Yochai Benkler - Transcription of the Tidia Workshop Talk Thank you very much for inviting me. I was thrilled to get the invitation. I've been hankering to learn as much as I can, about what you are doing, and, even just today's conversations that I've had have been enormously instructive. So, I'm grateful to the opportunity to come and learn with you and from you. I want to talk about something that our contemporary understanding of economics suggests shouldn't be there. And studying something that shouldn't be there is difficult. The analytic framework that I will take today, tries to explain the emergence of peer production of information, knowledge and culture. And emergence of sharing networks for material physical capacity. Within fairly traditional welfare economics, but with the heavy emphasis on transaction costs economics, this is not the only way to think about it, it is not the only way that I think about it. I will be talking in a couple of days, more, at the university, about the political theory and the social theory that goes along with this. But today, I wanted to focus based on what I understood to be quite central to the Project of the Incubadora, TIDIA, about the economics and the organization. So, here is what I want to try to do. I want to try to describe in very rough terms what peer production is. And, I'm going to refer a little bit to free software, but not much. Then I'm going to, very quickly, outline parallel events that we see in a computation storage and communications through distributed computing, and other more sources of shared computation and communication resources. Then I want to ask, basically, three economic questions: One: What is the motivation? Two: What are the feasibility conditions or the enabling conditions that make it possible for something that ten years ago we would've thought to be utopian, to actually be a sustainable mode of production in some domains of information, knowledge, culture and communications? And finally, I will ask what I think is a mildly less important normative question. But one that is nonetheless important, within the analytic framework economics, just why is it efficient, or at least sustainable, to use social relations as an alternative transactional framework to markets and firms. And then, I will close by doing something that begins to work towards how one design such platforms and works on such platforms in terms of solutions to the commons problems, that is to say, solutions to cooperation and coordination when you have given up exclusive property rights as the primary modality of organization and control over behavior. So, lets start with "why are we all here?" or "why are many people here?". Certainly, in the United States, this is the single strongest knockdown argument against anyone who says: "free software is just a fad" or "peer production doesn't really work". This blue line over here is the gross rate market share of the Apache web server. It is free software; down here, this is Microsoft's share of the market, and over here is the share of various other systems that you see up here. The basic reason that this is a knock down argument, is the first resistance to arguments about free software, and later on, as I'll talk about much more peer production, more generally, the same phenomenon when it occurs in other domains, is how do you know it works, the markets have always worked, its not a market system, and for anyone who relies on the market. As a reason to resist this notion of stability, this is a very powerful argument. Because the people who are putting, who are making these systems work at this level, are Amazon or Google - in the case of Linux, not in the case of the web server - they use their own server, the US department of defense, for example. The organizations that depend on stable servers adopt Apache. The same thing with slightly more vague numbers can be said of the GNU/Linux operating system. This is share of working servers, not share of desktops. This is a snapshot from 2001. I haven't found a better one. Judgments and any event go between 20% and 40% of the market, but again it's the same kind of argument. Companies and people for whom it matters whether their computers crash or not, who operate a server, use free software and that's important. Now, another little tidbit is IBM is the biggest patent holder in the technology field in the United States. And here are some revenues of IBM from just two areas. One is what shows up on their annual reports, as intellectual property transfer, licensing and royalties. And the other is what shows up in their press releases every year, not in their annual report, they don't have a separate line, as Linux-related services. So, in 2000, they didn't do any Linux-related services. Then, they made the big announcement that were going to invest in Linux-related services, and ever since then they've been reporting increasing revenue. Another way of looking at the same thing is, if you look at all of the revenues from this two areas alone together, in 2000, intellectual property related licensing was this kind of a share out of this two. And then, over the next three years, here is what happened. This is the largest patent holder in the technology field, in the United States. Ok, that is all I planned to say about free software. Not because it is not important, but because it's an area that many others have, and do speak about, and I have to speak about it later in questions, and tomorrow, if that's what people want to focus on. The thing that is less appreciated is that free software is one very salient, very important, but only one example of a much broader social and economic phenomenon. And that is the emergence of peer production, and more recently I've been focusing on the parallels on sharing of material resources. So these are two discrete economic phenomena. Their explanations have some overlap and are somewhat separately. So, I will try to keep them as separate as possible, while allowing us to learn from both. You need to keep in mind that I'm not talking about a single analytic framework, but I am talking about their economics of human creativity, their economics of a certain class of material goods. And both of them, for these different reasons, are converging on having some substantial social transactional framework, work [further]. So, peer production... What is peer production? A general category that is descriptively defined as various sized collections of individuals that effectively produce information goods, without price signals or managerial commands. This is very important. It's a negative definition. Here is something that is in fact observed in the world, that can be understood as effective production and that doesn't rely on the two mechanisms that all standard economic sees as defining the universe of production models: price signals or managerial organization. So, it's a negative category. Here is the set of things that are happening that we need to study. Sharing of material resources is a similar definition. These are large- scale practices of effective productive sharing. In this case, it is not sharing of human creativity, but sharing of components of material resources. And particularly what I'll talk about here are computer computation cycles, storage and communication's capacity in wireless. The components are material goods, they are economic goods. They are not public goods like information, and the outputs are not necessarily public goods. The outputs might, themselves, be rival, like computation, like storage, they might not be scarce for a particular reason, and we will talk about that, but they are not public goods. So, two quite different phenomena. Let's first focus on what peer production is. To some extent we have to realize that peer production is not an entirely new phenomenon, right? If we look at this definition: variously sized collections of individuals effectively producing information good, without price signals or managerial commands. That's us in academia, right? And has been for a very long time. The dean doesn't tell me what to go study, and if he does, that doesn't help him, and neither do I decide what to study based on my sense of selling the output as a good. So, a large chunk of our basic science, our basic academic framework, is already on the form of peer production. What's new is that we talked this relatively discrete phenomenon that still required full time obligations and full time participation from all of the participants. And now, we are translating into something that can be done in small and large scales by everyone, for everyone, in any kind of framework. Not only in those relatively narrow set of fields that we've decided to put public funding into the formal academic framework. So, at a very unstructured and minimal sense, the web is a form of peer production. You just think at the simplest level of what you used to need, an almanac or simple