The Evaporative Cooling Effect in Social Networks (2010)
blogs.cornell.edu
blogs.cornell.edu
I'm the author of the original piece which was originally penned to warn Quora that their social software design seemed purposely designed to drive evaporative cooling. With the benefit of 15 years of hindsight, Quora has thoroughly evaporatively cooled to an extent beyond even my imagining.
2) shalmanese, how would you relate this back to platforms like Reddit that are driven both by community members as well as volunteer janitors?
Users on modern social platforms optimize output to maximize favorable responses, thereby gratification. There are not really ceilings, barrier to entry, or way around the system, so it's resistant to spams and manipulations.
Classical BBS systems did have this problem. It was said that a community beegins with interesting people posting interesting topics, then uninteresting people joins to read interesting topics, and ends when uninteresting people starts posting uninteresting topics.
What was missing was feedback signaling, and social media got past this at some point during 2010s.
I feel like we must be using different internets. Spam and manipulation are rampant on social networks lately, far beyond what they used to be, and while there aren't really barriers to entry there absolutely are barriers to reach: you're not as widely followed as the spammers, your stuff will be drowned out.
As evidence I offer: any cursory glance at Facebook or Twitter, both of which have likes and retweets.
I can't shake off the thought that this statement might be more truthful than it deserves to be. Some of social media accounts are closer to what you have described, some are more like what I have. The Dead Internet Syndrome must be not spreading uniformly, but there must be significant disparity across fields and bubbles, deepening divides between common folks without clean freshwater supply and those privileged that has access to spam-immune input source.
My Twitter timeline is... not great, not terrible. Reddit is out of question.
Twitter turned into garbage when Elon decided to make it pay-to-play. Giving Blue Checks ranking boosts and extra power when they block others ruined every reply section. On any even vaguely political tweet you now have to scroll forever through a bunch of illustrated profile pics hurling insults before you can get to a real discussion.
Twitter is o-kay. They seem to have largely gave up investments on African-Indian spammer program and it's on its way out. Gratification mechanisms outside of the feedback loop such as paid boosts and reward cash are clearly detrimental to creator performance, so they were destined to be filtered out. Pushing blue check contents is like pushing AI clips in style of Tarkovsky to TikTok junkies, it never works.
I think Twitter users by this point as a collective consciousness must have learned that weaponizing Bluesky/Mastodon transition to trivialize corporate influence is a viable short term strategy, considering how slow and tame changes on the platform has lately been. Twitter had always had such mutually toxic and manipulative relationship between the company and its users.
Twitter has been garbage for a long time before that.
Since a pure chronological feed would be unusable for anyone following more than a few dozen people.
Or for anyone searching any terms more popular than the most obscure niches.
So there has to be some system deciding winners and losers effectively.
Nobody likes that they are the biggest social internet sites, but they are unambiguously the biggest, by a very large margin, and they like to copy each other's worst profitable parts.
Because only the most noteworthy fraction of the population are well… noteworthy.
So the only reliable factors to boost the vast majority of the population way above their peers would be money, endorsements, etc…
It's not a tautology for a site to bias for it though. That's a decision.
I don't step outside on any random day and get immediately blasted in the face by dozens of One Weird Tricks and AI-generated images of Jesus crossed with shrimp.
There’s probably over a million literally deranged people on Facebook.
Even if you only come across a tiny fraction of them, that’s still way more then you could possibly ever encounter in real life, in one physical community, due to simple probability and population density…
As an example, it could be Nextdoor-y[1], bound to physical location rather than global. And that's not a stretch either: it's what Facebook was in the early days. They decided to change it, to become what it is now.
[1]: I do not in any way mean to imply Nextdoor is doing things right, just showing that "social network structure" is a decision, not some kind of inevitability.
Even Harvard kids doing exchange programs overseas signed up for it in the first year of operation.
Questionable. If you're paid for engagement (directly or indirectly) there's strong economic incentives to bait and troll people.
For advertisement purposes, total engagement triumphs, so this is perfectly fine and lucrative for the platform itself, but the quality is not necessarily maintained.
