This is the second of a weekly series of posts on various aspects of social software design I find interesting, here is the full list. Each of these posts are written over the course of a few hours in a straight shot. Contents may be mildly idiosyncratic. To vote on what I should write about next, go to this Quora question.

Ice cream sundae

The people who most want to meet people are the people who the least number of people want to meet. The people who are the most desperate to date are those who the least number of people want to date. The people who are the most eager to talk are the ones who the least number of people are interested in hearing. It is the ignorance of this fundamental principle that I see at the heart of so many failed social software designs. This is what I call the Evaporative Cooling problem and one I believe must absolutely be tackled head on by the designers of any communal gathering product unless they want to see their product descend into a squalid lump of mediocrity.

The Evaporative Cooling Effect is a term I learned from an excellent essay by Eliezer Yudowsky that describes a particular phenomena of group dynamics. It occurs when the most high value contributors to a community realize that the community is no longer serving their needs any more and so therefore, leave. When that happens, it drops the general quality of the community down such that the next most high value contributors now find the community underwhelming. Each layer of disappearances slowly reduces the average quality of the group until such a point that you reach the people who are so unskilled-and-unaware of it that they’re unable to tell that they’re part of a mediocre group.

Evaporative Cooling is a dynamic that can apply to both real world and online communities but the affordances of the Internet make it particularly susceptible to Evaporative Cooling. By looking at real world social structures, we can get some clues as to both what causes Evaporative Cooling and what are effective ways of preventing it.

Example the first:

Moving to San Francisco, it was amusing to me, unearthing the social structures around networking that go on here. There is the public “scene” of parties, events & mixers. Alongside this is an entire shadow community of private, invite only, exclusive events which is where all the real work in the Valley is done. It is possible to live your entire life in the Valley, wandering around amicably being blithely unaware of the shadow ecosystem. You could go to the same events every week with the same mix of aspiring entrepreneurs, social media marketers, CEOs of dipshit companies, bloggers & the occasional A-Lister who is forced to be there out of professional obligation.

But, if you’re halfway decent and capable of networking, you’ll soon find yourself with an entrée into a small part of the shadow economy. How far down the rabbit hole you choose to go is purely a function of your innate function and drive. For every layer of exclusivity, there’s almost certainly one more exclusive that you’re not aware of. Some of these venues are well known; TED, Davos, Sun Valley. But for every one of these you’ve heard of, there’s certainly at least a thousand more equally as exclusive gatherings you haven’t. After a while, you start to subscribe to what I call the Groucho Marx rule. You stop attending any event which would have you as a participant.

Lesson the first:

Openness is a major driver of Evaporative Cooling. If anyone can join your community, then the people most likely to join are those who are below the average quality of your community because they have the most to gain. Once they’re in, unless contained, they end up harming the health of the community over the long term. Communities that are allowed to select their members in some way are much more immune to Evaporative Cooling. Unfortunately, most viable internet businesses have no choice but to set their business model to open. The nature of most Web 2.0 businesses is that they depend on extracting a tiny bit of value from a large number of users and are betting on their fuck you exit from massively exploding in scale. Building a thriving community that tops out at 10,000 members over the course of 10 years isn’t going to pay the bills.

Example the second:

One of the communities that I’m part of down here is BayCHI. It’s a community that’s been around for 20 some years now and the quality of the talks and people who attend is still excellent. It seems to have only minimally succumbed to Evaporative Cooling. Why is this? A large part is due to what I call Social Gating. Social Gatings are mechanisms that allow participants to self-select out of the group. In the case of BayCHI, the social gate was the nicheness and unglamorousness of the content. The only people who would choose to participate in this group in the first place are those who find the talk sufficiently interesting to take 3 hours out of their life. This, by itself set a minimum bar.

Lesson the second:

Social Gating is a powerful force and, unlike direct exclusion, works in a much more scalable fashion at Internet sized growth rates. However, it is also a much more subtle one and requires a deft hand to get right. Nicheness is just one possible social gate, charging money is another popular one. But there are an entire constellation of more nuanced ones. Spelling, for example, is an interesting social gate. Just seeing a forum in which ppl spel liek thiz instantly polarizes you onto one side or the other. At the other extreme Quora, in it’s very early days had an incredibly Orwellian system in which Quora staff would routinely directly edit the contents of your answer to fix spelling and grammatical errors. I’m planning to dedicate an entirely separate Social Software Sunday blog post to Social Gating so stay tuned (pro tip: If you want to see it faster, go to Quora and add it to the list).

