The Technium

The Bottom is Not Enough

[Translations: Japanese]

I wrote a book, Out of Control, heralding the immense power of bottom up systems. You know: smart mobs, hive mind, web power, amateur hour, decentralized webs, network effects, and collaborative work. Twenty years ago Out of Control made a wide-ranging and exhaustive case for the remarkable things which decentralized, out-of-control systems can accomplish in biology, technology, and cultural realms. Two decades later I’m still keen on the untapped potential of emergent bottom-up systems.

Ft Hdr.5-1

But throughout my boosterism I have tried to temper my celebration of the bottom with my belief that the bottom is not enough for what we really want. To get to the best we need some top down intelligence, too. I have always claimed that nuanced view.  And now that crowd-sourcing and social webs are all the rage, it’s worth repeating: the bottom is not enough. You need a bit of top-down as well.

The reason every bottom-up crowd-source hive-mind needs some top-down control is because of time. The bottom runs on a different time scale than our instant culture.

Here’s how I come to that conclusion. I call myself an editor first, and author second. I think the top-down function of editors — to select, prune, guide, solicit, shape, and guide the results from the crowd — is essential to excellence. From the earliest days of the web, when Wired originated one of the first commercial content websites, a key unanswered question was how much influence editors should wield? In the early 1990s, adhocracy folks such my friend Howard Rheingold (whom we hired to oversee Hotwire, the online content site for Wired), argued for the editor-less crowd. I was on the side of editors.

Howard was at the forefront in the then totally radical belief that content could be assembled entirely from the collective action of amateurs and the audience. I had no doubt that a lot of good stuff could be assembled this way. But I thought of that crowd-sourced content as just the start. I believed then, and still believe now, that the role of editors — what we might call middle people, the PSL (publishers, studios, and labels) — were NOT going to go away. I thought that by adding a mild, smart editorial choice on top of the bottom’s work, you’d have something much better.  Howard believed that we’d get further faster just relying on people with strong voices, lots of passion, and the willingness to write. We’d call those bloggers now.

My argument was that the publisher roles would change drastically, but the need for some top-down selection and guidance would only increase in value. As the amount of content expanded, the demand for some intelligent guidance and selection would be worth a lot to some people. The work of most unedited amateurs was simply not interesting or reliable enough for me. 

Ten years later no one was as shocked as I was to see the Wikipedia disprove this notion, and show how well the bottom could work without any editors at all. Howard was right. For better or worse, the Wikipedia now represents power from the bottom up, the decentralized apex of editor-less knowledge, out-of-control goodness, and infamous the hive mind.  Wikipedia is not the only hive mind out there. There’s the grand web itself, and other collective entities, such as fandoms, voting audiences, link aggregators, consensus filters, opens source communities, and so on, all basking in a rising tide of loosely connected communal action. 

But it doesn’t take very long to discover — if we look hard and honestly — that none of these innovations is pure hive mind, and that the supposed paragon of adhocracy — the Wikipedia itself — is itself far from strictly bottom-up. In fact a close inspection of Wikipedia’s process reveals that it has an elite at its center, (and that it does have an elite center is news to most). Turns out there is far more deliberate top-down design management going on than first appears.  This is why Wikipedia has worked in such a short time.

Time is what ad-hoc systems need, which we have so little of. The main drawback to pure unadulterated darwinism is that it takes place in biological time — eons. Who has eons to wait during internet time? Nobody. Which is where Jimmy Wales enters the picture.

Jimmy Wales, head honcho of Wikipedia, and and an emergent association of volunteer editors created a practice that allowed the Wikipedia to be smarter in a few years than pure dumb bottom-up evolution would allow. Rather than try to struggle to devise waterproof bottom-up legalistic rules to keep persistent vandals from messing up articles, uber-admin Wales would simply ban them unilaterally on the advice of his elite editors. A human editor could discern much better than any rule whether the trolls meant trouble or where harmless. Saved years of wasted effort trying to tweak and tune an emergent anti-troll system.


It is important to remember how dumb the bottom is in essence. In biological natural selection, the prime architect is death. Death powers evolutionary selection. Death is one binary bit.  Either off or on. What’s dumber than that?  So the hive-mind of evolution is powered by one-bit intelligence. That’s why it takes millions of years to do much.

We are too much in a hurry to wait around for a pure hive mind. Our best technological systems are marked by the fact that we have introduced intelligent design into them. This is the top-down control we insert to speed and direct a system toward our goals. Every successful technological system, including Wikipedia, has design wired into it. 

