I am trying to implement a simple system for trust prediction in social networks. I wonder (since "my system" only performs trust prediction of one user towards another user) if it is common to use that kind of system to predict trust of a user towards some kind of different object of social service.
So, for example, I want to recommend some kind of music video to the user: let's call him Jake. Should I run some simple algorithm - let's say counting how many users that Jake trusts "like" this video - and make my decision whether to display this video on that simple criterion, or should I treat this video as a separate object and perform a "classification process" that will determine a trust value for it?
Which approach is more common?