encyclopedia or fact book for, when you say: "I want to know the birthday of someone or something else". Today, you would go to Google to do that. You would go to Google not on the assumption that when you find the result you are going to have to pay for it, but on the assumption that someone, somewhere in the world, found it interesting enough to study this person to put up on a web site the information. And you can cross reference a couple of places to make sure you got it right. But the web is a very primitive and unstructured version of peer production, where we have no cooperation. But we have parallel coordinate behavior, for reasons that are social, without being organized. And nonetheless produces this vast information good we call the World Wide Web. That's interesting and important, but it's still only very simple and very primitive, as a social model. Beyond that, we begin to see, and we can systematically see, in the various stages of the communication's process. Emerging much more stable models of production, and systems of production, that are producing discrete components, and the way that I organized this, is just to think of it as just three big buckets. You can think of authoring, saying something, producing the initial content. You can think of filtering or accreditation, right? What of the universe of statements out there is worthwhile, or interesting, or relevant to anyone as the second discrete component, and you can think of distribution as the third component. And so, I will just quickly focus on each of this three to identify to you that we are in fact seeing on the web a large number of these kinds of peer production systems, where large-scale groups, or medium-scale groups of people, cooperate in order to produce content relevance and accreditation, and distribution. The simplest and most intuitive, and that's why I like to start with it, is Mars Click Workers. This is the interface of Mars Click Workers. NASA basically took down images from Mars, that they use to take the actual images and then use full time scientists to map the craters. Then, there was an experiment for a few months. What happened if they took these images and they made these little snippets out of them, and they put them out with the simple interface that allowed people to click around in three spots around the edges and then they would compute some kind of a location for the crater. And they tried it for six months. Eighty five thousand people played this game. Very often, they would have twenty, thirty people look at the same little two square inch - unfortunately they use inches there - interface. And the result at the end, when they compared it to the same areas mapped by a fully trained scientist, was, to quote the paper, practically indistinguishable. So, this is very easy and intuitive to see when you take a task that used to be organized in one way. You restructure, you create small modules, you put it up in a way that people can participate, not by spending their entire life, and needing a salary, etc. But, by giving five minutes here, ten minutes there, instead of watching TV, and you produce an output that is practically indistinguishable, from the one in the other model. So, that's why this is intuitively easy to see as a difference. Closer a little bit to the academic process is Kuroshin, or K5. This is basically a system that allows people to write editorial type essays. I used it not because it is the most important or best, but because it's very interesting in the fact that it has this model of peer review. Someone posts, as you can see, I've been using this now, for two years. This is when Saddam Hussein was still in power and people were challenging a variety of proposed actions in Iraq. But, that's all behind us now. So, someone wants to put up an article. They put it up on an interface. It goes up as a pending story. And then, people begin to vote: should it be allowed to stay or not - participants, registered participants. If a certain number of people - ninety five in this particular case, as the platform provides it - say that it should be on, it goes on the web site. If it should be kicked out, twenty people, say, it's kicked out. People, as you can see, can vote, they can abstain, they can choose whether it should be posted on the front page, or on a section specialty page, or whether it should be dumped. They can also provide specific comments. They can provide editorial comments - things that we are very familiar with, from the peer review process. All of this is done before it is posted to the front page. And then, you can repost, the same notion of a peer review process. But instead of having one or two reviewers - who are already accredited by their academic positions - what you have is at this point about fifteen thousand people, doing it for each other in numbers of hundreds looking, or at least dozens looking at each particular article and making each particular piecework, and [it's work]. Of course, we see Slashdot, with now something like half a million users, quarter of a million registered, active users, producing what is effectively, not the only, but a very, a basic required source of a newsletter, for anybody who's interested in technology, again with a slightly different, but still a sophisticated system for comments, and reviewing comments, and throwing comments that aren't worth reading below the radar screen and raising comments that are by having multiple people review each comment and give it a... Who knows Slashdot, here? Who knows of Slashdot? Ok, so I don't need to explain. To the extent that I can see through this light. Now, let's talk about what is increasingly becoming the most sophisticated example, in one sense, and that's Wikipedia, and I take it you've already talked a little bit about Wikipedia. Wikipedia is an encyclopedia written by volunteers, at the moment there are something like, I think it's now gone up to twenty thousand or so volunteers who participate, in multiple languages, creating definitions of encyclopedia style for many areas. Now, the interesting thing about it is that you can actually go to it and you can find topics that you know something about. And you can see for yourself, whether you think it's worth something or not. Now, it started out in 2001, in English, and very slowly, we have other languages added as well. The Portuguese version, for example, had relatively slow growth over the past year, year and a quarter. But, as of this summer, there were about eight thousand articles with two hundred characters and one link and ten thousand overall. Now, three months later, there are already sixteen thousand articles of two hundred characters and a link, and something like twenty five thousand articles overall. The number of participants has also increased and now stands at about two hundred and fifty in the Portuguese edition. And there are multiple, multiple editions. The German and Japanese editions now, each, have over a hundred thousand articles each. The French version has about forty thousand, as at about two months ago, but the speed with which this is growing is tremendous and again, you can't show by comparison to free software, a real market comparison between this two, but at least because it's an encyclopedia and you know your own area you can find two or three things that will give a sense for yourself. Here if you look at, for example, the definition of Mathematics in encyclopedia.com, which is the Columbia encyclopedia, a popular commercial encyclopedia you will see that actually Wikipedia compares very favorably, we won't go in to the whole thing, but the other thing that is interesting about Wikipedia is that, unlike Kuroshin, unlike Slashdot, which have hard technical constraints, where, for example in Slashdot, you can't moderate, you can't say whether a comment is good or bad, more than five times in any 24 hour period. So there is a hard technical constraint, to make sure people don't try to put it away. Wikipedia is different, in this regard. In a sense that anyone can post, anyone can change, anyone can change a definition, what happens is a combination of social norms: we are all here to produce an encyclopedia. Enormous transparency, so you can immediately see the changes that others have made. And, people who go back have a very easy mechanism to revert back to the definition that was there before. So, for example, you'll find something like five, if you look there is a history of every definition, if you look at the definition and you see somebody tried to insert something malicious - in a sense of a joke or maybe just very wrong - it will usually take on the order of five to ten minutes for someone to see this, and pull it out. Because for any different subject there are five, six, seven people who are interested into this subject. They all have automatic notifications that there has been a change. One of them will have a few minutes to take a look at it. If