Ꮢ๐ϲoᴄo Ⅿoԁem Ᏼasіlisk is a user on mastodon.social. You can follow them or interact with them if you have an account anywhere in the fediverse. If you don't, you can sign up here.
Ꮢ๐ϲoᴄo Ⅿoԁem Ᏼasіlisk @enkiv2

Bad idea of the day:
A "mastodon recommended follower" back-end that picks at random a recent-poster from the same node that has no network overlap at all (or, if that's not possible, one with as little network overlap as possible).

In other words, rather than optimizing for best match, optimize for worst match & see what happens to social interaction.

· Web · 21 · 31

@enkiv2 I don't think this is a bad idea, but it ideally would be not just on same node

I suggested limiting it to a single node just because nodes often have very different cultures & rules. But, on small nodes this wouldn't work.

Expanding the pool to all already-federated users works too, except on *very* small nodes and *very* large nodes.

*very* small nodes might need to request the local timeline from peers in order to find enough users.

@enkiv2 The different cultures & rules is exactly why this is valuable.
Offline we've always had ability to see meet locals. It's meeting the weird foreigners that the internet really excels at.

Yeah, but at the same time it pays to be careful when the cultures differ too much. Going into somewhere and just assuming that friendly signals are universal or that your needs are the same is a good way to start wars.

People are already pretty isolated in social bubbles *within* large instances. (For instance, if it only worked on mastodon.social we'd still see benefit)

@enkiv2 AKA chats in 2001 :) That being said, discovering totally unrelated people was a cool thing of early internet, and having a bit less of social bubble probably wouldn't hurt. On the other hand, reading instance timeline allows exactly that (for now).

Reading instance timelines really isn't feasible on large instances like mastodon.social as a mechanism for finding people to follow.

For one thing, people who post a lot are going to hide people who post less frequently, particularly if they aren't in the same time zone. (If you have relatively few relatively inactive people you follow, then having no overlap in awake time with some of your followers is fine: they appear in the home TL anyway.)

@enkiv2 I'd hate this on Twitter but if this was an opt-in on Masto with only well-moderated instances I'd actually be interested

Just like on twitter, actually accepting follow recommendations would need to be opt-in.

Oh, yeah. Like . I dig.

Come to think of it, this doesn't need to be a feature. A bot could do this on request.

@enkiv2 but yeah, this could be a fun "I want to follow a new person!" bot

One way to make the bot opt-in is to have the set of suggestions be equal to the set of people following the bot. Then, it could take requests for people to follow, but also (for instance) post messages on fridays.

This would be pretty straightforward but it also means that the pool of potential suggestions is both small and already listed in one place.


In that case it couldn't really be a bot because everyone who knew how to opt in would have at least some social connections in common. Sort of defeats the purpose of such a bot. (Unless someone like gargamel promotes it.)

@enkiv2 I mean if it's done right, one could totally get people to promote it

@enkiv2 I think the average Mastodon user is capable of playing nice better than meeting a stranger on the street. Mind you I'm pretty good at both so it colors my experience.

@enkiv2 @noelle "Challenge your perceptions by forcing yourself outside your echo chamber."

@jeffcliff @0x1C3B00DA @kik @noiob @bea @RussSharek @KitsuneAlicia

I wrote a bot that does this (specifically, follower roulette with its followers).

It's here: @FollowFriday

@enkiv2 people actually get exposed to diverse opinions and get less bubbled up and detached from reality?

@enkiv2 it's easier here than on other social networks. Still the majority or online interactions are quite void of actual diversity of thought as people seem to fear having to be confronted with an actual person with opposing views. Debating the TV talking heads you disagree with is easy, cause they never answer back

I am aware of the discourse. Everybody in my bubble was talking about filter bubbles and their ramifications back when the term "filter bubble" was still un-coined and we called it "the daily me".

I'm interested in automated systems that make opt-in engagement with genuinely different people easier. Often that means avoiding pure randomness in favor of sorting by connection inverted (i.e., find the pair with the least overlap).

Anyhow, in case you missed the rest of the thread, I wrote a bot to handle recommendations.



> optimize for worst match

these four words resonate with me way more than they should with any functional human beings

I think when it comes to anything involving the words social and/or media, "optimize for worst match" is a very important and severely underrated technique and should be part of the stable of rankings *always*.

(I also think that everything that has a set of semi-independent chunks of text should come with a random page feature, though. So, YMMV.)


I mean, as a human essentially "optimized for worst match" to existing in humanity I just can't see much up side in the notion

I mean, "optimize for outliers" or something probably wouldn't be bad, but "let's match up Klan members and black people" just sounds like the stereotypical daytime talk show game of putting actual enemies in contact for maximum mayhem, maybe I'm misunderstanding


although I suppose this is in some hypothetical space where "genuine bad faith" isn't allowed, and then I suppose "worst match" kind of lines up with "outliers", so

honestly I'm probably just not understanding it somehow

Klan members & black people is a different kind of "worst" than I mean. Like, that's predictable based on a single axis.

I'm thinking more like: match up people whose hobby is vandalizing statues via elaborate knitting projects with people who want to teach their children Volapuk as a first language. Mismatch in terms of what they care about instead of in terms of
vehement disagreement about a shared topic.


that's definitely a different kind of "worst" than I had assumed

yeah, THAT kind of "worst match" would be interesting for people to get outside their usual spheres and typical interests

it's just the kind of thing that gives me pause because it's precisely the kind of notion that's been co-opted by the shitbag regime of late. "look, you gotta start thinking outside your little echo chamber. so here's why the compound I live in is ruled by a living god and we worship the Orange Man."

Yeah. I'm trying to figure out what it means to have meaningful prioritization of new ideas, and how to separate that from the shitty "you don't agree with my hateful views because you haven't heard my speech enough times" thing, because the first is really valuable and the second really isn't.

For a first approximation, just having no topics of interest in common might be a start. (And maybe that can be estimated with differences in language use.)