dale has had enough of window-lickers [...]
[...] de que um estudante francês, Nicolas Kayser-Bril, foi muito lúcido ao cunhar o texto “A economia dos comentários“. Neste caso me refiro às reações desproporcionais a postagem de Fernando Rodrigues sobre [...]
[...] I was delighted when I came across a very well-thought-out post on the blog of Nicolas Kayser-Bril, a media economist. Given the publication I write for, it should [...]
[...] temps, vous savez que je me suis penché sur le problème des commentaires plusieurs fois (voir The Economics of Comments, par [...]