it is easy, if somebody has just come up with: "Barbie really exists and she lives in..." This just happens to be an example, I have fun research. I was researching Barbie definition on Wikipedia, the other day. Somebody changed the definition and said that: "Barbie really exists in some island, somewhere..." It was about five minutes before somebody else just took it off and put back the original, and put the other one in another functionality of "talk". So, Wikipedia is very interesting, also because it uses a lot of transparency for purposes of social regulation of the behavior as opposed to strong technical constraints. And we will talk about this closer towards the end. Now, Google, for the single most important application of part of Google, Google uses peer production for accreditation, right? Google doesn't write an algorithm that will decide which will be first based on its employees sense of how one judges best. Google's algorithm goes out and polls thousands of websites on the web and say, "who do you think is the best?" and the way that we know that you think it's the best is you've linked to them. So of the universe of websites that have a particular word, or a particular combination of words, which of this is the one that people around the web, for their own purposes - one person wants to show that they are wrong, the other person wants to show that they are right, one person wants to sell them, the other person wants to give them away for free, everybody for their own reason has put up on the web, either link or no link to this particular site. What Google does is it goes and it polls this votes. So, Google has taken its single most important value, the ability to say which are the most relevant things to you, and put that in a peer production mode. And if you look for Barbie, you see, and this is just a little bit of presaging the politics of this, the first one not surprisingly is barbie.com, but very quickly after that you start seeing that Barbie is also a very challenged cultural product, so you see "Adios Barbie, a Body Image Site for Every Body", the challenge of the perfect body, etc, that the doll represents. Now, again if you compare this to the market as it were. Here's Overture. Overture is a different search engine, and Overture the way you know where you're ranked, first, second, third, is based on how much the person who has ranked you paid to Overture. So Overture is a market system. It is like advertising. I want to be seen, I pay Overture and then when somebody looks for Barbie, my link comes up first. Again, if you are interested in market testing of efficacy - all of you, I'm sure, know Google. I'm not sure how many of you know Overture, and that is... some... at least rough market test of the relative efficacy of peer production, in this case, of accreditation. Yahoo's great innovation, when they started out, was, unlike AltaVista and like [?] all of those old search engines, we have people, sitting and looking at web sites and saying, we have paid employees looking at web sites and saying these are worth your time. So here is what Yahoo does in terms of law journals and Internet law. On the other hand, here's the open directory project, it's a... the list keeps going on. Open Directory Project, sixty thousand volunteers basically doing that, looking at web sites, saying here's the category they fit and they're worthwhile reading. Again, this is now being adopted by a good number of the major search engines as their index based searching algorithm. So we are seeing systematically the adoption of the products of peer production as competing head to head with market based production, with price based production. And not always, but with the sufficient regularity, coming up with outputs that are either measurably or, at least, one can appreciate them, as being better. With regard to material sharing, I'll spend less time on this, accepting questions. We see this happening in computation, the various "@home" projects, if you've heard of SETI@home or Folding@home. These are situations where, in the case of SETI@home, four and a half million users download a screen saver. When they are not using their computer, the screen saver contacts SETI@home, downloads a little piece of computation problem that's intended to scan radio astronomy signals from the SETI program, and then they compute to see if they can find signals and send it back up. Next time that they are free, they call for another piece of computation and send it back up. Just to give you a little bit of a sense a few weeks ago there was a huge announcement that IBM produced BlueGene which erased the shame that American super computer manufactures felt for a few years when the Japanese NEC Earth Simulator for the first time overtook the other major American super computers. And this is a very rough measure. It's problematic because SETI@home is not usable for everything but it makes for a good graphic. This is the difference between IBM, BlueGene, and the Earth Simulator in terms of teraflops per second, [?] floating point operations. This is what during the same time SETI@home produces, from pulling together the left over computer circles of people who aren't using their laptops and using them as a screen saver. Equally interesting we see the emergence of open wireless networks. We're beginning to see it a little bit now with WiFi, more than that we see very interesting designs for mesh networks. The basic point here is instead of having wireless networks that rely on a licensee, who puts up a broadcast transmitter or cell towers and controls and prices for usage, what you see is people buying laptops and computers with unlicensed wireless capabilities and then implementing in software the ability to help each other transmit messages. So that essentially my laptop is a user when I need to connect and a cell tower for my neighbor, when I don't, handing off networks. Again, we have excess communication capacity because we don't communicate continuously and we use it to help each other build the last mile - some very interesting potential applications to solve the last hundred meters, or so, of basic transport capacity. Critical here, in terms of the analogy, is it's done effectively through sharing networks. I'll keep mine own in return for you keeping yours on and we'll share. Storage, just imagine, that six years ago, someone would say "I would like you to build me a data storage system", that would be available to tens of millions of users around the world, twenty four hours a day, seven days a week, that would be robust and sustainable to the point where even if you took the central file, the central index down if you send soldiers down, they broke down the door and they shot down the server, it would still... The data would still be there and within a few months or some period we could get the data back and continue working, you would say "That is a dream! You will never build such a thing! Or if you build it it will take millions of millions of millions of dollars and there would only be one in the world", etc. That's P2P music sharing systems, right? It's a radically distributed data storage system, whereby users take parts of their hard drives, in excess, and share it to create a universal data storage and survivable date storage system. Freenet is the example of trying to use the same mechanism to do it for a publication system as opposed to a music distribution system. But it is a fascinating fact, that what we've seen is the development of an immensely robust and available data storage system. Globally available through shared excess resources that people use in interactions through social sharing. Basically exchange of: "I'll give you your files if you give me yours", without the mediation of explicit pricing. Finally we see combined mixes. Skype is very interesting Voice over IP application, Kazaa architecture and actually started by the people who started Kazaa. Fantastic voice over IP application, but again, instead of building a quality of service pricing of packets into the network, what you see instead is users downloading a piece of software. It looks like this, and as you can see, not everybody I'm talking to now is available, and very, very high quality relative to the switched network, let alone to the other voice over IP frameworks. Again, the same thing, excess capacity among users. So, what's common to all of these, these are widely distribute physical goods with some degree of capacity, they have excess capacity, of transmission capacity, of storage, of processing, they are privately owned by many distinct users and these users pull their excess capacity, not through a price system, not through a secondary market, in fact there are a couple of organization they are trying to pay people for doing the same thing that SETI@home does, and they've been very small and not particularly successful. So they are actual [?] performance networks, capacity