What’s new is only this: never before have we been able to make systems with as much “hive” in it as we have recently made with the web. Until this era, technology was primarily all control, all design. Now it can contain both design and no-design, or hive-ness. In fact, this Web 2.0 business is chiefly the first step in exploring all the ways in which we can combine design and the hive in innumerable permutations. We are tweaking the dial in hundreds of combos:

1) dumb writers, smart filters, no editors.

2) smart writers, dumb filters, no editors

3) smart editors, smart filters, no writers

…ad infinitum. 

The exhilarating frontier today is the myriad ways in which we can mix out-of-control creation with various levels of top-down control. We are rushing into an expanding possibility space never accessible before. It’s the 5th dimension of “no one is as smart as everyone.” A recurring insight (but still worth exploring) is simply: what happens if we turn it inside out and have the audience/customers in charge? As Clay Shirky puts it: here comes everybody! But pure unadulterated dumb mobs is the easiest, perhaps least interesting new space in the entire constellation of possibilities. More potent, more unknown, are the many other combinations of everyone and someone.


It’s taken a while but I think we’ve learned that while top-down is needed, not much of it is needed. Editorship and expertise are like vitamins. You don’t need much of them, just a trace even for a large body, and too much will be toxic, or just pissed away. But the proper dosage of intelligent control will vitalize the dumb hive mind.

Yet if the hive mind is so dumb, why bother with it at all? 

Because as dumb as it is, it is smart enough.

More importantly, the brute dumbness of the hive mind produces the raw material that smart design can work on. If we ONLY listen to the hive mind, that would be stupid. But if we ignore the hive mind altogether, that is even stupider. 

Yet  because the hive mind works at a slower pace on a different time scale, there’s a bottom to the bottom. I hope we realize that a massive bottom-up effort will only take us part way — at least in human time. In the realm of encyclopedias, we want totally reliable articles that are the most authoritative in the world, the most understandable in the world, and the most current in the world. We want news that is relevant, with a low noise to signal ratio. We want research reports that are unbiased but comprehensive and consistent.  We want expertise.

We are unlikely to get the level of expertise we want with no experts at all.

That’s why it should be no surprise to anyone that in the long run more and more design, more and more control, more and more structure will be layered into the Wikipedia. (The Citizendium is one hint of where it is headed.) I would guess that in 50 years a significant portion of Wikipedia articles will have controlled edits, peer review, verification locks, authentication certificates, and so on. That’s all good for us readers.


I know it is heresy, but it might be that the Wikipedia model is not good for very much more than writing universal encyclopedias. Other wiki projects to construct textbooks, species listings, and a search engine have not succeeded — yet. Perhaps the article length is fortuitously the exactly right length for the smart mob, and maybe a book is exactly the wrong length.  We’ll see.

However even if the 2006 Wikipedia process turns out to not be the best way to make a textbook, or create the encyclopedia of all species, or dispense the news, the 2056 Wikipedia process — with far more design in it — may be.  It may be equally heretical to suggest that the improved editor-assisted hive mind will write far more of our textbooks, our databases and our newspapers than anyone might believe right now. 

Here’s how I sum it up:  The bottom-up hive mind will always take us much further than even seems possible. It keeps surprising us in this regard. Given enough time, dumb things can be smarter than we think.

At that same time, the bottom-up hive mind will never take us to our end goal. We are too impatient. So we add design and top down control to get where we want to go. 

The systems we keep will be hybrid creations. They will have a strong rootstock of peer-to-peer generation, grafted below highly refined strains of controlling functions.  Sturdy, robust foundations of user-made content and crowd-sourced innovation will feed very small slivers of leadership agility. Pure plays of 100% smart mobs or 100% smart elites will be rare.

The real art of business and organizations in the network economy will not be in harnessing the crowd of “everybody” (simple!) but in finding the appropriate hybrid mix of bottom and top for each niche, at the right time. The mix of control/no-control will shift as a system grows and matures.

Judged from where we start, harnessing the dumb power of the hive mind will always take us much further than we can dream. Judged from where we hope to end up, the hive mind is not enough; we need an additional top-down push. 

Since we are only at the start of the start, it’s the hive mind all the way for now.

  • bennett johnston

    The best hive mind search technology I have seen is Unlike delicious-type tagging schemes, it’s simple top-down architecture seems to promise efficient and effective deployment of the hive into better search algorithms.

  • MikeT

    Control-freak. This column is an example of how managers spend all their time in useless endeavors attempting to justify your existence and absurd salaries. We programmers, engineers and workers are the ones actually doing the job. Guess what? WE DON’T REALLY NEED YOU. GO AWAY.

