From Serendip

Learning, Working & Playing in the Digital Age
John Seely Brown

The following is based on a tape transcription of a talk by John Seely Brown at the 1999 Conference on Higher Education of the American Association for Higher Education A pdf file of the tape transcription is available from the The National Teaching and Learning Forum. Text and figures of the html version provided here, with the permission of John Seely Brown, are the same as those in the available pdf version. To make it convenient for Serendip visitors, we have somewhat altered the layout of materials and added some accessing elements and links. Brown was introduced by Russell Edgerton, Director, Education Programs, The Pew Charitable Trusts. His introductory remarks are available here. Comments and thoughts on Learning, Working, and Playing in the Digital Age are welcomed in a Serendip forum area.

Thank you—but first a brief preamble about documents given what Russ just said. The etymology of the word "document" relates to ancient French docere, meaning to educate and to communicate. And "text" comes from texere, meaning to weave together, as in textile, to weave together fragments of thought, to communicate and to educate. So the document and text actually have a very rich grounding in the very essence of learning.

What I want to do this morning is to provide some evocative comments rather than give a coherent, logically argued talk. That is, these comments are meant to be idea sparkers that will, hopefully, evoke additional ideas for yourself concerning how the world might be changing and how we might actually recast or reframe some of the classical problems of education and distance learning in quite new terms. My interest here today is in looking at the notions of learning, working and playing in the digital age and how today's kids—growing up digital—might actually be quite different from what we might first think. But, more particularly, how by stepping back and looking at the forces and trends underlying the digital world, we may have a chance to create a new kind of learning matrix, one that I will call a learning ecology.

I became interested in learning ecologies because of their systemic properties. We need to view higher education from a systemic perspective, one that takes into consideration all of the components—k-12, community colleges, state and private colleges and universities, community libraries, firms, etc.—that make up a region. This, in turn, raises additional questions about how we might create a regional advantage such as in the Research Triangle in North Carolina or in Silicon Valley. For example, is there a way to extend science parks, that typically surround universities, into also being learning parks and from there into being learning ecologies by combining the knowledge producing components of the region with the nearly infinite reach and access to information that the internet provides? And, if so, might this provide an additional use of the internet in learning—one besides just distance learning. But first, let's consider what the Web is and see how it might provide a new kind of information fabric in which learning, working and playing co-mingle. Following that we will then look at the notion of distributed intelligence which has a great deal to do with the social basis as well as the cognitive basis of learning, and how those fold together. Then we will look at the issue of how one might better capture and leverage naturally occurring knowledge assets, a topic as relevant to the campus as to the region or to the firm. Finally, we will come to the core topic of how all this folds together to lead to a new concept of a learning ecology.

So let's look at the Web, or how it is evolving-rapidly evolving. It's important to realize that the Web really is a transformative infrastructure very much like electrification was for the United States at the turn of this century. Electrification changed nearly every aspect of how we lived, how we worked and how we learned. It changed the architecture of our factories, homes and cities and nearly all of our social practices. Consider the light bulb, radio and TV as just a few examples that transformed how we lived and learned.


It took 20 or 50 years for electrification to take hold and for society to enact new social practices that leveraged the potential of that infrastructure. And so it will be for the Web. But to see this I think it is crucial not to think of the Web and the internet as just a network of computers but rather as the beginning of a fundamentally new medium, a medium as in TV, radio, theater and books. But this medium is going to have properties that are going to be very hard for us to understand because it's going to be a two-way or interactive medium whereas nearly all our past media have been fundamentally one way. The fact that the Web is best thought of as a medium was manifest in a very shocking way yesterday, when Comcast offered about 50 billion dollars for Media One. Comcast is a cable provider. Why is it worth so many billions of dollars to them? I'll tell you why. They see the capability of transforming the Web and the internet into a fundamentally new broadband medium. What that medium will evolve into, believe me, none of us really know. But, we can be certain that the properties of this medium will evolve or, more accurately, co-evolve between the constantly evolving capabilities of the technology and social/market needs. This medium will have some very interesting properties and surprises. Recall that when movies were created, they imitated theater. And then over a period of 10 or 20 years they evolved their own genres, using new capabilities such as fades, dissolves, flashbacks, time and space folds, special effects, etc. All these were fundamentally new constructs that fostered new film genres that became radically different from theater.

What's very interesting to me is that in TV or in movies, genres seem to develop about every four or five years. Watch a movie that is about four or five years old and you think something is weird. It's out of kilter, a little bit. Well guess what? On the Web, genres appear to be evolving about every six months. If you see a Web site that's more than six months old, it screams out at you—something is wrong. This Web site is tired. The Web as media will be developing or enacting new genres in Web time—an interesting fact since genres are social enactments, not just technological enactments.