Example the third:

Another event that I attended this week that had a remarkably high quality of participants was Warm Gun. Among the people in the room were the Director of Design at Facebook and the Director of Design at Google. How did Dave McClure get these two in a room? He put them on a pedestal, literally. They were invited to take part in a panel discussion on how designers & engineers could better work together and it was the inducement of special treatment that made these very busy & high value contributors deign to be in the same room as us design peasants.

Lesson the third:

Unequal roles of participation can help shift the gradient of power and kill the evaporative cooling. When the community is small, such processes can be managed through the social layer. High value participants are treated as special because they have recognition & reputation from the community. But, as the community scales, these social mechanisms break down and often, if nothing is done to replace them, high value members get especially miffed at the loss of special recognition and this accelerates the Evaporative Cooling.

Explicit reputation systems like karma are probably the most popular way online communities have implemented unequal roles. But, for some reason, online communities seem particularly resistant to the type of elitist promotion structure common in real world institutions. In Academia, high school students have to fight to become undergraduates. Undergraduates have to fight to become PhD candidates. PhD candidates have to fight to become adjuncts. Adjuncts have to fight to become tenured and tenured professors have to fight to become Dean. I can’t even think of a single online community that bears even the slightest resemblance to this sort of power structure. This is something to ponder for a later piece.

Example the fourth

Finally, I will examine what I consider to be one of the most successful technological systems ever at scaling while maintaining quality: Facebook. I joined Facebook when it was less than a million members. Since then, it’s managed to grow by a factor of 500 but the quality of my experience has dropped by only maybe 50%. The reason why is because when some random person is participating in Facebook from Brazil, it has an absolutely negligible effect on my experience. Because every user only ever see their tiny corner of Facebook, every user is in direct control of their own experience. Lest you think this is a property that is intrinsic to Social Networks, Orkut was brought down precisely by those random people in Brazil. Facebook’s design, especially in the very early days, was especially conscious of this design dilemma and designed around it masterfully.

Lesson the fourth:

There are two fundamental patterns of social organization which I term “plaza” and “warrens”. In the plaza design, there is a central plaza which is one contiguous space and every person’s interaction is seen by every other person. In the warren design, the space is broken up into a series of smaller warrens and you can only see the warren you are currently in. There is the possibility of moving into adjacent warrens but it’s difficult to explore far outside of your zone. Plazas grow by becoming larger, warrens grow by adding more warrens.

These are the two fundamental patterns of social spaces. Every social space can be decomposed down to a collection of plazas and warrens. In Facebook, your profile, friends and newsfeeds are warrens but fan pages, groups & events are plazas. Twitter is mostly a warren with the exception of trending topics which is the one plaza. On forums, the front page and topic listings are plazas but each forum thread is a warren.

Plazas and warrens both have their unique set of tradeoffs. Warrens are notoriously difficult to get started. New users, stuck in empty warrens often don’t know how to connect to hubs of activity. The onboarding process is crucial and still not well understood (Friendfeed found that people needed to add at least 5 friends to have a reasonable chance of sticking with the service). On the other hand, plazas only need to be started once and then they remain a hive of activity for new users to participate in from the first day.

Plazas are much more visible than warrens so it’s easier to watch and understand your community. In communities, like in justice, sunlight is often the best disinfectant and the neglected spaces often become thriving breeding grounds for all sorts of social pathologies.

But the one absolute killer feature of warrens is that they allow your community to become almost perfectly scale free and grow like mad without ever sacrificing quality. This alone, makes them a design element that’s heavily worth studying to figure out what are the good social designs.

It’s also interesting to note that the real world is intrinsically warren while the online world is intrinsically plaza. In real life interactions, the physics of sound mean that we can only ever talk to a few people at once. Every person gets a “personalized” social life. To give every person the exact same content takes special work. Online, the easiest model to program is to serve the exact same bits to every requester. To provide “personalized” content takes special work. It is interesting to observe how this difference has influenced the evolution of these two mediums.


