You Can Predict What a Platform Will Become From What It Counts

If you want to know what a platform will feel like in three years, do not read its guidelines. Look at what number it displays next to a piece of content, and how that number is produced. Everything else follows, and it follows quite mechanically.

I ran a forum through two changes to this and watched the culture reorganise around each one within about two months.

The first version showed a raw reply count. What that rewards is anything that provokes a response, which in practice means disagreement, questions that are slightly wrong, and mild controversy. The place was argumentative and quite lively and had a persistent problem with people posting confidently incorrect things, which I initially read as a personality problem in those individuals. It was not. Being wrong in an interesting way was the highest-scoring available behaviour.

The second version showed a score from votes. That rewards agreement with the median reader, which is a completely different thing. Within a couple of months the arguments had thinned out and the top of the front page was consistently pleasant, uncontroversial and slightly boring. We also acquired, for the first time, a real problem with unpopular-but-correct posts disappearing, and with a small number of members who had learned exactly what the median wanted and produced it reliably.

Neither was better. They were different failure modes, and each was fully predictable from the counter.

The general form of the rule: whatever is counted becomes the goal, and the goal shapes not just what is posted but who stays. That second part is the slow one and the one nobody watches. Under the reply-count regime, the people who were good at provocation stayed and the people who wanted a considered discussion drifted; under votes it reversed. Two years of a counter does not just change behaviour, it changes the population, and by then it is not reversible by changing the counter back — we tried.

Applied outward, this is why I find most commentary about platform decline unsatisfying. It usually locates the cause in leadership, or advertising, or a category of user. Those things matter, and they are downstream of a more boring question: what does the interface count, and what does the ranking do with it. A platform that counts time spent will produce content that extends sessions. One that counts shares will produce content that is shareable, which is not the same as good and is very specifically not the same as true.

The one thing I would tell anyone building a small community: you are choosing this whether or not you think about it. A default forum installation has a counter in it, chosen by somebody else for reasons that had nothing to do with you, and it will shape your place more than any rule you write. Ours did for two years before I understood that I had made a decision by not making one.