Discovery here IS bad, and — I hate to say it — the solution to that is an (entirely optional) algorithm that surfaces potential follows and compelling content right out of the gate.

Algorithms aren’t inherently evil. Hell, “show posts in reverse chronological order” is in itself an algorithm.

What matters are 1) choice and 2) the intention with which the algorithm was designed.

#Mastodon https://mastodon.social/@mimsical/110865836818850808

I was able to bootstrap myself here relatively easily because I was both part of the deluge of Twitter users in November of 2022 as well as highly-motivated to make it work because I hated what Twitter was becoming and inherently grokked the benefit of federation.

That’s not going to be the common case.

If Mastodon is to expand far beyond low-hanging fruit like me it absolutely needs to meet users halfway.

@jeff would it not be enough to ask new people "type in a few hashtags you're interested in or select from this well-curated list of popular hashtags" and then auto-follow these hashtags for them?

Boom, their timeline gets filled out immediately with stuff they're interested in, no need to have any kind of "recommendation algorithm" involved.

Sidenote: this is less of a problem on smaller instances with lively and relevant local timelines, I feel.

@rysiek @jeff Also a hashtag doesnt make a post interesting. A topic doesnt make an op ed impactful. A genre doesnt make a book well written. Faves and boosts dont either, but if people with similar tastes to me find certain things interesting, I might too, and Id like to see it.

@SarraceniaWilds an algorithm doesn't either, as I'm sure we agree.

I'm not going to die on this hill. But I do feel there is value in the direct human connection, so to speak, created by what I see in my timeline coming either directly from people I follow, or boosted by them.

I can always say: "this is exactly why I got this in my timeline, and this is the person responsible for it, for good or bad."

I find this kind of trivial transparency and humanity refreshing.

@jeff

@SarraceniaWilds of course as long as I can disable or not use any kind of recommendation algo, I can have my trivial transparency and humanity of my timeline, and you can have your recommendations. Which is great.

But the point about power dynamics of such recommendation systems remains.

@jeff

@rysiek @jeff And as for transparency I dont understand why you think an algo cant be open source.

@SarraceniaWilds where did I say an algo "cannot be open-source"? 👀

Algorithmic transparency requires more than just openness of code, though. If the algo, for example, uses any kind of machine learning, apart from the code itself you need information on the datasets used in training, how they are labeled/categorized/etc, and so on.

That's why I wrote "trivial transparency." There is a way to have transparency with recco algos, it's just *more difficult*.

@jeff