This is nonsense; there are people gaming the system constantly who have to be actively fought. There are whole industries of gaming the feedback. And the feedback process itself distorts the content: it gave us "clickbait" and "youtube face".
I've been watching evaporative cooling of Twitter happen since the takeover and my move to Bluesky, where new users appear in waves every time some new stupid feature is inflicted on the remaining Twitter users.
Twitter has gone past that point years ago, possibly more than a decade ago. It's a warzone of drug resistant attention gamers and wannabes with cash to burn for as long as I remember. Maybe it wasn't as much as it is now during 2007-2008.
IMO, clickbaits and even wooow faces can be considered improvements so long that judgement criteria with presenters and audiences are aligned. Ragebaits are bad, open mouth brainrot thumbnails are disgusting, but a clear and content representative thumbnails would be good - the differences are not in levels of amplification relative to unmodified baseline, but in directions(is the "Inception braaam" bad? I love it and I think it's same thing as clickbaits.)
Evaporative effect as laid out in the article is a situation where "players" of social media as a videogame exhausts motives to play it. The game must continuously supply dopamine release to creators, whether by rewarding ever sillier thumbnails stronger or more insightful comments better, to retain useful players for content supply to continue. Again IMO, Twitter had achieved a near steady state cycle of gratification and content drop by architectural design, careful userbase formation, and useful set-in-stone precedents, relatively resistant to sabotaging and/or manipulation.
Is that entire thing a major net negative to this planet? Maybe. One could just say it and few would differ. It's supercharging scholarly experts across various fields and enabling invasive cultural pressures, so it seems neutral to positive to me.
It is interesting that HN still seems to be very high quality (though I haven’t been using it super duper long to truly judge). Does HN have any healthy gatekeeping mechanisms aside from its (ugly) UI to keep it high quality?
Yes, the mods.
Haha yes. You get access after getting a bunch of karma.
I think I've only accidentally flagged about posts by mistake...
Oh no, I truly believe it’s ugly.
Aside from the users being elitist egocentric pricks and the frontpage being indecipherable to anyone without a CS formation?
I don't think so. But yeah the layout is mostly what does it. Before reddit redesigned itself, it wasn't so bad. Normies want Instagram-like feeds to endlessly scroll through, and they don't like reading because it hurts their brains. So they gatekeep themselves.
The Evaporative Cooling Effect in Online Communities — https://news.ycombinator.com/item?id=1777665 (2010, 35 comments)
The submission here is also more of a discussion of that post, so this thread would be a discussion of a discussion, not direct comments on the article.
Why don't social networks provide recommendation profile transfer. Sort of like stepping into someone else's shoes. E.g. I would like to view how twitter looks like to Simon Willison. This would also make it easier for people to break through filter bubbles.
Obviously you don't see private journals someone is reading, still a good way to discover new things.
[username].dreamwidth.org/read
Which is not at all how actual humans and relationships work. We each bring different value to the table, along multiple dimensions.
Sort of. In practice there are valuable and less valuable contributors.
Plus all those multiple dimensions are not of equal value themselves.
(n and m are arbitrary, I'm not insisting on log base 10, and the natural log e might well be a better fit.)
This is typical of almost all large network functions which exhibit power laws, Zipf functions, or the like.
Measurement itself is difficult and subject to both cost and error, as well as variability over time.
In particular, I am not sure why people are so mean online. I try to be kind to other people even if I disagree with them, but pretty much cannot find anyplace with other people that feel the same. I feel like it is infecting me, and I am not as kind to other people as I used to be or would like to be. I probably need to just stop talking to people online entirely for my own mental health, and contribute to the evaporative cooling of the entire internet.
The internet isn’t great for nuance. When commenting or communicating online, I’ve often tightened up my language for succinctness.
On HN this is particularly challenging because we get a mix of experts and non-experts (and people that have different levels of pedanticness) - so almost any possible statement is going to be attacked for both lacking nuance or being too detailed (and frequently both at the same time). However the people here are nicer, more open minded, and more accepting of nuance than most places online.
To get someone to do something.
There are very few corners of the internet conducive to debate. (Here is one of the exceptions.)
Full paper: https://ojs.aaai.org/index.php/ICWSM/article/view/7275/7129