Friday, January 13, 2012

The Disqussion

Yesterday Fred blogged about Disqus's report on its user base, with the title "Pseudonyms Drive Community". Part of the definition of "quality" were number of replies and number of likes.

Now, Likes are variable but don't reflect that; intent differs from person to person. I like a lot of things, but I like some things much more than others. Disqus gives us no way to differentiate that.

An example is that I tend to like @awaldstein 's first comments; I appreciate them, I learn from them, and I want to point that out to others.

In later comments down the thread, I might like a joke he makes. I appreciate them less, I don't really learn from them, and when I Like them it's just to pat him on the back.

So how can such a variable signal be given such weight in their analysis?

And quantity as an indicator of quality, well, that really bugs me. Some of the best comments require no response. They're great on their own.

Some very long threads have tons of replies that are completely meaningless to the original point; some original points are weak on content but strong on provocation.

I dropped in late to the party--4 hours after the original post, and posted this.
Number of times a comment is replied to? Really? That's a signal of quality? 


Seems to me it's a signal of engagement. 
Or in some cases, enragement. 
Not simply what the sage meant. (drop me a beat...)


The totality of the statement is a bit over the top. Some, not all good comments are from pseudonyms, just like within that group some, not all comments are useful, reliable, or experience-based. 


Quantity is not a signal for quality. As much as I like Disqus, this 'research' doesn't reflect the quality of the service.
My comment was voted up pretty quickly, became the top comment and stayed there (yay me!).

Should velocity be a signal? Maybe. But velocity doesn't speak to the quality of the content (though by default everything I say is amazing high-quality content, of course;) )

Likes and number of replies don't do the job; they're inadequate signals. A better approach would be to have yet another button (or slider, or stars) that lets you vote on quality.

On StackOverflow, the best answers are voted up. Wait! Are they really the best? With programming there are tons of opinions on the best methodology, but fundamentally the suggested approach has to work, and if 50 people vote for something, it's likely the approach works.

But I think that's weak as well, though it's stronger than what D presented. I can find good solutions through that, though sometimes the better responses have far fewer votes than the most popular ones.

As @messsel pointed out later in @fredwilson 's post, sentiment, quality, etc are tough nuts to crack.

Which is why, I think, people reacted so badly to the claims in @danielha's original post: people care about quality, and we see Disqus as a thought leader (with about a million blogs using Disqus it's clearly a force), and they've based quality on weak signals.

Disqus is one of my favorite companies. We complain because we like what they are doing, but want them to be ideal. And my guess is that many of us differ on what an ideal Disqus should be. As it should be, of course.

I'd like to challenge Disqus to a live video chat among people who care about this. What a great discussion it could be. In fact, it would have been great to have a live video chat sometime yesterday when the topic was hot...when's that feature coming, Daniel?

No comments:

Post a Comment