calibration networks, and they haven't taken off, even though they are very explicit about "we will pay you x cents for so many hours of computation". And these people pull, as I said, their excess capacity, neither with managerial commands, nor with prices. So this leaves us with three economic questions: 1) what are the motivations, why do people do it, 2) what are the feasibility conditions and 3) is it more efficient and if so, when? The open source software economics literature has been the primary place where this has been studied and it maps the diverse appropriation mechanisms that the participant use to participate in free software, obviously because this is the broadest phenomenon and the one that's seen most often, it has attracted, and the one that has the most dollars attached to it, it is the one that has attracted the most attention from economists, so Josh Lerner and Jean Tirole, Steve Weber, Eric von Hippel, have done some very, very interesting work. And basically they mapped as their intrinsic and extrinsic rewards, things that are both the fun of it, [hedonic gains], people who actually value being part of the community, and it is something that is worth their while for purposes of participating in the community. There are extrinsic motivations, both on the supply side, in the sense that if I participate in that I'm investing in my own human capital, I know now how to build this systems. Tomorrow, someone is going to want to do an implementation, and they will hire me, not for my intellectual property, but for my human knowledge as somebody who can implement it in the company. As well as reputation, someone who is looking for a good computer programmer sees that I developed that particular component of this particular program and therefore hires me on a service model. Remember, IBM has more revenues from Linux-related services then it does from patent revenues. So, it is important to understand as base line for all of this applications, that while they are not based on prices it doesn't mean that there are no ways for participants to make money related to their participation, and this is a classic case, it's is a situation where, either because I've learned how to do this by participating, I'm now valuable, not to sell what I produce, but as a service provider in a relationship, or I now have better reputation, as well as on the demand side, this is very much the situation. Both of these are very much the situation IBM and HP. I can have better service contacts because I run better systems. The systems are better because the software is better or more robust, people want more of my servers, now I can sell more now that I know how to do it using a better piece of software. Now, the fact of diverse motivations explains why this is sustainable. It doesn't yet tell us about why it is that people may or may not join a project depending on whether it has a price or it's done for social motivation. This comes out of a much older debate on Motivation Crowding Out Theory, that starts around 1970 in an argument between Richard Titmuss and Kenneth Arrow. Titmuss wrote a book about blood donation in the United Kingdom and in the United States, where he basically showed facts that were clear and interpretations that were disputed. The facts that were clear was that Britain, which had an all donation system and no money, had less hepatitis in its blood, didn't have as many shortages and therefore produced better blood for the needs of the British health system then the United States, which at the time very heavily relied on blood purchases. Titmuss's theory was that people won't donate if you can buy the blood and therefore all the people who would've donated will stop coming and the people for the wrong motivations are donating. So, the background fact is, following this debate, the U.S. system did in fact move to an all volunteer, all donation system and that's is in fact how it runs to this day. Kenneth Arrow argued and said "You haven't proven this. What you haven't proved is that it's possible that the people who donate, donate in United States anyway, and the people who donate, donate in Britain anyway, irrespective of whether there is a price, because the people who donate are motivated for social reasons and people who pay, pay for a price, so if you remove the price, all the people who are selling, even though their blood isn't particularly valuable or is potentially tainted, will stop donating, and so, on average you'll get a better blood supply. Which is plausible, so the question was empirical, there's no model. So the person who's done probably most work in this field among economists is Bruno Frey at Zurich, and his work has focused on a psychological model developed by Edward [?] and others. And it focuses on the way in which people are driven by intrinsic and extrinsic motivations, and the basic model there is people feel rejected when you offer them money for something that they feel that they should be doing as a decent human being, and so what ends up happening or when you threaten to punish them for doing something, or not doing something, that they understand that decent human beings do for it each other. The basic model intuitively can be understood as, "what do you think? I'm not a decent human being? Why are you offering money to do this?" And that there is a displacement between this. It doesn't mean that offering money doesn't in fact increase an activity, it just means that it changes the social meaning of that same activity, so the people who are focusing within the social framework don't rely on it. There is another model by Roland Benabou and Jean Tirole, and the basic idea is that the people take cues from others who are in authority, when they're offered fine grained monetary incentives and wondering... They have a sense, "Ok, so this person, who is in authority, doesn't think I'm doing things properly. I trust them, therefore I am probably under performing and I continue to under perform". These are the theories. What I find to be more persuasive are the empirical studies in this field. There is a very good survey by Frey and Jegen, from 2001. Over the past almost ten years now there have been a series of surveys, one survey of managers regarding the efficacy of incentive contracts, people basically saying, and this is actually a finding, Osterloh and Frey found the same thing, that when you find something that is particularly difficult to measure, it is very hard to price and contract for. And so what you get within firms is that when someone has a lot of know how and there is a question of how to get them to share it, if you create incentive contracts, and basically tell them "I'll pay you a little bit more if you share, if you transmit your knowledge", you get actually less compliance then when you build a sense of team and combination and social relations where people feel the success of this other people is also my success, as opposed to my success is in transmitting this. And people end up producing better, rather than worse, knowledge transfer within these frameworks. A number of case studies showing that communities that don't want undesirable regulations, like a nuclear power plant in the vicinity, etc, respond better to arguments about contributing to the common good then to tax breaks in the form of money. The interesting study about parents picking children up. Parents paying a fine when they don't pick their kids up from the kindergarten on time and as it turns out when they pay a fine they just "oh, it's ok, I'm just paying now a baby sitter" and they stop coming or they come later. So, although some might say that just means the fine wasn't high enough. So that's the theory, that's one model of a theory which basically says when we have social motivations and money is introduced into the transaction, it reduces the social motivations, although it might raise, obviously, the monetary motivations. There is another set of literatures about social exchange and social capital, which is... For example, take peer to peer networks. The sharing there is instrumental, right? I want the songs, you want the songs. There's a very instrumental component, it's not just about feeling good about participating in the community or having a good sense of yourself. The anthropology of gift from Marcel Mauss and Malinowski on has focused very heavily on the fact that gift exchange cultures are very instrumental. They have rules in terms of marking, who is where in the hierarchy, and what the relationships are. These are not about, although there it's also mixed a lot with mystical and religious components that do map quite nicely onto the psychological explanations about the intrinsic side. The social capital literature, James Coleman, Mark Granovetter, and others, focuses very much on instrumentalism. "What job you can get depends more on who you know than on what you know" is a classic statement, and of course people schmooze