Evaporative Cooling is a fundamental social dynamic and one that is corrosive to the long term health of communities. This post contains barely 1% of everything I could write about Evaporative Cooling but I’m already at 2000 words and I’m not looking to write a novel here. They say ideas are worthless and execution is everything. Since I’ve gotten to the Valley, I’ve heard probably close to 100 pitches for social products in random conversation. About half of them involved a meeting place dynamic of one kind or another and about 80% of those, as they were conceived, would be killed dead by Evaporative Cooling. It is absolutely essential if you’re to be designing a social product that you deal with this issue up front or you’re just a dead man walking.

To be notified of the next Social Software Sunday piece as it’s posted, you can subscribe to the RSS feed, follow me on twitter or subscribe via email:




  • Pingback: Quora

  • Paul Shapiro

    I think you mean “phenomenon” — and the fundamental principle is not ignorance! It is ignorance of the principle and problems that leads to useless projects and if they are ignorant of things in life then their life will exhaust itself. Interesting first sentence.

  • Anonymous

    This piece is very well-written, and your analysis is intriguing. However, you need to fix your spelling errors! “bares” -> “bears”, “it’s” -> “its”, etc. 🙂

    • Hang

      Fixed, thanks.

      • xian

        one more: “At the other extreme Quora, in it’s very early days had an…”

  • Eliezer Yudkowsky
    • David Gerard

      Beware, however, the cautionary example of Vox – they built the Milton Keynes of social networks, were actually surprised when no-one wanted to live there in the first place and it lost out to the Ankh-Morporks next door.

  • David Gerard

    Heh. I was thinking “has it stopped on Wikipedia yet?” and then noted the warrens and plazas.

    The typical participation time on Wikipedia is around eighteen months. People give all sorts of reasons for leaving, but this figure is very similar to the participation time in MMORPGs, so I wonder if it’s fixable.

    One common Wikipedia life cycle is: people very productively write stuff in warrens, get drawn into the plazas, get disillusioned with the sum of human stupidity and leave.

    I’ve been around Wikipedia for six years, which obviously proves I’m not smart.

  • David Gerard

    Heh. I was thinking “has it stopped on Wikipedia yet?” and then noted the warrens and plazas.

    The typical participation time on Wikipedia is around eighteen months. People give all sorts of reasons for leaving, but this figure is very similar to the participation time in MMORPGs, so I wonder if it’s fixable.

    One common Wikipedia life cycle is: people very productively write stuff in warrens, get drawn into the plazas, get disillusioned with the sum of human stupidity and leave.

    I’ve been around Wikipedia for six years, which obviously proves I’m not smart.

    • gwern

      Thinking about this, I think that you could distinguish warrens and plazas on the temporal dimension as well.

      I mean, a section of Wikipedia could be a plaza spatially in the sense that you could register and participate everywhere and see everywhere if you just add enough to your watchlist. But temporally, any given section is more like a warren – only a few editors stay around as most editors come and go.

      (The converse, something that is a warren spatially but a plaza temporally, is a little harder to imagine. It would be something very narrowly focused but which tons pass through at some point. A blog post like this, perhaps.)

      What this perspective suggests is to try to really edit for the long-term. Take the eventualist viewpoint.

      I will give an example. Back in 2007, there was a controversy and a major reference was something then covered under BADSITES. There was a considerable kerfluffle. I lost the battle to link it and tell the complete story. So, I reasoned that the page was like a warren and the current plaza-dwellers would leave in a few years. Sure enough, come 2010, and no one even noticed its restoration.

      • David Gerard

        How to win edit wars: outlive your opponents.

        Thing I’m not sure about: how Wikipedia’s emergent social network nature affects that it’s a project with a purpose, rather than purely a social network.

        See also LiveJournal, which is also a strange hybrid, and currently in great controversy with owners who don’t care about half the social network side – is there such a thing as forced evaporative cooling of social network?

  • Pingback: links for 2010-10-10 « Blarney Fellow

  • Jon

    Phenomena is plural – phenomenon is the singular.

  • SB

    I think Digg is a nice example of evaporative cooling.

  • Joe K

    very informative article. I think evaporative cooling is “The Cancer” those 4chan-ians are always whining about.

  • Joe K

    very informative article. I think evaporative cooling is “The Cancer” those 4chan-ians are always whining about.

  • Ebun Omoni

    Very interesting piece. You used Quora and FB as your primary examples but what other companies do you think are doing/have done a good job to combat the evaporative cooling effect?