  • tms


    Great read. Thanks for taking the time to put it all out there.

    Now, the great project begins. Ask me about “the nucleus.” It’s not written yet, but the idea has been born in my mind so watch out world. Once the idea is born, it’s real…it exists.

    I believe in each group throughout the world, there’s a small portion of it that “gets it.” The task, bringing these people together.

    I’m not talking about politicians or “world leaders.” I’m talking about the people who really get what we’re hear for, how we’re all so closely connected.

    The nucleus will change this world.

  • Kevin

    A very interesting post. My book We Think is published early next month in the UK and covers very similar ground: mobs are not automatically smart and crowds not always wise. What are the conditions under which a mass of participants can start to contribute, connect, collaborate and create robust, reliable and complex products? It turns out none of these online communities is as open, democratic or egalitarian as many think. They generally have a social core and require leadership, albeit of a particular, open kind. If you email me your postal address I will see you get a copy.

    Charlie Leadbeater

    • Kevin Kelly

      Charlie, good points. Address sent.

  • These mixtures of top down and bottom up control in command structures are indeed a useful way of getting “where we hope to end up”. However, thats all they will do, allow you to reach a known goal, or to replicate a known system.
    To my mind the most exciting aspect of systems that display swarm intelligence is their ability to do startling and unexpected things. The danger in maintaining roles with titles like editor or manager, is that they will edit out these new and unexpected abilities of the system. This can happen either through lack of insight and imagination or through fear. Institutions and their individuals often fear radical shifts for pragmatic reasons such as protecting their own interests. This usually resolves to maintaining the status quo.
    The problem is not that a mixture of control methods in systems is not advantageous, its how to prevent the limiting tendencies of top down or hierarchical systems from negatively impacting upon these new hybrids possibilities.
    Hierarchical systems have been the most prevalent
    in our cultures for so long that their logic is entrenched in peoples perceptions. moving away from them will be difficult. especially as the most efficient systems are likely to be combinations of both bottom up and top down.
    The key question then is how do we structure a hierarchy to facilitate a meshwork?
    Time and duration are important here. Bottom up meshworks form fast and loose relationships enabling them to move quickly. Top down hierarchies consolidate their structures over time to minimise internal inefficiencies. I would argue that its exactly these durational qualities that make hybrids difficult propositions. Combinations of the two like an editor who uses a meshworks as source material. fall in to the trap of attempting to force a meshworks into a hierarchy. This combination is limits the productive generative capacity of the network of many to that of the containing institution. If we were however to try to augment a meshwork with some degree of meta control then we would need to use the institutional logic present in the dominant system . This would mean that any power base would be as just another node of a network. positioned not above it but within it and of it. The transition towns network uses such a model. Once a network is formed a small section of that network subdivide to act as a core group. They are responsible for those tasks best suited to a central group. These core groups further follow the logic of meshworks by limiting their potential duration. The first job of the core group is to design its own demise. To timetable a point at which they will be replaced by another group or set of groups.
    The bottom may not be enough but the top certainly is not either. When we come, therefore, to designing their mixture its worth remembering that the bottom can generate the top but the top by definition can never generate the bottom. If we want a functional mixture of these systems we still need to do this through the bottom up approach.

    • Kevin Kelly

      Simon Johnson: I am not sure what you mean by “When we come, therefore, to designing their mixture its worth remembering that the bottom can generate the top but the top by definition can never generate the bottom.” Can you explain?

  • We shouldn’t be so surprised about the wisdom of crowds — after all, we once suspected that enough monkeys armed with Selectrics could probably recreate the works of Shakespeare.

    In media models, where organizations try to appeal to mass or niche markets, there has always been risk.

    Not every song becomes a Billboard Heatseeker, nor every book a New York Times Editors’ Choice. The PSLs provide an editorial layer that helps manage risk and screen inappropriate content. Whether fully or partially automated, the screening process comes from somewhere, even at celebrated crowdsourcing examples like Wikipedia or Craigslist.

    I suspect future value will be driven by having a certain number of the smart elites as exclusives to your venture, or havg tremendous number of a certain type of daily interactions with smart mobs.