with each other so they could become friends, so that, when a job opens up they can hear about it, that's very instrumental. So we have social transactional frameworks in a multiple literatures that are instrumental as opposed to psychologically driven by an intrinsic sense of self as well as some empirical work on this. For our purposes here though we don't really need to nail down which is the one single theory, partly because I don't think that... I think each of them is looking, each of this lines of literature is looking at different parts of the same phenomenon, because what we care about is, what is driving people within a social transaction network overall, as opposed to the particular individual levers. And the basic point is, human beings are diversely motivated. They have a reward function that includes material motivations that can be expressed in money. Social and psychological motivations which can be both instrumental and non-instrumental, and, and this is important, the different motivators have a complex relation to each other. They're not necessarily additive, but they're also not necessarily subtractive. And the easiest way of thinking about this is you go to dinner with friends, if at the end of dinner you leave some amount of money on the table and leave, you wont be invited again, right? The activity [?] you won't become a more desirable guest, although if you bring a gift to begin with, you might be, because there's a certain sense of exchange and, as I always like to say, if dinner is not entirely obvious, sex usually is. Now, these are obviously culturally contingent and cross culturally diverse, what counts as what a respectable person, what somebody who's a peer and a friend, or an acquaintance, does or doesn't do, is and isn't expected to do within a framework for what's paid for, changes across cultures, somethings may be easier to do in one place than in another, and this offers one way of seeing to what extent a particular culture is or isn't amenable to these kinds of mechanisms. When we look, for example, at the @home project, SETI@home, Folding@home and all of this, we see that they actually are working within a model of motivation that actually suggests a very broad motivational range outside of money. So, for example, one of the things that they do is they provide motivational feedback, information about the amount of work that each user contributes to the collective efforts, information about the scientific context, so some of them for example have interfaces that tell you what you doing. The most interesting one actually is this one: ClimatePrediction.Net, has people participating running a model of different CO2 levels over time and how they change the world. What's interesting and unique about ClimatePrediction.Net, is that you download an entire run of a model, and the screen saver that runs is the world that you are predicting, on your screen is what is represented, so you actually can see your particular contribution and then you can go to the web site and compare to the others. So its very well rendered to make the lay participant able to say "oh, this is what the model that I'm running is doing, and here is how it maps on base line model on others". Others like Folding@home are usually pretty pictures with more or less connection to the actual model. As I said, the commercial efforts of not being particularly successful, the reason that I have this up here is to show you the difference in terms of what it is that is rendered, right? Here is how much money you've made these month, here are your earnings over your life time, right? These have not really taken off too well. Papers published using the process are also shown, there are competitions for who provides the most cycle, including teams like national teams, or universities or, Linux users is one team on Folding@home for example, and online forms for discussing the topic of the project. So, what you see is there's no single motivational theory driving the people who are designing this project, instead we see what anthropologists called agonistic giving, just to say, "I give it to you to show that I'm more important than generous". This, you see, in the user of the week, month, year, galaxy, the greatest user of all time. We also see, however, a lot of non-agonistic giving in the assumption, right? Give this to help humanity, and in surveys of SETI@home people, a large chunk of people basically choose help humanity rather than have the most user of the week signs. We see both individualistic systems, assumptions, and [solidaristic] assumptions, and when we look at [solidarism], we see both old, like as I say, Alliance Francophone is one group in Folding@home, SETI@home allows people to do it by nation, but also synthetic, there's one very successful group in Folding@home called "The Knights who say Ni" so, a lot of new fangled stuff as well. Whether reciprocity plays a role or not, less so in computation, it's not at all clear, but it's very obvious in peer-to-peer networks that reciprocity is central to the process. And Kazaa, for example, very interesting here in terms of explicit pricing and monitoring versus implicit and broad gage, if you compare what Mojo Nation did with Gnutella around 2001, 2002, was an attempt to actually create coins of a sort, called Mojo, and the more you contributed, the more of this you got, and then explicitly you could get more resources of their network, and they tried to create an explicit trade system abstracting to units of currency, so that the more you contributed songs and bandwidth, the more you can download. That didn't work, that didn't work. What did work, and what is working for Kazaa, is something much, much more subtle, which is to say, if you contribute more and if the files are not corrupt, you then have, not even in your first search, but you have deeper search capabilities into the system. So the system rewards you in much vaguer and coarse grained rewards for being a good citizen, and that seems to have worked better. And that's consistent with this notion of social motivations being undermined when you make the rewards too explicit. So then lets talk about what are the feasibility conditions for this to happen and here is where the biggest difference is between the two kinds of uses. So, let me be careful about distinguishing human creative labor from material resources. The things that make it feasible for human creative labor to be more efficiently produced in some cases is one that is highly variable, different people are creative in different ways and even within, and this is very important, even within individuals there are some times of day, week, month, year, lifetime, where we are more or less available to different kinds of creative activity, but our jobs usually are structured along some dimension of our creativity, and assume a certain lifetime creativity. So, a lot of the things and times when we might or might not be more creative or left relative under utilized, and this is very important. Creativity is also personal, and specific, and non fungible. It's intrinsically available to individuals, but it's not something that you can fully specify and say "here, let me spend a month, writing a twenty page formal brochure", hand it to you and you'll have all my creativity. It just doesn't work that way. And anybody who's tried within a company to do knowledge management and formalize knowledge transfer knows that transferring human know how and creativity is basically not achievable, you gonna achieve some intermediate [?]. So this is another reason, and this is a reason why prices and managerial commands are relatively weak in this area. The other thing that's enabling is network connectivity of many individuals to diverse projects, at diverse times or rather all the time. So, if you have human beings, whose availability and value for a particular question at a particular time, varies in short and long time scales, if you have millions of people connected to millions of problems, the right person at the right time for right problem will be available. If it's very hard to reach a problem and work with it, and I'll show this graphically in a little bit when I talk about efficiency, the harder it is. The other thing that's developed, and this is what I tried to focus on when I talked about Mars Click Workers, but its true of all of this, is that absolutely critical for pulling diversely motivated contributions is modularization, that is to say, the taking of a big task, the breaking it up to small discrete components of variable magnitude, so that the person who has five minutes at ten o'clock at night can contribute something, and there might be tens of thousands of those, and the person who can contribute a whole work day because IBM is paying him to do it, can be effective, and the people who are sort of hanging out in the university, and have long nights where they can do things, are somewhere in