The problem with using extant votes is that votes are associated with a given post, not associated with a given provider.

The votes in question would be the ones that the first party cast for the second party, thus elevating the level of trust that first party has for the second party.

The curator who voted it up?

Oh, I hadn't considered, but maybe increment the trust the first party has for third parties that also upvoted the second party. But that might be too easily exploited.

That also means that "why am I seeing this thing?" needs to be a more transparent and answerable question.

It could be graphically displayed like those annoying tag clouds... nodes with larger trust values are displayed larger, then color the edges in the graph to show which parties were responsible for it arriving in the users feed.

If you continuously rescale Trust values to be between 0 and 1, normalizing to a float based on a moving window, you can integrate some sort of decay function over time. This also solves the problem of your interest in a given provider changing over time. More recent Trust investments will be larger than older ones.

I like it.

Every system is exploitable.

Sure is.

I suspect it wouldn't be terribly popular, however. (See the current New tab as it is.)

I generally visit that after I've digested my feed. But then it doesn't show me the tags I want to see so then I manually edit the url...