  • Jeffrey Kluger

    The bottom is indeed not quite enough, but that’s not to say that it’s not almost enough. There are few crowds more balletically coordinated than a large school of snonemall fish moving toward the same goal–usually a food source. They swirl around obstacles, surf currents, duck predators and do it all with a seeming intelligence that belies their simplicity. How they do it with little to no effective means of communication has long been a mystery, but biologist Simon Levin of Princeton University figured it out. Every school in transit to some goal must have a small group of lead fish that know what and where that goal is. They’re what are known as the “informed individuals.” In the average school, it takes only about 5% of the total group to know where they’re all going and the other 95% will follow. What’s more, the larger the group, the smaller the executive committee of informed fish must be. Often they represent as little as 1% of the total. As for how the rest of the school follows them in such choreographed unison instead of in a chaotic swarm, the key is body length. Every fish tries to stay no more or no less than one body length away from all of the other fish around them. As long as each one does that, there are never gaps that open up in the school or clumps that slow it. No one fish knows very much–but no one fish has to. All they need do is perform their own simple role, and they create a very complex whole.

    • Kevin Kelly

      Jeffrey Luger, your example is a perfect case of what I was suggesting. In my book New Rules I talk a little bit more about this “executive function.” The fish school illustration is wonderful. Thanks.

  • Wonderful post Kevin.
    A colleague and I recently completed a major Australian research project on user-led innovation for the Smart Internet Technology CRC.
    It can be found here:

    ‘User-led Innovation: A New Framework for Co-creating Business and Social Value,’ co-authored by Darren Sharp and Mandy Salomon. The sources of innovation are shifting rapidly from the traditional 20th century model of commercial R&D labs, elite universities, private companies and government agencies to user-led innovation. Today’s users have much greater input into the creation and dissemination of the media, knowledge and culture they consume. Open Source software, virtual worlds and media-sharing communities are at the forefront of new modes of user-led innovation that challenge established boundaries between producers and consumers.

    This new CRC report reveals the major drivers of user-led innovation and explores how it is affecting organisations’ relationships with key stakeholders. It investigates how user-led practices generate business and social value through a major case study of the virtual world Second Life. The report canvasses a number of pathways for organisations to leverage the participation of their audiences, customers and citizens in the interest of co-creating new products, services and platforms.

    The research draws on extensive interviews with some of the world’s leading thinkers on the social, economic and legal aspects of user-led innovation including: Eric von Hippel (MIT), Yochai Benkler (Harvard), Jimmy Wales (Wikipedia), Siva Vaidhyanathan (Virginia), John Howkins (Adelphi Charter), Michel Bauwens (P2P Alternatives) and Mitch Kapor (Linden Lab).

  • Richard York

    Mr. Kelly:

    I was fascinated by the latest essay in The Technium about bottom-up development.

    I think there is one area of bottom up development which may or may not have occurred to you and others: the popular initiative and referendum process which affects politics and civics in both California and my own state of Oregon.

    Even though the popular initiative exists in 22 other states, it is most open in our two states. One can easily view the people’s initiative as one of the earliest exercised in ‘wiki’.

    I have been serving on a citizen’s committee created by the City Club of Portland which completed a two year study of the Initiative and Referendum in Oregon. So I am acutely aware of its use and abuse. However, it is only after reading your article that I realized that what we have in Oregon and California is, essentially, an unedited (or at least un-mediated) wiki.

    And, the consequences of that absence of editing are painfully apparent in both states. Both our states’ systems allow unfettered access to the ballot for anyone with enough time and, more commonly, money to gather signatures for their particular causes.

    I will try to be brief. Originally, the initiative was promoted by the progressive side of the political spectrum. They used it to force major changes on corrupt and recalcitrant legislatures. They also used the initiative, once it entered the constitution, for causes such as direct election of US Senators and women’s right to vote. And, they promoted labor rights and civil liberties using the initiative.

    In recent years though, the initiative process has been used by conservatives to limit taxation, circumscribe the political activities of unions, and restrict the rights of certain social groups they don’t like, i.e. gay marriage laws. It has also been used by the left to compel expenditures for schools and health care.

    Whatever the cause espoused, the initiative in these two states has caused the enactment of laws with little or no contemplation of their consequences.

    So, the committee I served on recommended one major change in the system: that all initiatives which have qualified for the ballot undergo legislative hearings. The recommendation is more detailed. The underlying idea is that by holding public hearings, both sides of the issue can be heard, and there would be time for serious research into the consequences of the proposed measure.

    In other words, our desire is to turn the initiative and referendum system in Oregon into an “edited wiki”.

    I think the whole system, which has existed in various forms for over a century is one of the great social experiments of the last 3 centuries. And, I hope you and others might take it into deeper consideration as you research and write about bottom up processes.

