Interesting response!

Yes, the node, edges, travelling salesman, those are my definitions as well.

The only difference is that I’ve always thought about this is with an undirected graph. That models conventional friendships pretty well, but now that I’m thinking about it, it’s probably not a good way to model modern “relationships”. Either way, with two nodes it’s no longer a travelling salesman, as it becomes a much easier problem to solve.

I would argue that the tyranny of the majority unlikely to not be solved with the current form of social networks. I’m not sure if I have time to write out my argument, but it stems from the fact that the most popular people in the world have very uni-directional relationships with most people; everyone knows of Mr. Beast, very few know Mr. Beast.

Ooof. Hot take !

Hot take, but I think the Model Y/Late Model 3 were really well engineered.

I would have thought this is already a solved problem.

If you model social networks like a graph, then you can measure certain properties. One property that’s very important for social networks is the “small world index”. The small world index is a ratio of (how many of your friends know about each other(clustering coefficient))/(how many people on average to connect with anyone in the world (average path length)). Basically, in tribal communities, it used to be that everyone knew one another, AND if you really needed to send a message to another community then it would take only a few intermediate people. The former gives you a sense of safety, and the latter gives you the sense of being able to change the world.

With the advent of social media and other things, the small world index has gone way down. The amount of your friends that your friends know on average is gone down, aka, everyone is very fragmented and the clustering index has gone down. This number has gone down faster than the average path length, because the average path length was surprisingly low to begin with. The net effect is that people feel less like they are part of a community.

Social networks always try to tune their algorithms to encourage interactions to raise the small world index in this direction, but it’s very hard. For one, only few percent of users actually generate any content, the rest are lurkers. You can’t have a high clustering if your “potential friends” never talk to each other, if you can call them that.

Another reason is that enabling small worlds inside social networks is sometimes at odds with revenue generation. In general, consuming content/having para-social interactions online is much easier than before the modern social networks. Thus, people have gotten much more picky with their time and interactions, and now the type of content that people expect to see online now must be of the highest quality. High quality content is easier to monetize, but it’s also harder to create. So this puts an artificial floor on who creates content (only people who are in it for the money), and thus we have fewer people who spread/share ideas. This decreases the clustering coefficient. Thus, paid social networks are against small worlds.

Follow me for more crazy ramblings.