How we actually think about the shaping of those genres is going to be a very interesting issue, because the Web has both a broad, one-way reach like broadcast, but also way a two-way reciprocity like in mid-cast. How are you going to have two-way broadcasts or how do we combine broadcast concepts with narrow cast (1-1) concepts? Confusing, to say the least, but also mind-stretching.

The second aspect of the Web that has interested me for some time is the fact that the Web may be the first technology, the first medium that honors the notion of multiple intelligences. Our current concept of literacy grows out of our intense belief in text, a particular focus that probably resulted from the power and omnipresence of one particular technology—the typewriter. Over the last 50 to 100 years it has become the primary authoring tool, a tool uniquely suited for text but terrible for sketching, painting or even writing music or mathematics. Indeed, almost all our past technology for helping the authoring process has been aimed at one particular kind of intelligence. With the Web we have for the first time a medium that could truly honor multiple forms of intelligence—abstract, textual, visual, musical, social and kinesthetic. The kinesthetic is the last to be served. So now we have the chance to construct, with almost no extra work by the educational communities, a medium that enables a kid to become engaged in his or her ideal way of learning—enabling a good impedance match between the medium and how a particular kid thinks. (This capability may be especially important as a child starts his learning journey. Afterwards, and after a sense of self confidence about being able to learn has been established, mastering a broader set of learning media will be easier). We may be at an inflection point where visual, musical, spatial and kinesthetic intelligence can be as easily served by technology as abstract and textual intelligence. This opportunity is a discontinuity on how we might leverage technology for learning in the future.

Let's go back to the last point about the Web being a new kind of medium with both reach and reciprocity. One of the unusual things about the internet and the Web as a medium is that it enables us to leverage the small efforts of the many along with the large efforts of the few.

Two very simple examples: consider a project called Pueblo that is happening in the Longview School in Phoenix, Arizona, in conjunction with some researchers from Phoenix College, a part of the Maricopa Community College System. These researchers have found a way to use a closed internet to connect a set of senior citizens acting as mentors with kids in the school systems. The result was that the small efforts of the many—the senior citizens—beautifully complemented the large efforts of the few—the teachers—and as we all know, kids are often more willing to listen to grandparents than to parents...jumping a generation. This experiment reflects a win/win situation because the senior citizens wired together actually created a sense of meaning for themselves, through interaction with themselves and the kids, while also acting as a powerful resource to the kids.

A second example: Hewlett-Packard and the Web. In this example, engineers at Hewlett-Packard use the Web to act as cognitive apprentices, or mentors, for kids wanting extra help on scientific, engineering or mathematical type problems. Again, the small efforts of the many—the engineers—complement the large efforts of the few—the teachers. Both of these examples barely scratch the surface of what could result from interlacing the small efforts of the many with the large efforts of the few. They also suggest that there may be powerful ways to combine the various resources of a region—some as dedicated and some as peripheral—but ones that leverage the ongoing practices of the region.

What I really want to stress is that this medium is as transformative as was electrification and with similar diffusion properties. The industrial dynamo was introduced about 1880. It took about 30 or more years for the effects of the dynamo to permeate our society—that is, to have all the necessary complementary assets in place for it to be really useful for society—complementary assets like utility companies, wired neighborhoods, power stations, etc., for it to become sufficiently infrastructural in order for it to become transformative of the way we live, work and learn. Social practices don't change overnight, nor do capital assets.

My belief is that the Web/internet phenomenon is basically as fundamental to society as electrification and is subject to many of the same diffusion and absorption dynamics as electrification. We've actually had the internet as the ARPAnet around for just about 25 years. And it is now just starting to have a major impact on our lives. We're just at the bottom of the S-curve of this innovation, a curve that will have about the same shape but a greater slope than the one pertaining to electrification. And as this S-curve takes off, it creates a unique period for entrepreneurs! It is entrepreneurs, be they academic, educational or corporate entrepreneurs, that will shape and drive this relatively chaotic phenomenon especially as it relates to learning. Entrepreneurs are great at challenging the status quo. Their power lies in their willingness to see differently, unearth and challenge background assumptions and then act on their beliefs, often overturning an assumption that others felt were unassailable. Our challenge and opportunity, here, is to foster the entrepreneurial spirit toward creating new kinds of learning environments, ones that leverage how we naturally learn coupled to or enhanced by the unique capabilities of the Web.

So, having said all that, let's step back and ask about today's kids, kids growing up digital. How are their brains different? How do they learn differently? How do they think differently? How are they different? Because after all, today's kids are today's customers for schools and tomorrow's customers for lifelong learning. So we all have a lot of motivation to jointly come to some understanding of how the "digital kid" is different. Let me first give an overview and then I will dig into this topic a little bit more.

