Going from "certain features that cause virality are not implemented here" to "hence things cannot go viral here" not two weeks after #JohnMastodon went certifiably viral is… a take.
Going a step further and claiming this somehow means fedi could not have supported social movements even more of a jump.
One way I could respond to that is: this whole network is a social movement, for Dog's sake! It started off as a social movement of people who wanted out of walled gardens.
But…
1/🧵
…I think there's a more important point here that is missed.
I don't think such "virality-enhancing" features generate more attention in the system, so to speak.
On and other algorithmic social networks these virality-enhancing features only *shift* that attention towards certain things, at the cost of other things.
Wondering why you get more interactions around here with fewer followers? My uninformed hot-take is: that's why. Our "attention budget" is artificially redirected.
2/🧵
So yeah, things are not algorithmically amplified — but nor algorithmically buried either. There is no artificial virality, but there is also no artificial non-virality.
The dynamics are different.
This does not mean things *cannot* go viral — they can, as #JohnMastodon shows if anyone needed any proof.
I strongly believe Fediverse *can* support social movements (it is one), and that interactions here might be more meaningful thanks to lack of certain "virality-enhancing" features.
3/🧵
@robryk @rysiek if you want to be more-literal about randomness: algorithms which we can't see might have literal randomness as part of their process. Mastodon has no such algorithm, and therefor is not random. As I can't speak to how various "sort by recommended" feeds work, I can't say with certainty that they use randomness. I know that in at least one case where I've written them myself, they did involve literal psuedorandom numbers to a limited extent.
> Mastodon has no such algorithm, and therefor is not random.
I think there's actually more influence of randomness here, given that whether I see a given post depends on whether I happen to look at it in a stream of date-sorted posts at the time that I pick by a random process that's not influenced much by other things in fedi. If I had a queue of unseen posts sorted by some weight (or a feed that rate-limits posts I see by picking the ones with highest weight when the incoming rate is too high), the weight would be a good predictor for whether I will see the post. Until the post is boosted by (more) of people I follow, that weight is uniform across all the posts in my Home feed.
@robryk @rysiek perhaps if I can rephrase, I am more-generally translating "randomness" into "entropy", which in the context of this discussion I would further translate into "the ability (or inability) to accurately predict the outcome". I believe that Mastodon, being a more-open platform, would (given datasets of similar size), be easier to build an accurate prediction model for. BUT, if one exposed the (bird) concept of an artificial boost, I would expect the opposite.
Randomness is a good model for the world, in the same way in which temperature is a good model for motion of molecules. You can similarly say that temperature is just a placeholder for all the variables we don't know about (motion of individual molecules).
What I'm curious about is: let's say a similar post was posted by a similar person at a similar time. Would you be more certain about predicting whether it'd become viral based on everything you can see at the time it was posted here, or on e.g. Twitter?
@robryk @rysiek while there are many factors in what causes virality to happen I would hesitate to call any of those factors "randomness". eg: weather is modelled using randomness as a placeholder for "variables we don't have the ability to know". So, if one takes the same definition for "randomness" here, then somewhat definitionally: a system which doesn't have the ability to artificially alter the results regardless of those variables would see more influence from them.