quantized pfp
quantized
@quantized
"suggested followers" ruin all new social networks by creating an inequitable/lopsided socio-economy that stagnates the culture around the early nerds / founders' friends
6 replies
11 recasts
66 reactions
Dan Romero pfp
Dan Romero
@dwr.eth
What’s a better model?
5 replies
4 recasts
33 reactions
Venkatesh Rao ☀️ pfp
Venkatesh Rao ☀️
@vgr
Sortition above a minimum quality low threshold Give everybody 50 randomly selected follows from a large group of 2000 active casters A bit like data availability sampling at human level Maybe give them a button to regenerate the set Show them a word cloud of casting topics in the set
1 reply
0 recast
1 reaction
Dan Romero pfp
Dan Romero
@dwr.eth
> Give everybody 50 randomly selected follows from a large group of 2000 active casters More or less what our onboarding model was with the auto follow UX last year. We stopped it because people complained but retention was best when we did it.
1 reply
0 recast
3 reactions
depressivehacks pfp
depressivehacks
@depressivehacks
Personally, I feel that it should be topic based. This would require being able to analyze the content that people are posting on a large scale, but folks are more likely to stick around if their timeline is filled with topics that interest them. I've been expanding DepressiveHacks across socials since October, and this is an issue I've run into on many platforms. Quite frankly, the original algorithm, when you've provided no user activity, is bad on many platforms, and the amount of input that it takes to create a feed that you're actually interested in is too high. This creates a time burden on users, and unless they're extremely committed to growing on that platform and want to put in countless hours to train their algorithm, they're likely to default to going back to spending their time wherever they had been previously. A timeline that is of interest from the start would make those initial 10,000 interactions much more authentic.
0 reply
0 recast
0 reaction