Starting about three years ago we, at PARC, started hiring fifteen year olds to join us during the summer as researchers—researchers that were given two tasks. Their first task was to design the workscape of the future, a workscape that they would want to work in. And the second, working with some California school superintendents in a summer intern program, was to design the school or learningscape of the future—again, one that they would want to learn in. With these projects we had an excellent opportunity to watch these kids. Again, this was not meant to be a training exercise or routine work. This was designed to be a serious research assignment, one that afforded us the opportunity to learn and one that actually shook me up when I began to see how they were thinking and what their concepts of ideal future workscapes and learningscapes were like.

Let me make one obvious comment. Speaking cognitively, today's kids are always multi-processing. That is to say they are always doing many things simultaneously—listening to music, viewing TV, talking on the cell phone, using the computer—all at the same time. They are, so to speak, multi-processors. Recently, I was with a young researcher, albeit one that was a bit unusual, that had actually wired a Web browser into his eyeglasses. As he was having a conversation with me he was actually bringing up my Web page to read about me. This was a bizarre experience, but except for watching his eye movements, I was pretty much unaware that while he was talking to me he was also reading about me. I was even unaware of how he was using his left hand in his pocket to cord in keystrokes to bring up my Web page, while seamlessly talking with me. I was completely blown away that he could do these things in parallel and relatively undetectable to me. It was a strange experience to say the least.

Many of us tend to think that kids who are multi-processing can't be concentrating. This may not be true. Notice that the attention span of most top managers range somewhere between 30 seconds to five minutes, which seems to be about the right span for most kids that I know. And then think about the fast context switching that all of us really do but don't want to admit, because most of us seldom spend more than five minutes concentrating on any one topic, at least during the daytime. So perhaps this notion of multi-processing so present in today's kids may, in fact, turn out to be ideal training for the corporate world. But let's go into this in a little more depth.

Let's look at a set of dimensions and consider the shifts along these dimensions that are now occurring for kids in and of the digital age. Although I discuss these dimensions as all separate, they actually fold in on each other creating a complex of intertwined skills comprising a kid's cognitive make-up.

The first dimensional shift has to do with literacy and how it is evolving. Today, literacy involves more than just text but also involves image and screen literacy. The ability to "read" multi-media texts and to be able to pick up and feel comfortable with these new rapidly evolving multiple media genres is decidedly non-trivial. In fact, we tend to think of watching a movie as requiring no particular skill. I'll tell you this, if you've been left out of society for about ten years and then come back and see a new movie, you'll find it to be a very jarring, a very confusing experience. I don't know how many of you watched "48 Hours" when they shifted their genre about eight years ago. It was jarring; it looked as if MTV had come to a traditional network. But now it seems old and dead again and they are, I believe, in the midst of doing a complete revamp of that genre.

What I want to suggest, though, is that the new literacy, the one beyond just text and image, is one of information navigation. I believe that the real literacy of tomorrow will have more to do with being able to be your own private, personal reference librarian, one that knows how to navigate through the incredible, confusing, complex information spaces and feel comfortable and located in doing that. So navigation will be a new form of literacy if not the main form of literacy for the 21st century.

The next dimension concerns learning and how that is shifting. Most of us experienced our formal learning in an authority-based, lecture-oriented school. And yet with the increasing amounts of information being readily available on the Web, we find a new kind of learning happening—it's not all that new; most of us did it informally anyway—having to do with discovery-based or experiential-based learning. We are constantly discovering new things as we browse and surf through these emerging digital "libraries," Indeed, Web surfing is becoming the new form of infotainment and may even become a prominent form of entertainment for the digital kid. But that by itself, even with this shift of the notion of literacy, is not too significant until you recognize a third and substantially more subtle shift, one that pertains to forms of reasoning.

Reasoning classically has been concerned primarily with deductive, abstract types of reasoning. But what I see happening to today's kids as they work in this new digital medium has much more to do with bricolage than abstract logic. Bricolage, a concept originally studied by Levi Strauss many years ago, relates to the concrete. It has to do with the ability to find something—an object, tool, piece of code, document—and to use it in a new way and in a new context. In fact, virtually no system today is built from scratch or first principles—like the way I used to build systems—but rather from finding examples of code on the Web, borrowing "that code," bringing it onto their site, and then modifying it to fit their needs. Today's systems are built up through an extensive sense of bricolage—by cobbling or "wiring" together code fragments and extending or modifying such fragments when necessary. The catch, however, is that if you are going to become a successful bricoleur of the 21st century, a bricoleur of the virtual rather than of the physical, than as you borrow things you have to be able to decide whether or not to believe or trust those things.

