Reccomendation Analytics
Ok so this is just something I'm v excited about as a data nerd ok
So assuming we have a linear gradient of relationships, and arbitrary point values assigned to them
- blocked -5
- unknown -1
- following +1
- trusted +5
And we have two users who already have a network. There's analytics stuff twitter does to reccomend who to follow, but it occurs to me that what we could do is assign "arbitrary" point values to the relationship between two users, done like so:
- user A is looking at user B
- we take the average of of how the people user A follows and trusts feel about user B, with trusted connections weighing 5x as much
- this score is displayed to user A as a "affinity" of some sort
- theoretical maximum affinity is 5*5 (you only have trusted connections and they all trust that person)
- theoretical min is 5*-5
It also occurs to me that this could maybe in some way be used to map social capital? Idk, would probably have to opt in for that, since creedy programmers trying to steal all the data ya know.
At any rate, this is a practical system, it could be used to determine suggested follows, which is a feature I really appreciate ... When it works properly. It would also be -completely- transparent and potentially editable by the user. Maybe even add in some advanced stuff like using it to generate network graphs or weighting affinity based on keywords. That'd be a premium feature though probably.
But again! I would like to state that this already exists in every social network, its just completely opaque.