    The full text of the report can be found here:

    Rick York

    • Kevin Kelly

      Richard, the referendum issue is a great example. Thanks for bringing it to my attention.

  • Chris

    Very nicely put. I’d suggest that the example of death is a somewhat misguided comparison to the bottom-up systems people have attempted to place on the Web. Each individual actor has many personal motives, in a human population, and what we end up with is simply an average. The better rule-of-thumb in a natural evolutionary setting is the group that survives, not dies; a mix of attributes turns into something that -can- survive, but has many undesirable traits. Death is your top-down editor in this case, he just happens to not be a very good one. Would that there were a better ;)

  • Your words place great trust in humans to wield technology to properly manage change.

    Don’t you think that our human hive and top civilization has imposed its own deadline? Namely global warming and peak oil. Just to deflect arguments — one could easily say that we have 20 to 30 years before we know the full extent of global warming. By then we will surely define the full extent of the problem with certainty.
    And prognosticating within this same time frame, Moore’s Law triggers discussions that an Artificial Intelligence Singularity will pop into being (whether we will know it is another matter). It seems that your Hive/Wiki is a human civilization interface that adds confluence to help make that happen. The viability of that Hive/Top system will be tested in preserving a habitable future.

    With an increasing rate of change ahead, amazing changes due.

  • Excellent insight, Kevin.

    I have a strong suspicion that as important as the removal of friction the user-contributory technologies is to enabling wide-spread *authorship*, the effect of that same fast tracking to prominence of talented *editors* will be as or more important, empowering and spotlighting new editorial voices to help us cut through the morass of data to carve out information and perspective.

    • Kevin Kelly

      Marc Meyer, I agree.

  • Both this post and the comments are very insightful. It’s very encouraging to see this being viewed from so many different perspectives, yet still as a common subject.

    I agree that all “bottom-up” systems require some form of governance (i.e. “top-down”). This is true for all distributed, self-organizing systems, from democracy, to software, to the human brain (all of which were cited by your visitors, with some great links).

    In my view, however, there is a major difference between an ideal manifestation of governance and how governance is established in most existing “top-down” institutions. Nick used the word “meritocracy”, and described a top-down structure emerging based on actual transparent merit. I believe that’s one key difference – that governance emerges from merit rather than status or capital, and remains flexible.

    Your answer to Nick was that editors make aggregate decisions. Toffler once argued that the only reason aggregate decisions had to be arbitrarily made was due to limitations in our information technology at the time the systems were established. As a technology-savvy editor, do you agree with this position? If so, can you imagine a “leaner” editorial process, perhaps one so lean that it’s not technically top-down at all? Something like an editorial gardener.

    I hope you don’t mind the shameless plug, but I have recently started a blog very much related to this very subject at I can only hope to be given the quality of feedback seen here.

    As far as your opinion that the bottom is not enough: does that mean you changed your mind back again about what’s possible? (

    I hope not.

    • Kevin Kelly

      Eyal, yes, I think a leaner top-down editorial process is possible. Your metaphor of gardening is spot on.

  • Kevin,

    I finally was able to write a riff of your thinking in this post.

    “On the balance of top-down and bottom-up”

    Yeah, global top-down works if it comes from the next layer of network above. It won’t work from within the network.

  • indir

    The reason every bottom-up crowd-source hive-mind needs some top-down control is because of time. The bottom runs on a different time scale than our instant culture.

  • Raphael Kasper

    K^2 –

    While I find myself in agreement with your conclusion that top-down influence is needed to make a bottom-up system work, I believe that the reason you give for reaching the conclusion displays something of a misunderstanding of evolution and its results. You state, at one point:

    “Time is what ad-hoc systems need, which we have so little of. The main drawback to pure unadulterated darwinism is that it takes place in biological time — eons. Who has eons to wait during internet time? Nobody. Which is where Jimmy Wales enters the picture.”

    One point that even mortal adversaries in the evolution controversies [Gould/Dawkins, for example] agree upon is that evolution has no goal, no purpose, no ultimate objective. Evolution proceeds, and organisms evolve, but there is no “progress.” To say that bottom-up systems merely need time to allow useful or innovative ideas to “evolve” is a misinterpretation of Darwinism, yet you pose this as the only reason to arrive at the [correct, to me] conclusion that some top-down direction is necessary. In fact, if bottom-up ideas develop in a manner similar to the evolution of living beings [a point with which I do not agree], it would not be lack of “time” but rather the goal-less, aimless development of the ideas that would lead to the requirement of top-down influence.