So in some interesting sense the need for making judgments is greater than ever. After all, who would necessarily believe something just because it was on the Web? If you found it in the Wall Street Journal you might have some reason to believe it, the National Enquirer, perhaps not. Because if you extract information from well-established subgenres such as certain newspapers, you can count on the authority earned by that genre and inherent in the institution(s) behind it. But on the Web, let's face it, there is virtually no institutional authority behind much of what you find. So in a way, if you want to be a successful bricoleur, you're going to have to be damn good at figuring out what to believe in as fact, not fantasy. Do you want to believe that essay, especially if you plan to clip part of it for a term paper? Do you trust that person's code sufficiently to build on it?

So we now have navigation being coupled to, basically, discovery and discovery being coupled to bricolage but you don't dare build on whatever you discover unless you can make a judgment concerning its quality or trustworthiness. Thus navigation, discovery, borrowing and judgment all get wrapped up together, especially when the student is engaged in using or building something that he deems important. Judgment is inherently more critical than ever in order to become an effective digital bricoleur.

But how do we make judgments? Do you do that socially in terms of recommendations of others you might trust? Do you do that cognitively based on rational argumentation? Do you do it based on the inherent warrants of the institution that might have sponsored it? What's the mixture of ways and warrants that you end up using to decide and act? With the net, not only is the need greater but so are the resources, but many of these resources are different than the non-digital adult is used to.

The final dimension has to do with action. One of the things that I find most interesting is how new systems get absorbed by society and how this absorption or learning process might be quite different for the digital kid than for us. My generation, speaking generally, tend not to want to try things unless we already know how to use them. If we don't know how to use some appliance, software or game, etc., then we tend to reach for a manual, ask for a training course or ask to be shown how to do it by an expert. Believe me, hand a manual to a 15-year-old or suggest going to a training course and he thinks you are a dinosaur. "A manual? Give me a break! Let me get in there and muck around and try various things and see what works." More generally, today's kids tend to get on the Web and link, lurk and watch how other people are doing things and then try something themselves.

Action then brings us back into the same loop in which navigation, experiential learning and judgment all come into play in situ. When, for example, have we lurked enough to try something ourselves? Once we fold situated action into the above story we shift our focus more toward learning in situ with and from each other. Learning becomes as much social as cognitive, as much concrete as abstract, and becomes intertwined with judgment and exploration. As such, the Web becomes not only an informational and social resource but it could also become a learning medium where understandings are socially constructed and shared. Said differently, learning becomes a part of action and knowledge creation. But to see how this all follows, let's take a brief detour into knowledge—its creation and sharing—seen both from the standard Cartesian position and from that of the bricoleur.

Knowledge has two dimensions, the explicit and the tacit. The explicit dimension deals with concepts, the know-whats, whereas the tacit dimension deals with know-how.

Know-how is best manifested in work practices and skills. Since the tacit lives in action it comes alive in and through doing things and in participation with each other and the world. As a consequence, tacit knowledge can be distributed between people in terms of a shared understanding that slowly emerges from working together, a point that we will return to.

Jerome Brunner made a brilliant observation some time ago when he said that we can teach people about a subject matter, for example, physics. That is, we can teach them the concepts, conceptual frameworks and facts of physics—the explicit knowledge of physics. But that does not make the student a physicist. To be a physicist he must also learn the practices of this profession. So learning to be a physicist as opposed to learning about physics requires growing (through situated learning) a column down the middle of the above diagram that starts to create a platform for the rich interplay between the tacit and the explicit. (This interplay is best characterized as "knowing" and it lives in the action of deliberate inquiry where the concepts, heuristics, laws and algorithms comprising the explicit function as tools for action–deliberate inquiry.) This, of course, is where real expertise lies. Learning this expertise requires learning the practices of deliberate inquiry of that discipline and how to best utilize the conceptual tools—the explicit—in support of that inquiry. And learning this involves a kind of immersion and enculturation—a way of seeing, interpreting and acting.

The epistemic landscape gets still a little bit more complicated because both the tacit and explicit dimensions of knowledge apply not only to the individual but also to communities of practice or the social mind. It is very easy for us to think that all knowledge is in the head, but when we start to consider the tacit dimension, especially as it relates to practices, we realize how you can know much more than the knowledge you actually have. Much of this knowing is brought forth in action, action through participation, participation with the world, participation with the problem and participation with other people, i.e., practices. A lot of the knowing comes into being through the practices comprising one's community(s) of practice.

Understanding how intelligence is distributed across a broader matrix becomes increasingly critical if we want to leverage learning to learn, because learning to learn happens most naturally from being situated and participating in a community of practice. Returning to Brunner's notion of learning-to-be, learning-to-be might be best thought about as enculturating into a community of practice, a community of physicians, lawyers, writers, readers, etc. Enculturation lies at the heart of learning. It also lies at the heart of knowing. Knowing has as much to do with picking up the genres of that particular sub-profession as it does with its conceptual framework. For example, how do you recognize whether a problem is an important problem, or a solution an elegant solution, or even what constitutes a solution in the first place?

