pull down to refresh
79 sats \ 3 replies \ @k00b OP 18 May 2022 \ parent \ on: SN release: profile photos and a few other things meta
Not a whole lot to say yet, but we currently have a single trust metric for users based on post/comment quality. However, in real life we don't trust people overall if we only trust them in one category of action. e.g. If I trust your ability to drive, I don't trust your ability to perform heart surgery. Meaning, we will likely develop several trust metrics for users:
- do we trust their upvotes (are they voting up quality stuff)?
- do we trust their posts (are they posting good stuff)?
- do we trust their comments (are they adding to the discussion)?
Overtime many more dimensions of trust will likely emerge as we add new features.
That sounds pretty sophisticated. I like the idea a lot.
But for the sake of compute performance this would need some serious optimization or keep an extensive caching logic that keeps a users different trust scores and dynamically updates exactly those scores that are the oldest or have changed a lot.
This is just a suggestion from me, you could do it entirely different of course. Anyway, it's pretty fun to think about an architecture like this - will it slash fast if it looks like a users account was compromised? Will interaction with people with high trust scores ramp up the trust scores fast? So. Many. Things. To. Consider.
reply
for the sake of compute performance this would need some serious optimization or keep an extensive caching logic that keeps a users different trust scores and dynamically updates exactly those scores that are the oldest or have changed a lot.
Not necessary to optimize aggressively today. Current trust algo (which we'd reuse for all these metrics) is fast and only gets computed daily.
Will interaction with people with high trust scores ramp up the trust scores fast?
This is how it works today.
reply
Alright, then I will exchange more comments with @k00b lol
reply