    Raphael Kasper

    • Kevin Kelly

      Raphael Kasper, about progress in evolution. This is a MUCH longer discussion but I think it clear — even though neither Dawkins nor Gould admitted so — that there is a long-term direction to evolution, while short and medium term evolution is without destiny. Gould liked to say that any long-term direction in evolution was due to the lack of anywhere else to go, but he could not get around the fact there there is a trajectory. Even with directed evolution — breeding — it takes time, more time that say outright genetic engineering, where you just jump to the state you want.

  • Some form of control is needed in these crowd based sites, if their purpose is to give factual information.

    Wikipedia and Google have become the cornerstone of the modern web. Once they start learning, we will be heading towards a technological singularity. However, that might be years from now. In the here and now, Wiki needs to become more factual, less vandalized. The editors need to stop behaving like tyrants.

  • yeah, just like the dark ages….again

    think monty python and the holy grail, society knee deep in poop.

    all of it “free” btw- ;)though you have to pay for the gruel to get the poop.

    we need no reminders that mankind can poop alot. Every new born does it, every VC does it. Every blogger does it.

    the real question is, do we want to be the borg, a race that can be defeated in 10 minutes by a network cable, a half naked frenchman, and the order to “sleep”.(poop)

    so far, seems like we do.


  • vanderleun

    Or, to paraphrase the kid about the structure of the universe, “It’s hive minds all the way down.”

  • You touched briefly on the fact that natural systems run on the death and removal of unsuccessful individuals. Another way that this differs from our system of innovation is that everything — not just life — is dear to us. We discard so little of our innovations, data, ideas, or plans. Think about the amount of information we have about the technologies of ancient civilizations, and the meticulous way in which we seek to discover and preserve more of it. Conversely, consider that around 99% of the organisms that have ever lived are extinct, many with no fossil record, and certainly very little influence in the dealings of current animals; Not only are a massive percentage of organisms extinct, but many evolutionary branches — whole phyla of animals — are totally gone.

    Until we get a lot less attached to our information, we can’t expect to match nature’s ability to create diversity.

  • Stephen Balbach

    Very nice article.

    The editorial control is of course tightly integrated with the design (software) of the site. Wikipedia works because the site design was built piece by piece through trial and error in the early days with the goal to create a 1-page encyclopedia article. To build multi-page books the same way, I think we need a new type of site design, the Wikipedia design doesn’t work as well. There are groups working on this, like the blog “Future of the Book”, it probably won’t look or behave like the software at Wikipedia because the editorial controls (top down) and bottom up need to be different. Wikipedia is good for 1-page articles because thats how the software was designed (article history tab, talk pages, title, etc..) it’s a hack to move beyond to a multi-page document.

  • maybe it is still a mistake to use the opposition top/bottom, experts/crowd too heavily. it is still a static model with old user roles. but these maybe are not representative for the structures we think of (and already have out there).

    i’d propose to discuss things not so much in terms of “editing content chunks”, but “circulation of statements”. part of this new ecology are all kinds of creating, filtering, editing processes, but they are neither feeling like “creating original material” nor like “editing something”.

    it is more like “being a vital part in a dynamic socio-semantic circulation process”. this fundamental feeling is what has to be created in organizations and individual workspaces as well.

  • This post reminds of Dan Dennett’s Virtual Machine position in the philosophy of mind, which he contrasts with Paul Churchland’s bottom-up connectionism.

    I think I side with Churchland on this one, though.

    The papers I’ve pasted are representative of the positions, I think, although not necessarily definitive statements.

  • Gordon Whiting

    There’s already a mix of elite + mob in our current everyday web, in that the contributors (to blogs, forums, bookmark site, social sites) tend to be motivated individuals who want to participate in the conversation. The truly dumb in our society are less well represented on the web than the really smart, and those really smart types are aware of cues, call the ambient editorial cues, of the web environment. A really good critic/commentator gets attention and respect, and an ambient influence flows from them. That’s what is (successfully) replacing the old top-down command-and-control model. And it does not take as much time as biological darwinism, as you argue. People paying attention to other smart people gets things moving along quickly. I think we are just getting used to the availability of this type of editorial influence, and as the cues become more understood, ingrained and accepted, intelligent information will be produced on the web in a less haphazard way.

  • I suppose we’ll have to collect and observe more examples of different balances between network and hierarchy in order to know what approach works best under various circumstances or stages of processes.