We tend to think of theory as always being at the top of hierarchy of importance and practice as being at the bottom. But think about what you do when you get a Ph.D. The last two years of most Ph.D. programs is actually spent in apprenticeship, cognitive apprenticeship. You are picking up or apprenticing to the practices of an expert. And why does this apprenticing tend to commence after you have mastered the conceptual framework? The framework acts as scaffolding to help you structure the practice that you pick up. So learning in situ and cognitive apprenticeship fold together through this notion of distributed intelligence.

I dwell on this point because each of us has various techniques, most of them invisible, but nevertheless techniques that we use day in and day out to be able to learn with and from each other in situ. Students develop such techniques inside and outside of the classrooms. Indeed, on a campus as much learning happens outside the classroom as inside. As does learning how to learn. And understanding and supporting how this all happens requires understanding the dynamic flow in the above two-by-two matrix. Furthermore, if we could use the Web to actually support the dynamics among these quadrants, we could create a new fabric for learning, as well as learning to learn in situ, which has to be the essence of lifelong learning.

A brief detour about how I, personally, became engaged in this two-by-two conceptual framework of distributed intelligence. But first, let me say that I hope all of us here today recognize that this two-by-two is the essence of IRL, the Institute for Research on Learning. I borrowed (bricolage) this two-by-two slide from Susan Stucky and Peter Henschel several years ago and ever since it has become for me an icon for capturing the essence of designing systems that support organizational and individual learning.

How did I get into this? Before coming to Xerox, I was at BBN building intelligent tutoring systems and AI-based job performance aids for teaching complex electronic troubleshooting techniques. When I arrived at Xerox I discovered Xerox was spending many millions of dollars a year training tech reps how to repair our office equipment such as copiers and printers. Xerox wondered if I could use any of these sophisticated artificial intelligence tools to enhance the ability of our tech reps to learn similar troubleshooting skills more cost effectively. This could be a substantial opportunity for Xerox since it has 21,000 tech reps around the world.

Well, thank heavens what we did not do was to start immediately designing a system. Instead we did something that seemed bizarre to many. We hired some anthropologists. To one, Julian Orr, I said, "Julian, I want you to go live with these 'savages'. Understand how they work. Understand how they really learn and come back and tell us." And he went off to live and to work and to drink and to even to go to school with these guys.

When he came back he said, "John, you're not going to be happy."

"Why is that?"

"Everything you've ever written about troubleshooting is wrong. You think you are an expert on troubleshooting. But do you know how these guys actually do troubleshooting?"

"No, tell me, Julian."

"First of all, what happens is whenever a tech rep gets stuck he calls in another tech rep and then, standing around the problematic machine, they start to weave a story, a story that starts to explain some of the particular symptoms of the machine. And then some fragment of the initial story reminds them of something else which suggests a few more measurements to make which in turn produces some more data that reminds them of another fragment of a story, and so on. Troubleshooting for these guys is really just weaving together a narrative, a narrative that eventually explains all the symptoms and test data of this machine. And when they have made sense of all the data, the narrative is finished and the machine is diagnosed."

So troubleshooting is really story construction, not abstract logical reasoning. They didn't even talk about partial differential equations, which was a shock to me. These guys were storytelling, taking fragments of past stories that they knew or had experienced, weaving them together with new fragments until they had "explained" the machine's behavior.

Now, what do they do with these stories? When they come back to the home office, they sit around playing cribbage, drinking coffee, swapping war stories. Amazing amounts of learning were happening in the telling of and listening to these war stories. In telling a story, the story gets refined. In listening to the story, fragments of past stories get reorganized and refined and so on and so forth. Learning was happening in a fantastic way in terms of telling and listening to stories.

We now understand more about the architecture of the mind and how it is particularly well suited for remembering stories. That's the beautiful part. The sad part of it was that Xerox, under the guise of "business process reengineering," thought that storytelling was a waste of time. So big posters were put up saying, "Do not tell war stories." And lo and behold, instead they were sent back to Leesburg (Xerox' training center) to get efficient, formal, reliable training in how to repair these machines. Julian also went through this formal training. And guess what happened? When the tech reps returned to their home sites, they told stories now about Leesburg. Through their storytelling about Leesburg they transformed much of their experiences into something more meaningful. Well, this was an eye opener for me, along multiple dimensions. (And, in fact, it greatly influenced my own thinking about the formation of the Institute for Research on Learning and the need to combine the cognitive with the social and to take more seriously how to leverage naturally occurring resources for learning.)