the middle of a few hundreds, and pull together, all of this can be pulled together into a single product, so you have diversely motivated individuals with lots of projects available to millions of people. If they are modularized, so that you don't have to define the motivation and the output in advance, in the way that you need with a firm process, with a company process. But in fact you can just let, just as I put on my web site certain things and others do on theirs, if every person can define for themselves "I care about this enough to spend the down time of my computer", or in this particular case, the five minutes to do Mars Click Workers or the two hours a week to maintain a Wikipedia definition. If you modularize the task, that's important. Now, this is important because it means that different kinds of information goods will be more or less easy to do in this form. An encyclopedia with discrete definitions can be modularized. A novel probably can't. I think a documentary film, for example, would be easier to do in this form with found footage, than something that would be recognizable as a mass audience movie. So this starts to let us chew in to where are the domains where this is most potentially effective; and obviously you need an integration platform. There are different enabling characteristics for the material stuff, for the sharable goods. First of all, they are lumpy, this is a technical term, lumpy, which is to say, given a particular period in the technology of fabrication, of the making of things, you can build discrete packages, right? We don't today, it's not cost effective today, to have one huge supercomputer, where all of us get computation on demand. Instead, we can have a slower PC, a faster PC, a super fast PC, and by slower we mean something that five years ago was blastingly amazingly fast. But that's just the way processors are going, as well as storage, etc. If I want any computation, I need to buy at least a slow processor, and if I'm a medium user, that means I've lots of excess computation capacity. So, I bought it from my own personal reasons, and that's the notion of mid grained, and this changes from economy to economy, right? Mid grained is that unit that is produced because of the technical model of fabrication, will it, will there be one to a thousand, one to a hundred, the more that the size of the unit that people will buy giving their own needs, is something that given the level of wealth in a society, and the demand for the functionality is a wildly distributed good, like PCs are for example in the United States, like mobile phones are in many other places, as soon as you get to the point where there's a large scale population, that's got these goods, they become shareable. By which I mean, we bought them as consumption goods, for our own personal use. But they have, attached on to them, capital capacity that we can now think about how it gets redistributed, or reused. We have large amounts of excess physical capacity, that is widely distributed in the population, in small chunks. What this does is, it doesn't mean that you can't use a secondary market for this, of course you can. But because I already bought this particular piece of capital equipment, based on my own consumption needs, I can now use the excess capacity for any one of the diverse human motivations that I can use, I might sell it off but I can also decide to share it. In a way that if this were a steam engine, I simply couldn't afford to do. But this is a large-scale capital good you could only aggregate the demand for its capacity through a market or a firm. It is the fact that the goods are small scale, and lumpy, that makes them available in the population for individual or house hold use and therefore available for sharing along side secondary markets and firms. So this is very interesting, because we come at a time when the most important inputs into the core economic activities in the economy, in the network information economy, in the most advanced economies around the world are widely distributed in the population, human creativity and computation and communications resources. So, if we think of this as a moment of great transformation in the social organizational of production, particularly in the advanced and relatively wealthy world where access to computation and communications resources is wide spread, we are seeing a genuinely foundational shift in what is scarce, what is scarce is human creativity, and a will to act. So this leaves us with the last question of "Is it efficient?" The first question I think should have already been answered and that is the question of incentives, right? Is it efficient? Will people have the right incentives? The answer here is that when every motivation crowding out decreases the activity, it will be efficient to use social resources. What can we say about when that's good? And this is very interesting for why it works for distributed computing, for distributed storage and for distributed wireless communications. If the capacity is distributed in very small quanta across a wide population. If in order to get the whole function you need, you need to collect thousands of little contributions, then the amount of money you can pay for each contribution is very small, but you've already made a market transaction out of a social transaction. So the crowding out effect kicks in, while the incentive effective of money is relative low because of the small amount of money. So that's a domain where we shouldn't expect there to be less incentive to sell for very small amounts than to give away for some decent social rewards. The incentives in general are a trivial question when you talk about large scale volunteer projects, because as I said, if you modularize it correctly, then the five minute increments of the volunteers, the two hour increments of the students, the full day of the people who can actually do demand side and supply side instrumental rewards can sustain the process. Most of the answer with regard to efficiency comes from transaction costs economics, and mostly falls into information gains and allocation gains. If you think, and this is a very... I'll go through this very quickly, there won't be a test at the end. If you think of... If you take the transaction costs theory of Ronald Coase and the property theory of Harold Demsetz, and map them on to each other, what you can basically see, is that any agent who has a particular resource stands at a set of choice points over time, that map all of this transactional frameworks. So if you think of an agent or resource down here, the first thing the agent needs to see, should I exclude people from this? Kick them out? Or should I not? And the first answer is how expensive is it to keep people out, versus how much benefit can I get from keeping them out? So, for example, woodlands in places that no one... That is easy to get access to a lot of land is not a place where you would exclude people. You wouldn't put up fences and put no trespass signs for one person coming every three months, who, if you ask them to pay, will go around to your neighbor anyway. So, you get no exclusion. Then when you get to exclusion, you have to choose, do you perfectly exclude, do you say nobody should come in to my thing at all. Or partial exclusion, some people can come in, and again it's the same analysis: how expensive is it to keep people out, how expensive is it to let them in, and I can go into this later, but there is both anthropology here, and transaction cost economics, that suggest that for somethings we'll say I just don't want anybody in here. My car, I won't stop to start transacting with people along the way, or maybe I will, we will see, hitchhikers, car pulling, etc. When I decide on partial exclusion, I can decide whether it's selective or non selective, non selective partial exclusion means I just know that somebody wants to use my things, not more, and if I'm better off in some way by giving them access, for example, because they will do same, then I'll use that. So the Internet Protocol is a first come first served model, peer to peer networks, clients or first come first served. I also did work... I won't talk now about body snatchers and slugs. This is a particular kind of car pulling, car sharing on the way to work in Washington D.C. and San Francisco that exhibits very similar characteristics. If I've decided that I can benefit more by being selective about who can come in, I still have the choice of whether I'll do it through the market, or through a social relations network, and again, it's the same analysis that's been going all the way along. Is it more expensive for me to charge money, relative to how much money I think I'm gonna make, as opposed to simply say, well I won't make that much money, but if I give this away to my neighbors, there will be all of social benefits, and it's cheaper to do. So, again, it is a transaction costs analysis