Eugen RochkoDuring the #Eunomia meeting, I was asked, who are the most influential accounts on Mastodon? I know what Mastodon thinks about popularity contests, so don't get me wrong, I am not going to publish what I find, but I am thinking of methodologies for determining who in this network has influence.
I didn't say "influencer". I didn't talk about Instagram users pushing sponsored products. I didn't talk about users who put "thoughtleader" into their bio as if that would make it true.
Anyway, I can't resist sharing some of my thoughts on this. Ranking by follower counts alone is boring. Admin accounts artificially inflated through the Tom effect.
Running the follow relationships through PageRank makes the results a bit more interesting, but I still feel like that's not it--the results are too influenced by my personal follows, and it's hard to agree that some of my personal friends have influence on the community just because I follow them...
I'd be curious to see how the results would look if you used the number of times a person boosted the person they are following as a weight in PageRank. Maybe that would be it.
Odo Tournesol@Gargron Why do you need to use the data to create hierarchies at all?
witchlectical runeterialist@Gargron i mean .social is going to be a bit stilted by your presence, just as witches.live is stilted by mine. this is why i think mentions are kind of a big deal here, it seems like the most popular people get in conversations or are tagged into things regularly
@Gargron its very different than like, twitter or insta or whatever where number go up, since interaction here is a lot more interpersonal
๐๐๐๐@๐๐๐:~#
Lumb@Gargron
What about if you assigned a weight to faves and boosts and hmmm is there a way to track just replies? I suppose just times people have been mentioned could do that.
But I mean, just the number of times people have had their toots boosted, faved, n replied to? Is that something
What about if you assigned a weight to faves and boosts and hmmm is there a way to track just replies? I suppose just times people have been mentioned could do that.
But I mean, just the number of times people have had their toots boosted, faved, n replied to? Is that something
knzk forever 
@Gargron hmmm. if i aint number 1 i'm suing
healyn@Gargron i appreciate your transparency, and i think the best way to rank users is how often people respond "god damnit" or "god damn" or some other variation to their posts
Trap Beat Spider
Maggie McFear@Gargron Next time just show them pictures of the best cats... which will also be hard to decide. :)
Gadfly (-booq-)@Gargron The question's a bad one, IMO.
It tries to collapses "influential" into one measurement, which is, like, totally BS.
@Gargron
Like is your question "people know their opinions and react to them and that drives conversation?" Then you're pretty high up there. Most people have SOME take, if you take an opinion.
But if your asking after, like, an Ender's Game style "their phrases and opinions percolate through the community" -- with all due respect, I don't think you've done much of that. Whoever came up with the term "hornt", tho..? DesPITE me (and you, and, like, everyone) not knowing their name!
@Gargron Alternatively, maybe you want "they cause people to take certain actions" -- heck, there's a 60+% chance /I/ hit that! Unless someone I didn't see was also constantly mouthing on about it like me, I singlehandedly caused both Pleroma main and at least one Mastodon fork to implement BBCode parsing!
@Gargron
Btw, as an instance admin I would love more tools to be able to look at metrics (messages in and out of the instance, most/least active users, etc.) just because I think it would be neat to be able to track that stuff.
Btw, as an instance admin I would love more tools to be able to look at metrics (messages in and out of the instance, most/least active users, etc.) just because I think it would be neat to be able to track that stuff.
revenant
The Book of KelsPretty sure I won't be on any list you generate.
Ben Lubar (any pronouns)[person who doesn't know what they're talking about voice]
Who do you think the most influential person on GMail is?
paramilitary organizer
Strawberry Toob@Gargron Wouldn't it just be the most well federated account?
๐@Gargron Lance Ulanoff
Neal Wilson โ
Lord Guysbert de Fieri 
@Gargron how about by doing a sociology study instead of trying to create a technological algorithm to mimic human assessment and considered study
infinite love โดณ@InspectorCaracal @Gargron what would the methodology of such a study be, though? methodology doesn't have to be technologic, but it does have to be consistently qualitative or quantitative.