To make a long story short, what did we do? What kind of system did we design, because of course as a technologist I was expected to build a system? We created a beautifully simple system, one that involved using two-way radios and no computers. We gave everybody in our tech rep community-of-practice test site a two-way radio, a radio that was always on, with their own private network. Because it was always on, they were always in each other's periphery. When a tech rep needed help, other tech reps in his community-of-practice would hear him struggling and if one of them had an idea he could move from the periphery to the (auditory) center, adding his fragment of "story" which usually suggested a new test to run or part to replace, and so on. And so basically we created a multi-processing, multi-person storytelling process running all across this initial test site. It worked incredibly well. In fact, it also turned out to be a powerful way to bring new people into the community since a novice could, as I mentioned earlier, lurk on the periphery and hear what was going on and in so doing could be a virtual cognitive apprentice. He could also move from the periphery to the center when he had something to contribute, very much like today's digital kids are doing on the Web.

The trouble with the above scenario is that all of these story fragments were being told through the ether and hence were lost to all not participating at that moment. And some of these fragments were real gems. We needed next to find a way to collect, vet, refine, coalesce and post them on a community knowledge server. Furthermore we began to realize that no one person was the expert. The real expert was not a person but was the community mind, the mind of the community-of-practice. If we could find a way to support and tap the community mind we might have a whole new way to accelerate learning and to capture and structure knowledge assets in the making and do all this with virtually no overhead. With this goal in mind a new system was designed, called Eureka.

This Web-based system has now been in use for over a year. It is particularly interesting because the tech reps in co-designing Eureka unwittingly reinvented the sociology of science as a way to make their ideas and stories more actionable. In particular, many of their ideas, many of the fragments of stories weren't actually trustworthy. In reality, they were just opinions, some of them kind of crazy, just like cold fusion was in science. In order to transform their opinions and experiences into "warranted" beliefs and hence actionable, the author would submit them to peer review, a process facilitated by the Web. His peers would quickly vet and refine the story. In addition, the author attaches his name to the story (or tip), thus creating both intellectual capital and social capital, the latter stemming from the fact that the tech reps who create really great stories (or tips) become local heroes and hence become more central members of their community of practice. This system has changed the learning curve of our tech reps 300 percent and will save us about 100 million dollars a year. It's a beautiful example of how some relatively simple technology can achieve power by tapping and supporting the social mind via capturing naturally occurring assets, refining them and then enabling them to be shared.

Having said all this, what are some other ideas that are beginning to emerge that might help us—either in the workplace or on the campus—capture, refine and share knowledge assets in the making? Are there ways to capture assets that are left just lying on the table, so to speak, and then use them in order to make learning substantially more productive, more productive in the firm, on the campus and even the region? The answer, I think, is yes.



Let me go through a couple of examples, ones that are meant to be suggestive. There are many other examples that I am beginning to see around the country, especially as entrepreneurs are now starting to see this as ripe territory. The first example is one that I first encountered at Stanford University and was created by Professor Jim Gibbons, the Dean of Engineering. I think he discovered the basis of this accidentally some years ago and has since refined the underlying concept over the past several years. His original insight emerged while teaching an engineering course in which some Hewlett-Packard folks were taking his course. Part of the way through the course the Hewlett-Packard people were transferred to a new location, making it impractical for them to continue to physically attend the classes.

What Jim did was simply to videotape the classes and send them the videotapes. Simple enough. The twist, though, was that once they received the video, the engineers would replay them in their own small study group, but replay them in a very special way. Every three minutes or so they would stop the video and talk about what they had just seen, and ask each other if there were any questions or any ambiguities that needed to be resolved. If so, they would discuss or argue about them, often replaying a prior part of the tape until they could resolve the question. Then they would go on and continue to play the videotape for another three minutes, and so on and so forth. This would, of course, take longer than just watching the videotape with no interruptions. These study groups were socially constructing their own understanding of the material. The results were that the students using this method out-performed the ones that were actually taking classes live. The results, now tried out with H-P engineers, college students, inmates of the California prison system, to name just a few, are that most students get over a half a grade point better by doing it this way than by physically attending the classes. This is not meant as a commentary on Stanford classes. Rather, it is an elegantly simple idea about how forming these study groups and letting them socially construct their own understanding around a naturally occurring event, the lecture, has turned out to be an amazingly powerful tool, especially given how easy it is to capture the original lecture on videotape. This technique leverages a naturally occurring asset and can be used by folks off campus without the need for expensive communication technology. Think about what this suggests for distance learning. And what does it suggest for on campus students, let alone community colleges?

The second example stems from research being done both at PARC and Cornell University. The PARC system is called Madcap and is an experiment on how we might leverage our forums. Every week we have a forum where we often get some wonderful outside speakers. We wanted to explore how we could leverage this naturally occurring asset better across the PARC community and also across the broader technical community. These forums have proved to be a really powerful stimulus to the whole Silicon Valley region. Of course we tape them and give anybody a videotape of any forum that they miss. But, in reality, almost no one ever replays the videotaped forum because it's very hard, unlike a textbook, to be able to skim through a video stream, looking for highlights. Might it be possible to use computation to automatically segment and highlight such a video stream? And, perhaps even summarize it.