  • Pingback: links for 2010-10-11 « Blarney Fellow

  • Tao Tao

    Fantastic analysis. I’ve had similar thoughts but could never put them into words so succinctly like you did.

  • Carrie Ashendel

    Something that has me a bit perplexed that you might consider for your social gating post: If the problem is that those who are most eager to participate are the ones whose participation is least wanted, then implicit mechanisms for selecting those who are most eager to participate should only aggravate the problem, no? I wonder if the key to using social gating while maintaining the benefits of openness is actually to match the selection mechanism to the type of person you want in your community in a way that least selects for eagerness.

    Here’s an example of how I think this plays out, taken from my undergraduate institution, University of Chicago. When I applied, many, many moons ago, their recruitment brochure had a prison theme, the intent being to select for those who had felt trapped by the drudgery of high school assignments – the people who, given the freedom of the internet, would have long since evaporated from the scene. Then, for their application essay, Chicago required that the applicant answer one of their unique, rather off-beat, and fairly challenging questions. By doing this, Chicago was requiring that the student be eager enough to go to the school that they take the time to write a unique essay for that app. The important part is that that essay represented the type of work the applicant would do on a daily basis as a member of the College. But while this social gating likely removed a lot of the riff-raff that would have otherwise applied to every top-ten school, it also might have lost some of the real geniuses, or anyone who saw the essay as representative of more assignment drudgery.

    My point is that the more a social gate relies on proof of enthusiasm or eagerness, the more likely it is to filter out the top, too, if the self-selector/filter isn’t matched to their interests and abilities. Gates that become less time-consuming for those at the top, like the requirement that one hack into the system to create their membership, potentially avoid the drudgery part of this problem, but even here, the geniuses are still left with the impression that they are entering a community that is below them. Would the geniuses be more likely to wander in and share their wisdom with this lesser crowd if it took less time and effort to do so? I think it depends on what their other options for sharing their wisdom are and how tight-knit the community they are entering is. I have a feeling that the appeal of exclusivity can sometimes negate the drudgery, but only sometimes, and less frequently when the options for participating include warrens. If the warrens are clear from the outside – for example, if a Chicago applicant knows of their brilliant theoretical physics group – the understanding that the filter is not set for that warren will likely make the drudgery of passing it more endurable.

  • Alan Pinstein

    Thank you so much for sharing this fascinating and insightful analysis. I frequently get a “that won’t scale” vibe from many startups, and your thoughtful analogy puts some substance to my intuition that has really gotten my gears turning.

    My company is working on building out some social aspects at the moment and your ideas about plazas and warrens should make for an interesting starting point on devising a way to make it scale seamlessly.

  • Pingback: Evaporative Cooling « I am so There

  • Pingback: links for 2010-10-11 – Kevin Burke

  • crimson

    Thank you for writing this, feeling the mediocrity in me already. I am certain things in Nyc are the same and I am therefore on the fringe with no networking aptitude. I am writing this in part-jest part self-appraisal.

    The reality is I am certain you would be one of the people that believes eugenics is a valid science, and over time your superiority is paramount to the lower classes. Yes social dynamics flux, but if they do not you will lack creative destruction… If the group does not implode and reconfigure to the evolving interests of its’ members, ideas will become ideology and thoughts will draw into a conforming group thought.

    Hurray for nature.

  • Someguy

    what happened to orkut was this:

    1. -orkut foruns allow administrators to choose the forum language. most of them have chosen english.

    2. – everyone respected the administrators choices, on every forum.

    3. – brazillian ppl came in, they showed to respect for the administrators or the forum ppl. they started to talk in brazillian portuguese, off-topic and sometimes even spammed the forums.

    4. – ppl tried to explain to those brazillian ppl that they should respect the rules.

    5. – brazillians didn’t care

    6. – everyone left.

    7 – now orkut is a spam social network full of dumb ppl.

    of course all portuguese speakers are shamed of the brazillian attitude. i am a portuguese native speaker and i, like every other portuguese, respected the rules and tried to talk in the forum choosen language. brazillians just dont care about other ppl. they don’t respect anyone.

    • Guest

      And you, with your sweeping generalization, is very respectful?

