pull down to refresh

money is the (mother fucking) moderator™️

What doesn’t really make sense is a post getting 1,000 up sats from 10 people, and then one person with 333 down sats being able to cancel it out. That needs to be rethought.

reply
135 sats \ 6 replies \ @k00b 5h

I agree.

Optimism, after Scoresby did, made the point that in tuning for sybil risk we underestimated "activism" risk.

I'm thinking we should make upsats = downsats = boostsats in terms of weighting. Game theory tells us what people should do when they are rational and energetic. It's a great starting point for system design, but alone it tends to produce systems that humans don't like, because on average we are irrational and lazy.

reply

I don't know how, but I think the number of up/down zaps (stackers) should be taken into account.

reply
22 sats \ 4 replies \ @k00b 5h

On that point I disagree because counting people, determining with 100% certainty that one zap = one human, is impossible today and will be more impossible tomorrow. Trust/reputation can be used to make better guesses about who is human, but it makes the system opaque and unfair. As is, it may be unfair, but it's transparent.

reply

I wonder if the answer isn't user configurable weights. The raw data is in the zaps table, but users can decide their own methods of aggregation and ranking.

Obviously, that's information overload for most people. But maybe some human-friendly names for a handful of baseline configurations (Wild West Mode, Trust Mode, Default, etc)

reply
169 sats \ 2 replies \ @k00b 4h

Beyond information overload and engineering overload, it'd fracture everyone's experience on SN.

Under high variation regimes, the things I say about my experience aren't likely true of yours. e.g. this discussion we are having would be much harder to have if there were even a handful of modes of SN.

There's something special about us all playing the same game at the same time.

reply

Yeah, it wasn't really a serious suggestion, as I think it's a "juice not worth the squeeze" feature, by any sane metric.

But I like thinking about the limits of possibility for these game theoretic systems of trust and attention.

The tradeoff between trust, transparency, accessibility to new users, robustness to bots/spam, and overall utility is a difficult balance to strike.

reply
102 sats \ 0 replies \ @k00b 4h

By fast food analogy: I like in-n-out because it serves hamburgers with little variation and has a relatively limited menu. It makes great hamburgers, and at scale, because it makes them with little variation and has a relatively limited menu.

I'm much more interested in building, and using, an In-n-out than a McDonalds.

reply

This really is a microcosm of the attention economy. This wouldn't matter so much if Hot wasn't basically the prime attention allocator.

Even if I set my visibility filter to -1,000, downzapped posts are still effectively invisible to me due to my inability to maintain perfect attention. I'd have to catch them on Recent about the time they're posted. (My other option is to find a comment worth clicking on when I do Recent -> Comments which then helps me find the post.)

reply

Hot and top are broken, recent is the way to go. That’s what I’ve been doing ever since I joined SN.

reply
100 sats \ 0 replies \ @Scoresby 20h
the prime attention allocator

I had been thinking about this mostly as a problem of how zaps and downzaps work, but it is interesting to wonder about solutions that give stackers different ways to decide how to allocate their attention.

reply
0 sats \ 0 replies \ @k00b 20h

master branch has a -inf filter minimum. I'll deploy it tomorrow

reply