We now have a prototype system for doing this designed by Dan Russell's group at PARC. First, we capture and store the digital video on a media server, which also marks and timestamps any time there is any uniquely identifiable event such as clapping, laughing, a slide change (the latter being identifiable by shifts in the color space). The audience can also use their laptops or Palm Pilots to take notes, notes that can be time-stamped and thus cross-indexed into the video stream. We also transcribe the audio stream. That's actually done by a transcription service that costs surprisingly little. All these "signals" are combined to make a soup of streams all cross-indexed with each other. The resulting structure becomes a very rich medium in which you can skim and pick out exciting moments where, for example, there was wild applause, laughter, heated argumentation, etc. From this structure you can tell when the energy in the room went up or when you or a colleague you know made an annotation and what kind of annotation it was.

This is a first stab at trying to find ways to capture and represent additional signals, signals that are created almost as a by-product of an audience listening to a presentation and then using these signals as structural indices to the video stream. The goal is to make this a richer knowledge asset than just the video so that browsing, reflection and focused conversations are more likely to happen. Also note that if you have a diverse set of people taking notes and who are willing to identify themselves, you start to create an ecology of annotations—diverse, overlapping, richly opinionated, etc. So, for example, when I note, "Wow this is an interesting idea," this annotation might be reinforced by ones made by others or it might contradict them. Either way that might be a useful signal. Comments might even cluster into natural equivalence classes that might provide additional structure. In any event, later, when someone sees such a tag, they can bring up a brief video segment to see what was going on at that point and see why I might have found it interesting while others might not. Or suppose you just wanted to look at the slides. Encountering something that is confusing, you can touch the slide, or a particular bullet on the slide, and up comes the video of the lecturer explaining that point. These are just some of the ways to capture, structure and transform a lecture into a knowledge asset rather than just a performance, a performance that only a small set of people actually get to experience and profit from.

The next system is an experimental system in use at Cornell University and designed by Dan Huttenlocher. Here they use dual video cameras, one on the lecturer and one that zooms in on any student asking a question. The video stream can then be automaticallly segmented, identifying exactly when a student asked a question or the lecturer changed a slide, etc. Once a slide is identified its image is passed to an optical character recognizer whose output is used to help create an index of the video stream content.

These examples are just the first steps toward capturing naturally occurring knowledge performances and coalescing around them additional tag structures in order to transform them into a more structured and useful knowledge asset. Of course, this asset, when viewed by others, can be part of another knowledge performance (and knowledge sharing) leading to an additional layer of annotations as the meaning gets further socially constructed and vetted. Although these examples are rough and embryonic, they are hopefully evocative of what could be done to capture and structure performances of all kinds that naturally occur on the campus and then to use the resulting knowledge structure in new contexts.

Let's move to the concept of an ecology and see how it might relate to learning environments. An ecology is basically an open, complex adaptive system comprising elements that are dynamic and interdependent. One of the things that makes an ecology so powerful and adaptable to new contexts is its diversity. Recall that with the prior examples of knowledge performances, it was the diversity of the comments that helped to create the texture of the knowledge asset and also enabled it to be used in ways that might never have been originally imagined.


Let's consider a learning ecology, in particular a learning ecology that might form around or on the Web. As a start down this path, consider the Web as comprising a vast number of "authors" who are members of overlapping special interest groups, many of which embody substantial expertise that exists in both written and tacit form. These special interest groups are often overlapping since it is quite common to find a member of one interest group to also be a member of another. Given the vastness of the Web, it's often possible to find a niche community or special interest group that exactly coincides with your own, idiosyncratic interest or, more to the point, a kid's interest. Furthermore, on the Web—as beautifully depicted by the New Yorker cartoon of a dog using the Internet saying, "On the 'Net nobody knows you're a dog."— a kid need not necessarily reveal himself as being just a kid. Indeed, I've watched a seven-year-old in New York have a conversation about penguins with an expert at the University of Pennsylvania. And I believe that that expert never had any idea that he was actually conversing with a seven-year-old. The expert probably realized that the other person was not a real expert on penguins but at least that he was really interested in and moderately well informed about them. Furthermore, in this kid's school there was no one, including his teachers, that shared an interest in penguins. But the kid, on his own, first found the interest group through navigation. Then he linked and lurked and at the right moment he transitioned from lurking to asking a question therein initiating a brief conversation with this expert. A small momentary effort of one expert inspired this kid.