  • Pingback: Quora

  • Pingback: Quora

  • Pingback: Leapfroglog - links for 2010-10-13

  • Grrrwaaa

    I don’t agree with the sentiment that only elites make a community valuable. I’ve seen a few communities evaporate because of over-dominance by elites… I mean, would you rather share a house with a random selection of people, or an especially invited set of design gurus and CEOs?

  • Al

    Although your solutions are certainly insightful, I fear that online social networks will have a hard time avoiding Evapoartive Cooling. In the end, desire for growth will outweigh desire for value and they will fizzle.

    I disagree about Facebook. While they resisted longer than others, I deleted my account two weeks ago for the very reasons you described. Perhaps you’re just a few tiers down from me. 😉

  • Pingback: Social Software Sunday #3 – All social software are inherently socio-technical systems « Bumblebee Labs Blog

  • sachxn

    great post on Evaporative Cooling…nicely explained…lesson and example approach….

  • Pingback: Korta klipp – 19 October 2010

  • Pingback: Korta klipp – 19 October 2010

  • GregoryJRader

    Very interesting analysis. One factor that does address some of these concerns is the extent to which a community “improves” its participants. Do members learn to contribute more over time as a result of their participation in the group or is their ability to contribute largely determined externally? If members grow as a result of their participation then your first lesson is not necessarily so detrimental. New members may initially dilute the potential value of the group but over time their contributions will increase.

    This might be achieved either through explicit learning and growth or simply as a matter of learning the group culture. New members might initially be somewhat disruptive but if the group has a strong culture then they will tend to either assimilate or drop out. A group with a weaker culture will be more likely to be overrun as new members with disruptive attitudes join.

    Thanks again

  • Pingback: Quora

  • Pmpodhorzer

    Uff, should I write? After all I am a lower South American. Your article is interesting but it borders simple racism. Do you want an example from the US? Rotten Tomatoes used to be a place to talk about cinema in a more or less non-trivial way. Nowadays is full of comic-book loving superficial Americans, which will kill you if you dare to say something about a piece of shit such as “The Dark Knight”. There you have.
    And you are making money = intelligence (of course, you are American). The Facebook ass only had one (very important) quality: he made his site in Harvard. Nothing very genious at work there.
    I find that intelligent people tend to find the forums where other intelligent people talk, and the subjects are so boring for superficial people that they simply do not bother to come back. As simple as that. Of course intelligent discussion =/= popularity.

  • Steve Bennett

    Very thought-provoking, and I can see how it applies to some communities I’ve been in. Generally, fortunately, ones that have resisted the ECE because of a high barriers to entry and intensely warren-like structures. (Now that has got lots of warrens, will it become less mediocre?)

    Would like to see the Examples and Lessons distilled into Principles, though.

  • Pingback: This Blog Needs No Name | December | 2010

  • Pingback: Quora

  • Pingback: Quora

  • Pingback: Design Articles, Recommended by Dan Saffer | Musings on Design

  • Pingback: Quora

  • Wayne Willis

    Good insights into “group dynamics” online. Clay Shirky addressed many of these same issues in his NYU course. “Evaporative Cooling” and “plazas and warrens” and “social gating” all need to be included in one’s interaction design for any site with social networking dynamics.

  • Quick Detox

    A very well thought out title, one that had nothing to do with what I was searching but found myself interested. Sunday, evaporative cooling effect…. So many ways you can sprinkle that in blog titles, good stuff

  • Pingback: The bar at El Farol, evaporative cooling, and the high performance workplace « archizoo

  • Pingback: Why Quora is in Trouble | Attack Of Design

  • Pingback: » Round up of interesting and creative stuff around the web think throw away think

  • Pingback: Social Networks Evaporation

  • Pingback: Quora正处在危险的边缘,不承认知识,智慧的价值化将导致其毁灭 | 互联网进化论-互联网研究中心

  • Certskey

     I was very pleased to find this site.I wanted to thank you for this great information.
    Testking JN0-101

  • Medion MD98150 akku

    so long to read it andy find some words.

  • Anonymous

    so long to read it andy find some words.

  • Vcfunding

    I think you mean “phenomenon” — and the fundamental principle is not ignorance! It is ignorance of the principle and problems that leads to useless projects and if they are ignorant of things in life then their life will exhaust itself. Interesting first sentence.

  • domain registration

    Thanks a lot for sharing this useful and attractive information and interesting posts from you. keep it up.