between these, and this is where Ronald Coase's first work on transaction cost economics, which earned him the Nobel prize, was first done in the nature of the firm, which is the second half also of my first piece on this area, called "Coases's Penguin, or Linux and the Nature of the Firm". But the same model that chooses between firms and markets really chooses all the way up and, and this is most important for us, chooses whether you'll go into the market system at all, or whether you'll go into a social system that could either be, again, similar to firms and markets in decentralized peer production, or in non-profit associations, organizations that are within a social transactional network. Let's see... We talked already about the fact that human creativity is very diverse and hard to specify. Markets and hierarchies need crisp definitions of what exactly it is that you need to do, how much exactly am I going to pay, and the other way around. And they need relatively structured data about what's been done, how it's been done, what's been paid, how it's been paid, so there is relatively high transaction costs, even in efficient systems, simply in specifying very clearly what's been done, and by whom, and what sort of reward. Social sharing in exchange systems on the other hand are much looser, they are much more, nobody says, I will water your plants this week if you help me carry three boxes next July two stores. We just don't do that, right? What we do is, we do things which all of us understand create some set of mutual obligations in a general vague cloud of obligations, but we don't need to specify on a per-transaction basis. So, social systems have similarly high set up costs, right? Just like we need property law and markets, etc, for markets, we also need social sharing systems and trust and systems of mutual dependence that we setup. So there are high setup costs, fixed costs for everything. But the marginal transaction costs through the social system is lower. It's also the case that centralized systems lose information relative to decentralized system, but gain some degree of controlling. Centralized and decentralized, here I mean something very specific, systems where the sensing of an opportunity for action and the action are... A decentralized system is where the sensing of an opportunity for action, the authority to decide on action and the action itself, are very close together. A centralized system is one where the authority to decide on the action is far from the space where opportunities for action are sensed and can be acted upon, right? And the information gets lost in the transmission to decentralized control system. So, there is a trade off in this system between information and control. Information is lost in the transmission from the locus of opportunity for action to the locus of authority and back. So centralized systems will in a sense be more information poor but more controlled. Decentralized systems will be information rich and less controlled, though not necessarily less dynamically stable and efficient. How does this work out as a matter of allocation gains, and this is something where I want to use a simple graphic and I'll leave the option theory, the option part of it aside. If you just imagine a situation where we have two companies, A and B. Company A has a certain number of agents A, and a certain number of resources at its disposal. You can think of it as computer programs or existing story lines in a film studio, or something like that. And company A will choose the best person who works for company A to work on resources one and four, given the talent pool within the company. But they're not gonna go out and search the entire universe, it could be that there is an A8 somewhere in company B, that could actually do fantastic things with R1 and R2, but it's not worth the transaction costs of finding who that A8 is, getting them transfered and working, and then sharing the benefits between the two companies. In a peer production community, A8 self identifies and has the information that she can work better on these things, then anyone else, and tries to. And then, exposed to [?] community says "Actually, hey, what you produced is great! We'll take that". But the information about the opportunity for action is lost to transaction costs when you separate resources and agents into different pools. And one of the things you get from peer production is enormous gains in information about who is the best person for the best role. There is another way of looking at this, but I won't focus on it at the moment, because I want to leave us some time for questions. The role of culture: it's important to recognize that the technological feasibility conditions that I talked about are threshold barriers to the various modalities of production, right? If your economy has to run on steam engines and large scale production plants, you will not get social production of cars. You will not get social production of refrigerators. Because that is the technology of production that simply makes it too expensive for people to do this. Now, it is true that are worker-owned firms and its true there's a literature on common property regimes for dams and irrigation systems. But these are the exceptional, and they end up looking very often, very similar to firms or quasi-governmental organizations. But beyond the threshold viability, similar activities can be undertaken in different modalities, depending on the extent which a particular culture is or isn't amenable to a particular community. Its institutions, the level of trust in it, the extent to which people are or are not sensitive to monetary incentives, the extent to which they actually see cooperation as something that is easy or hard. And it is possible, though I haven't studied it, and I haven't seen a good study, that there's some recurring iterative pattern where if the more you participate in this and gain confidence that they work, the more you're willing to contribute. It's a hypothesis, I have not yet done the work to show that it's correct. So, overall we can say that, under certain technological circumstances, it becomes practically feasible to execute production problems through a class of solutions that rely on social sharing and exchange, rather than on the price system or on firm hierarchy. These are typified by decentralization of the authority and the capacity to contribute to effective action. They rely on social, information flows which are vaguer than market or firm. Information flows, social motivation structures, rather than prices or commands to motivate and direct action. If we compare this to the literature of the gift economies, the gift literature focuses on the production of social relations through the exchange of things, and peer production focuses on the production of [?] things we value materially, through social relations, it's the same incentive universe, the social norms literature focuses on the production of behavioral constraints and institutions through social relations, but the domain of operation is usually in markets and firms, as opposed to actually providing both the motivation and the organization signally. The social capital literature focuses primarily on the degree of increased productivity in the market or in the firm, from the fact people have more social capital. Ok? So, again, the motivations and the information comes mostly in the market, the social capital eases the market, lowers transaction costs, makes things more efficient. And as I said, the common property regimes or club good literature are typified by crisp distinctions between who is in and who is out. They're not that open. They rely on stable repeat player relationships. This [?] of social production or peer production operates in the domain of the production of things in services that are valued materially, not only or primarily the production of norms or social relations. The inputs and outputs can be private economic goods, not only public goods, and not only club goods, or common pool resources. They can even occur among strangers as we are seeing in the examples that I talked about on the net, not only among intimates and long standing relationships. So, the potential for these to be a very broad phenomena as opposed to a discrete subcategory is quite interesting. There is a funny characteristic of the efficiency of these. When you plot these kinds of social relationships, like car pulling on the way to work with co-workers, or free software. If you map a three dimensional map, of whether the level of acquaintance, familiarity, whether it's really intimate relations as opposed to general human connectedness along the Y axis, along the X axis, whether it's a multi-purpose relationship or a special purpose relationship, and whether it's decentralized or centralized. You can map a lot of things that we see in the social capital literature, in the social norms literature, along