    There’s a lot of knobs and levers to adjust and take into account: Accountability, visibility and reputation. Personal power and gains, vested interests. The complexity of the issue, the speed of change. Level of trust, level of risk. The degree to which participants see themselves as having shared interests…

    It’s crucial to be open and realistic about what’s in fact going on. It seems that there’s a popular myth connected to web 2.0 or the open source movement about a new type of communism or socialism arising in which everyone shares freely and on equal terms. Certainly, that’s part of the magic. But at times it’s as if the movement is in denial about the fact that there is typically a strong core with very centralized power.
    Rather than treating power, leadership or vision as taboo, it would be really interesting to look closer at the nature of power in networks. Is it power that someone grabs and uses for his or her own purposes? Or is it the kind of power that’s bestowed on someone – the sort of power those that make others greater have?

    With a more nuanced understanding we might be smarter in designing processes that fit a particular purpose or context.

  • T

    It’s a somewhat useful insight that the best results come from a combination of bottom-up and top down. What would be really useful is to know how the role of the “top” needs to change as you go to a bigger bottom. An obvious difference is that editors don’t assign stories, they select them. But how do they solicit? How much do they edit the ones they select; they should clarify the writers intent, not make it fit the prevailing paradigm that they would have gotten from staff writers. How do you get good writing, new insights, discover news that other sources can’t find? How do you keep the “hive” enthused and producing?

  • Jonathan

    The top-down approach is a necessary aspect to a practical mind as it imposes rules and guides the process. Steering a mind from the bottom up requires complete knowledge, the system is far too dynamic to be controlled. If there is a goal, there is a god that defined that goal. That god works from outside of the system, beyond the reach of worldy laws.

  • Marc Rettig

    Seems a shame to pick a nit after such a thorough, thoughtful, useful article. But it’s bugging me, so I have to vent….

    I’m thinking that reproduction, not death, is the one-bit intelligence that powers evolution. Still one-bit, the point still holds. But as far as evolution is concerned, who cares when you die so long as you’ve managed to reproduce.

    • Kevin Kelly

      Marc Retting: Yes reproduction is also just one bit, but it really doesn’t decide (or change anything). It continues what was decided by death/life. Change in evolution begins with mutation (random change) which is weeded, amplified, extended and made real by either dying off or living (to reproduce).

  • The open-writing/editing model is similar to the free market model…and we know that free markets are never totally free. In fact, even the most laissez-faire economists recognize that the invisible hand sometimes needs to be held.

    The approach of open-and-guided clearly has the best opportunity for survival. The interesting thing, of course, is that the guides rise up from the environment in the precise same way as the content. The Wikipedia elite was not elected. They simply emerged.

    Thanks for the thoughts.

  • Hybrid systems are what we desperately need politically as well. In Switzerland, where citizens as well as Parliament have made law for 160 years, you never see government torturing, spying on citizens, breaking the law via “signing statements,” adding trillions of debt, etc. They call this hybrid “co-determination.” Works for couples, too!

    The best project for better and NATIONAL ballot initiatives like the Swiss have is led by former Sen. Mike Gravel. YOU can vote to ratify it at, much as citizens ratified the Constitution at the Conventions.

  • gregory

    hive mind is similar to the wisdom of the crowds phrase, but another older concept is really in charge, the collective unconscious, the group mind, consciousness itself

    we and all of our technologies are just along for the ride

  • ec

    HCE – here comes everybody

    Original source for this formulation

    James Joyce Finnegans Wake

  • Nick

    Any member of the hive/swarm can be also editors. The aggregation of these editors and the editors themselves emerge. At this point you have the benefits of bottom up (meritocracy), but without the drawbacks of top down (possibly dumb). And perhaps this ability to have members of the hive become editors is just another top down design. But then again, it can be hive designed. I guess it does not matter if the ability to edit is by hive/swarm design or by top down design, as long as the merits of editing are hive/swarm controlled. If/when the merits of editing is top down, the hive can vote for or against it.

    • Kevin Kelly

      Nick, yes editors can come from the bottom, and like rulers in democracy, their power ultimately rests on the bottom, but still they make decisions without having everyone decide on each point.

  • Phil

    This reminds me of the book _On Intelligence_. One of the brain mechanisms described in that book had neurons that processed abstract data feeding predictions back toward the neurons that collect raw sensory data. When the predictions don’t pan out brain activity increases in an effort to explain the input and sometimes to build new predictive mechanisms. I think that the role of “editor” in systems that display emergent behavior is to serve as a sort of repository for data gained through this kind of feedback loop. It seems very possible that feedback, coupled with prediction and correction mechanisms, could provide a context in which the emergent behavior of a bottom up system is very similar to that of the mix of top down and bottom up that you describe. In fact, the American system of government, which is a republic rather than a pure democracy, seems to be a fair approximation of such a system, with politicians floating positions (which are like predictions) and the electorate providing feedback that directs but does not completely control the system.