@trwnh @InspectorCaracal Conducting a study is an algorithm too. Now, I am not a sociologist (I minored in philosophy, not sociology), but it seems to me like asking people head on who they are most influenced by is not a good way to find out who they are most influenced by. You would maybe find out who they think they are influenced by. And at a much smaller sample size...
@Gargron @trwnh I'm not a sociologist either but imo it would take an extended period of observation by sufficient people doing the study to identify all of the circles of influence by behavioral patterns and passage of ideas. And the people doing the study would have to remain uninvolved in the dissemination of ideas and to be up-front and aware of their own potential biases.
@Gargron @InspectorCaracal hmm, interesting point. you could measure how fast memes spread. but maybe not so easily because there's no way to track who you boosted something from, as in tumblr. tumblr had a reblog tree extension/lab that was pretty interesting at identifying influential nodes
James Webpage@Gargron @trwnh @InspectorCaracal one time @[email protected] set everyone on fire you could do that and watch where it spreads to.
Sydney โThe Wordโ Falk 
please let me know if I have any when you check, b/c I'm pretty damn sure I won't be on the list, but if I am I should get off it somehow
@sydneyfalk What is the purpose of this message?
it's a bad joke
(b/c I know I'm not someone "with influence", and because I would probably try to destroy said influence if I was aware I had some)
The Big Mammoth@Gargron why is this important? is there any good you can do with such a list?
Drew Brees@Gargron I can only assume "Mastodon's Sweetheart and top booster, Anthony Davis" was among your answers, and your gratitude is more than enough of a thank you.
@anthonydavis Honestly I wanted to toot about the methodologies for determining what influence is and how to measure it, but all the immature responses make me want to keep it to myself.
Frinkel, but is a coin@Gargron does "constantly shitposting about TF and clone TF" count as being influential because if so i'm extremely influential 
@Gargron
classical "engagement" approach:
- boosts
- faves
- followers
- reply rate
aggregated in total, for highest-scoring toot, and also normalized to toot count, or normalized to number of followers. also,
- endorsements
downside is these measures can easily be gamed, but within a margin of error it's the most numerical you can get. works fine if you assume "influence" = "engagement". if not, you'd need to have another proxy for "influence" (which is unmeasurable directly)
Vx. Princess "size_t queen" Grace 
@trwnh I don't think gaming any of the stats is a concern when talking about private, one-time research.
@Gargron perhaps not in context of the study, but it's a variable that has to be considered nonetheless. thankfully mastodon de-emphasizes engagement farming, but it's not immune. surely there exists at least one account trying to maximize its engagement (not necessarily in negative emotional contexts, as twitter is prone to).
@trwnh Yes but you don't need to think about the research itself feedback-looping into the metrics, like it would do with live leaderboards.
@Gargron I minored in math at school, and I can tell you that there are a few ironclad rules for determining who has the most influence in a given network.
- name rhymes with "space"
- exactly 846 followers
- made chili recently
The numbers don't lie, Eugen.
Jason Scheirer@Gargron just be lazy and apply PageRank via followers/follows on the network
@Gargron Eugen, i have a sincere question here. What purpose does it serve determining that?
@Gargron thanks, gotcha
Emelie ๐จ@Gargron obviously the accounts that posts the most cat pictures
ersatzmaus@Gargron a count of boosts or favourites by accounts that _don't_ follow the candidate directly, maybe?
Bert Latamore@Gargron Why should we care? On Facebook advertisers try to determine who the influencers are to try to get those people to recommend the advertiser's clients' products. Here we don't have to put up with that sort of stuff.
@BertL Who said that you should care?
alienne rottenmeier@Gargron
its obviously cats, ๐ ฑ๏ธ and ๐ lately
its obviously cats, ๐ ฑ๏ธ and ๐ lately