Now, with the Web, these virtual communities of niche interests spread around the world interweave with local, face-to-face groups, where kids interact directly with one another, in school or outside. A new, powerful fabric for learning starts to emerge, drawing strength from the local and the global. A kind of cross pollination of ideas happens as a group of local kids, all of whom participate in different virtual communities, carry ideas back and forth between those virtual communities and their local ones. Now recall our emphasis that informal learning often involves a joint construction of understanding around a focal point of interest and one begins to sense how these cross-linked interest groups, both real and virtual, form a rich resource for learning. Of course, not all conversations, whether face-to-face or virtual, even if focused and well intended, lead to productive learning. Judgment, navigation, discrimination and synthesis are more critical than ever; again, congruent to our hypothesis about digital kids.

I've been really struck living in Silicon Valley and also spending a fair amount of time in other high-tech regions, that each region can be analyzed with respect to the quality and diversity of its knowledge producers and knowledge consumers. The classic way to view the knowledge producers in a region is to look at the system of higher education—K through 12 schools, community colleges, colleges, universities, libraries, civic centers—and see these elements as the region's producers of knowledge, knowledge that gets consumed by its inhabitants.


The above matrix lays out the classical consumers of knowledge—the firms, the enterprises, students and so on. But in most regions there is a very rich interplay between these two axes, although an interplay that seldom gets much attention. If the region is geographically compressed enough, you start to get all kinds of informal face-to-face connections between the knowledge consumers and the producers, especially as students work part-time in the surrounding firms, as new firms spin out of universities, as older firms retrain their employees, etc. But that's just the beginning. It is my guess that in the 70s and 80s we were preoccupied with the importance of science parks, the areas surrounding our major research universities. In the 90s a significant shift started to happen. Science parks were transforming themselves into learning parks. But few saw this as a significant shift. We keep talking about them as being science parks, yet these parks are affording increasingly rich intellectual and educational opportunities.

Besides all of the traditional learning resources that high-tech regions offer—high quality K-12 schools, colleges, libraries, public lectures, etc.—learning resources typically consumed by the inhabitants of the park/region, we see a flow of resources the other way. Corporate research centers and high-tech companies are increasingly providing adjunct professors, guest lectures, thesis supervision, richly textured case histories to the universities. They are also providing consulting and sabbatical opportunities for professors and graduate students, thus providing opportunities for the academy to become better grounded in real world problems. Although such intermixing is not fundamentally new, the degree to which it is happening is new and various kinds of cross linkages are growing.

The traditional producers of knowledge (e.g., faculty) are also becoming consumers of the knowledge that their traditional consumers (e.g., grad students, firms in the region) produce. In my opinion, this is healthy, very healthy indeed. But for a rich and highly textured mixture to emerge, physical proximity and density helped. For example, did different companies frequent the same restaurants and was there much overlap between the favorite hangouts of university students and the R&D folks of the surrounding firms?

Now let's overlay on top of this physical/social region, the Web, and look back to the example of the kids participating in local face-to-face groups but also tying into virtual ones. Again, on the Web there is seldom such a thing as just a consumer or just a producer. Basically, each of us is part consumer and part producer. We read and we write, we absorb and we critique, we listen and we tell stories, we help and we seek help. This is life on the Web. The boundaries between consuming and producing are fluid—the secret to many of the business models of Web-based commerce.

From this perspective the Web does three things. First, the Web helps to establish a culture that honors the fluid boundaries between the production and consumption of knowledge, recognizing that knowledge can get produced wherever serious problems are being attacked and followed to their root. Second, with the Web, it is easier for experts—in the academy or in the firm—to casually interact with others and thus to act as mentors or advisors for students (or knowledge consumers) of any age—that is, folks that want to learn. Third, the Web provides infinite reach, rendering accessible resources far beyond the region; yet, the power of this reach is greatly enhanced when the results of this reach act as cross pollination, providing new grist, new points of view for communities of practice of the region.

In essence the Web augments the knowledge dynamics of a region, increasing its diversity and expanding its learning resources by leveraging local expertise—in a lightweight way—for mentoring. More generally, it enhances the fluid boundaries between knowledge production and knowledge consumption and between the local and the virtual. The Web helps to build a rich fabric that combines the small efforts of the many with the large efforts of the few. It enables the culture and sensibilities of the region to evolve, not only by enriching the diversity of available information and expertise, but it tightens the feedback loops of bootstrapping. It increases the intellectual density of cross linkages. And it enables learning to happen everywhere—a learning ecology. And the lurking (or informal benchmarking) that happens in local hangouts can now get augmented by the Web, one feeding the other. In other words, a self-catalytic system starts to emerge reinforcing and extending the core competencies of the region.

Let me end with a brief reflection on a very profound shift that I believe is happening—a shift between using technology to support the individual and using technology to support relationships. This shift will be very important because with it we will discover new ways, new tools and new social protocols for helping us help each other, which is really the very essence of social learning. It is also the essence of lifelong learning, a form of learning that learning ecologies could dramatically facilitate. And being able to create learning ecologies in a region is a first step to constructing a culture of learning, more generally.

Thank you.