your score:
Discussion
it seems to be a good metric of compatibility since it tends to capture a lot of the data relating to which people are interacting with each other
I want to explore the question more of how the score relates to clustering. but for now, I have to go to fiat mine
i've been waiting for someone to build a thing that analyses the relative frequencies of engagement in the form of reactions (and the emoji in it, if any) and replies, to composit against the follow list graphs. i think that would change the results quite a bit. engagement isn't necessarily trust or approval tho, to really dig into that, reactions would need to be analysed relative to the user and who they make them against, as well as using LLMs to pick out the tone of replies
I've been saying that for a long time too.
It would probably improve my results. Not sure.