various areas and, depending on the culture, family can be more or less centralized, but generally is high on the intimacy, etc. What is interesting about free software is that it's very close to the origin. It's various special purpose, very decentralized, and much closer to general human connectedness. And the reason that this is interesting is that although it's a social interaction, it's a relatively fungible and easy in-out one. So, it is the one that is least likely to exert, in a market sense, inefficient pulls. Where I stay in this not because I'm the best person to do this, or I have time now, etc, but only because I feel obligation to stay. Which, from the perspective of a theory of communitarianism is bad. >From the perspective of a market, in fluid labor, creative labor, to new things or fluid capacity to new things it's actually very good. So it is an attractive characteristic from the production perspective of this particular category of uses. So essentially we can say that we have four transactional frameworks, we have the price system, we have firm hierarchies, we have government regulation, and we have social sharing exchange. Many different kinds of activities can be undertaken under different modalities, which one will be more efficient in any given context, is a function of the technology of production, the availability of opportunities, the relative information gains, and basically overall we can say, of the transaction costs and the benefits of participating in each one. Obviously where motivation is resistant to sharing claims, the price system will be more effective. For example, one of the things that I've been working on is using sharing for purposes of critical infrastructure. So if banks want to have bank records, backup storage, it is going to be very hard to get people to contribute it in a social sharing network, as their backup. But not so hard for banks to say, "here is a few dollars, or a few cents, or a few whatever it is, out of what, [?] pay me, if you devote some of your hard drive to backup of other people, so that I have a diverse backup system". Firm hierarchy is particularly good, I think, where the excess capacity is internal, that is a [?] for enterprise systems on the distributed model. Government regulation may or may not be a residual, social sharing and exchanges are particularly valuable, where lots of smalls contributions are required. Particularly easy where instrumental exchange is possible, I can open wireless networks, distributed storage. Processing, actually, is a little harder. SETI@home is a little harder to do instrumentally, and you see all of the work there on trying to generate some identity type frameworks. This leaves us with the commons problem, which I will go through very quickly. The basic point is: we think we know how to use property to organize and coordinate behaviors. Different kinds of commons have different kinds of solutions. Information is only a provisioning problem. It is only a question of what are the incentives to produce, it's not on allocation, because it is a non-rival public good. Once the information is out there, there is no problem with over consumption. If there is a problem, there is only a problem of under production on the provisioning side. Shareable goods do present an allocation problem, right? The excess computer cycles could go to this problem, or that problem. The excess capacity of communication can go to this neighbor, or that network, or the other network. The solutions online to these have been technical, with proposals for using law to get some regulatory gain in open wireless networks, and that is something that I'm doing a lot of work on, with the Federal Communications Commission in the United States. Because our participation, that I just described this notion that you get deeper search capabilities, if you spend the time to make sure that your files are available for longer and are clean, is another way of solving this. Basically if we think of levers used in commons, we see formal rules, some of then can be contractual, like the GPL or Creative Commons, some can be organizational, like standards and the [?] of WiFi. We see technological mechanisms, which fall largely into clear constraints on behavior, and new affordances, or new abilities to behave in ways that facilitate social coordination, so Slash moderation, the fact that you can't post more than a certain number of times in a minute or an hour, the fact you can't moderate other people's comments more then five times in a 36 hour period. Those are constraints. Wireless standards, say collision control, are hard technical constraints intended to make it difficult for people to do things that would undermine the total collective action. Affordances are in a sense more interesting, or equally interesting. The transparency of Wiki, that I showed you earlier, the fact that you can see what people have done that is very easy, and is very easy to change is a kind of an affordance that lets the social relations kick in. Moderation on Slash, the fact you can actually say this is interesting, this is not interesting. Collaborative filtering, that is to say, what I'm interested in, and you're interested in, we can rely on each other's judgment. This is another technical affordance that lets people decide together what's relevant and credible to them. Reputation systems, like troll filters, in many of this moderated systems, but also what we see on eBay. These are technical mechanisms, the reason that I see them as different. The technical mechanisms, not to prevent you from acting in a bad way, but to enable the group of users to use its own social mechanisms to police each other and to achieve coordination. And obviously we also have just plain redundancy. The fact that lots and lots of people are doing the same thing, means that if one, or two, or three fail, there is still someone who's concluded it. Click Workers, BitTorrent are examples. Social mechanisms, social norms as behavioral constraints or guide. So Wikipedia has a whole site on Objectivity. We are writing an encyclopedia, it's not just opinion. So there are social norms and people say: do I join this group or not? Can I live with this or not? And so social norms actually play a role in making sure that people behavior ways that are coordinated. But they can also include discussion platforms for airing differences, "projectlets", Wikipedia "talk", blog comments, so, mechanisms that allow people to engage in the social mechanisms. [All the way] opt to quasi-formal mediation or arbitration, even on Wikipedia. There is a mechanism where people, when they really disagree about the definition, they can go to community mediation. Here is my definition, here is yours. Let's find a formal mechanism of deciding this. So there are social mechanisms that can be embedded in this systems. So, in general we can talk about... Again, this is one of those small letters, that... There won't be a test at the end. Transactional frameworks, social sharing firms, price system government, or just parallel existence, like we have on the web. On this axis, levers of constraint and affordance. Contracts, norms, social norms, technology, and law. And what we see here as we are trying to think about designing this kinds of platforms, like the Incubadora, for example, is that you see, you can in principle think of purely technical... You're disagreeing that that is what it is, lets talk about that. Purely technical... I don't know, that's why I said I came to learn. Purely technical coexistence is one mechanism, the more interesting thing is that the development of this concept of social software, software that understands that it is facilitating a social relationship, and that it will only succeed, not if it is the best technical implementation of a particular use, but if it's the one that will best harness the social relations of the users, in a sense Clay Shirky describes as "the group is the user, not the individual." And the group can be self sustaining through social norms, maybe through law and contracts like the GPL, or the Creative Commons, how these together pull in order to make the group most sustainable without relying on the market. So, to wrap up, we're at a point where technological threshold conditions enable individual human agency, human action. They increase the domain of effective action for diverse motivations, without focusing us in the productive arena on purely instrumental monetary reasons for action. Peer production is focused on human creativity, resource sharing from computation and communications resources, for different technical reasons, is also now available for social sharing and exchange. So what we are seeing is the emergence of social sharing and exchange, of peer production of human creativity of information, knowledge and culture, emerging as a substantial modality of economic production in the digital network environment. Thank you.