Karma, Policed
KMO published a Substack article yesterday, which I wanted to comment on.
The article deals with the decline in the quality of online discussions on any topic. So... the thrust of his argument is that the big online platforms that have been kind of accumulating more and more, a bigger and bigger percentage of our discussions, have been moving to implement various algorithms designed to optimize specific metrics. And the metrics had to do with reactions that people give in response to various posts. These reactions can come in different forms. For example, the most primitive form is the ubiquitous like button, but there are other feedback mechanisms. For example, Reddit utilizes so-called karma, which the way I understand it boils down to like a sum total of upvotes and downvotes on your posts. Now, that may seem like a logical way to create like auto-regulation for each board or subreddit, but over time what it does, it actually constructs a very effective echo chamber. And the meaning of those upvotes and downvotes becomes simply upvote means I agree, downvotes means I disagree. People who I disagree with will be voted out of this board because their posts will become invisible. And so this group devolves into an echo chamber. KMO described it much more eloquently than I just did. And I really recommend that you go and read the whole article for yourself. But I wanted to make two comments.
First, if anyone still remembers a big online empire that was called Gawker that went bankrupt as a result of a somewhat frivolous lawsuit, was it 10 years ago or maybe it's not quite, maybe eight years ago, that empire was presided over by a publisher called Nick Denton. Now, Nick is a somewhat controversial figure, but it's completely irrelevant to what I'm going to say. Now, one thing to notice on Gawker properties, comments were kind of important. They were actually the main reason people often read this website. So Denton was very concerned about maintaining a certain level of vigor and insightfulness within those comments. And I think it was somewhere around 2011 when he went to the South by Southwest conference and gave an extended talk on the subject of online comments. And I think during that talk, he made some excellent points. One of them was that he was very aware of the fact that if you let people vote for the popularity of each comment or poster, which would then result in comments that represented the majority opinion rising to the top. And for Nick, with his muckraking instincts, that was the exact wrong thing to do. Now, the second point he was making, which was kind of similar, is if you keep doing it, you keep rewarding people expressing that prevailing opinion, then you're actually attracting the exact wrong type of commenter. It's the commenter who has absolutely nothing to add, nothing interesting, nothing insightful, nothing new. And if that person is attracted by having a high karma, high number of points, stars, upvotes, or whatever, that actually makes it only worse. So in summary, this is the person you definitely don't want to give too much visibility to. On the other hand, if a story has a villain, and that villain shows up in the discussion board, you should definitely give that person number one spot, regardless of how popular or unpopular they are. Of course, they're going to be unpopular. This is the villain.
So I think Denton's talk from more than 10 years ago and KMO's article from yesterday, they both make the same general point, which is that the comments whose visibility is based on this kind of social signal, upvote, or karma, or things of that nature, they're going to be fundamentally broken. So it's a wrong thing to optimize for. And it results in poor, more echo chamber discussions.
Now, I don't want this to happen to this platform. And of course, there are major differences between Gromco and Reddit, for example. But still, as we build our tool set for moderation, and for discovery, we're going to try to do our best not to fall into the same traps that all the major platforms seem to have fallen into, as argued masterfully by KMO in his article yesterday on Substack, which you should probably go and read yourself, because I probably didn't do justice discussing it here. But once you do that, feel free to come back and leave a voice comment with your opinion on this whole issue. Because that's just good karma.