  • nmw

    I see you have not simply limited the formatting of this text to HTML: you have also used css, rss and you have incorporated javascript code — was this YOUR decision? Are you at the top or at the bottom of this decision making process? Why did you write the text in English?

    I like the visual imagery of bottom up vs. top down — this is as vulgar as the crowd, and it’s the crowd that “creates” language and (according to Wittgenstein) it’s the crowd that gives everything it’s “meaning”. Perhaps an even better example of the “wisdom of the crowd” than is… But were the crowd to limit it’s language use to that of babbling babies, then we would not know “progress”.

    Progress is built not on geological time but nonetheless it is “handed down” from generation to generation. And it is indeed intriguing to view “death” as a critical element of progress (no doubt an “ode” to Darwin’s theories?).

    With the invention of writing, it was possible to transport language across larger spans of time than just the fleeting span of moments when sounds dissipate into thin air. With the printing press, it became possible to mass-produce such texts, and thereby we began an era of standardized language.

    In the past decades, it has become possible to mass-produce publishing — indeed, “desktop publishing” is currently fading as a “new and improved” concept, gradually paling it’s own death. Left in the wake of this epochal change is a renewed multiplicity of language, a plethora of frames of reference — such are the best of times, and such are the worst of times….

    But it is agreed: time will kill off the nonsense. “Babble” and “Google” will give way to “House” and “Home” — and the language we use (and thereby “create meaning” with) will be tomorrow’s “state of the art”.

    Before Susan Sontag died, I met her in Tübingen. She was a wonderful woman with a great mind — something I had only discovered shortly before the speech she gave there. She explained how she carried language with a certain responsibility — a responsibility she shared with “top” poets such as Shakespeare and Goethe. And she also explained how she was more than willing to hand down such words of wisdom to quench the thirst of those who were ready to read it off the tip her tongue.

    Such is the “wisdom of the language”.

    :) nmw

  • Oliog

    If the hive were to be invested with any degree of innate wisdom I imagine it would be the wisdom of the Cheshire cat who observed that any way is the right way, if you don’t know where you want to go. The problem I think begins and ends with the notion that there is a predetermined target; that someone knows where he wants the hive to go! And, in your marvellous book all those years ago, you illustrated how easily the hive can be directed ..

    The other obvious concern will always be to decide just where any editors would sit — outside or inside the hive. It carries, in my confusion, some shadows of the ontological proof of the existence of God. It also echoes some of the Quantum issues about whether the observer / reviewer is a significant part of the experiment .. Indeed, there are many parallels with Quantum Theory.

    And, in this context, it is difficult to see any positive influence for editors. In their most identifiable role one can see the stultifying effect they have had on published literature where consistent variation is so rare it allows Reviewers to operate from a very limited range of Templates to declare on what is acceptable. Each role, reviewer and editor, can be comfortable interchanged because both have the same target; both know where they want to go; and it is far more convenient if they follow the same path. And I, immediately, declare my prejudice here as a self-published writer on

    I do believe, however, that, to reap the benefits to hive power, the vital role falls to reviewers rather than editors – were we to succeed in separating the two. Editors are a potential virus, either purposefully or accidently, whereas reviewers should form the immune system and, as such, exist simply as part of the system, promoting or retarding elements with no greater authority than their reviews deserve.

    The essential element here, it seems to me, is that the hive is an organic system and as such has no purpose or target beyond survival, and possibly procreation (the marvellous fruit of the tree). Editors would simply steal any resulting offspring and make them dance on top of barrel organs for pennies.

  • Your conclusions make sense mainly for publications which deliver content on a regular schedule. It doesn’t seem to apply to more personal and expressive kinds of writing or for individual blogs. Editors help with setting goals and themes and targeting specific audiences.

    Actually though I would guess that editorial functions have changed significantly over the last decade. Now editors are less focused on filtering or polishing text than simply recruiting or motivating talented individuals to contribute.

    I’ve always found the studio module of running a TV series to be fascinating. It’s a different kind of editorial management; you are managing an imaginary universe of characters and situations. I have to wonder what would happen if a TV series suddenly decided to choose their season’s episodes from the fan fiction site instead of a stable of writers. In many ways, fan fiction is more adventurous than the scripts which actually make it on the air (though I’m sure the ones that do are tightly polished).