Changing TraditionsBy CIOinsight | Posted 10-10-2002
Expert Voice: David Gelernter on Knowledge Management
It's an old story: Researchers at Xerox Corp.'s Palo Alto Research Center developed the basic elements of the personal computer graphical user interfacethe desktop analogy, the mouse, multiple windows and moreback in the 1970s.
It was first commercialized in 1984 by Apple Computer Inc. in the form of the Macintosh, on which Microsoft Corp. based the desktop analogy for its now ubiquitous Windows operating system. The upshot: Computer users are still heavily dependent on a 30-year-old technology. Aren't there alternatives?
One technologist who believes fervently in alternatives is David Gelernter, a professor of computer science at Yale University and a cofounder and chief technologist at Mirror Worlds Technologies Inc. Gelernter and his team have developed a software program intended to revolutionize how personal computers save and display information.
The goal: to present all informationword-processing documents, e-mail, pictures, music, everythingas a stream of time-ordered files that can be reorganized instantly into substreams by topic.
Executive Editor Edward H. Baker recently chatted with Gelernter about the technological and philosophical basis for his software, and his belief that the computing public is being shortchanged by the myopic views of today's post-Internet IT industry.
CIO Insight: The software your company has developedradical as it isis usually lumped into the category of knowledge management. Why does knowledge management need to be revamped?
Gelernter: I think the field of knowledge management is struggling to express the fact that it wants to move up an entire conceptual level from where conventional software has pegged it. It doesn't want to deal with traditional operating system ideas of files or even applications or dataor for that matter, information. All this is irrelevant. People want to connect directly at a higher level to the knowledge or the information that defines their lives, and they don't want to be boxed in by an operating system or any particular machine.
In thinking about the term "knowledge management," how do you distinguish between information and knowledge?
Well, it's really an artificial distinction. I think information and knowledge were seized on as an alternative to data and the data processing view of things. But if there is a distinction between information and knowledge, knowledge is supposed to be something at a higher level. Knowledge implies context. You can't have an isolated little tidbit of knowledge. You've got to understand what it's part of, you've got to have some idea of the big picture.
If I'm researching some topic and I come up with 147 isolated tidbits or news stories on Nexis or in e-mails or whatever, in no particular order, I have a bag of random information. But if I take those 147 bits of information and arrange them into a big picture so I understand that first this happened and then that happened and this led to this and then this happened in consequence and in the future we're expecting that, if I arrange those 147 bits and pieces into a narrative or a story that gives me some context, some feel for the big picture, then I've graduated to knowledge.
And yet no one's proposing to call this the Knowledge Age.
No, they sure aren't, and it's a striking fact. If you read the newspapers and see the unfolding story of U.S. intelligence agencies, for example, we are supposedly in the Information Age, and in some ways we are fantastically badly informed. And the reason we're badly informed is that we keep failing to put two plus two together. We keep failing to see the big picture. That's the story of Sept. 11 that emerges as we hear more and more tidbits of information that the FBI had or the CIA had, and the catastrophic tragic failure to put these pieces together and to see the big picture.
These are big issues, but I think many people have this kind of experience on a daily level as we get hit by stuff all the time at a higher and higher rate. We're like asteroids getting showered by little space fragments. I think lots of people have the feeling that it's no gain to them to be hit by more little tidbits and fragments and jagged packages of information if they don't know the big picture, if they can't add it up, if they don't see where it's leading, if they don't understand the story line.
Yet ironically, we have the wrong information structures, the wrong knowledge structures, so increasing the amount of data makes us worse and not better informed, because we're buried under more and more stuff and we have less time to put the pieces together, we have less time to think about it, we have less time to mull over the big picture and let it emerge.
Yes, no one is calling this the Knowledge Age. I think the nation does not feel well informed. I think businesses don't feel well informed. I know individuals don't feel well informed. And it's remarkable that more people aren't asking themselves, "Why is it that in this day and age, when we have access to more data by orders of magnitude than we've ever had before, why don't we feel better informed?"
CIO Insight recently interviewed Etienne Wenger (May 2002), who believes that corporate knowledge arises out of what he calls communities of practice. How does that fit into your view of the knowledge crisis?
I think communities of practice is one way of saying what we've all said repeatedly, which is that the idea of corporate memory or communal memorya community's joint experience or narrative or history or storyneeds to be accessible. We want to take communal memory or corporate memory away from being a sort of vague, poorly defined idea. We want software to support that idea directly.
There's a lot of informationor in a broader sense, a lot of wisdomin any company's experience or university's experience or department's experience. Mirror Worlds Technologies is a company that's done a bunch of things and learned things and built things and so forth. That's true of the Yale Department of Computer Science and a million other communities and organizations.
Yet we still treat that collective experience and wisdom as a kind of footnote, a negligible commodity that you can use to the extent you can remember it, that new people don't have any access to, that comes and goes as staff come and go. Instead, we should be able to treat it as a fundamental resource and maybe the most important asset that any organization has. And we want to see it. I want to see it as a tangible thing. I want to see it as a narrative on screen, and I want to be able to follow it, and I want to be able to find out what this organization was doing last year at this time, and I want to know what this organization is planning to do next month.
Imposing a Narrative Structure
Imposing a Narrative Structure
Why is a narrative structure so important to this process?
We were looking for the bottom layer, the fundamental organizing principle, and we wondered: Should it be a tree, like in a standard file system hierarchy? Should it be hypertext, that is, a random network? Should it be a bag with no particular order at all? All those seemed wrong. Instead, it seemed that the fundamental structure of life was a good basis for building any kind of knowledge or information system.
We wanted information to be arranged the way life is arranged, and life is a narrative. There's no way to live except moment by moment. Every human being has the experience of going from one moment to the next to the next, living a timeline with a past, present and future. That's the universal skeleton key to how experience is organized and how memory is organized.
If your information is stream structured, you've got to at least consider the possibility of treating the past as something you can deal with rather than unknown territory. You've got to consider the possibility of history. You've got to consider the possibility that time passes, things change, the world evolves. You've got to consider the possibility that this is not a static world. We aren't fixed in the present, it's not always going to be this way, it didn't always used to be this way. Things change, things happen, things evolve.
Yet user interfaces have always been thought of in spatial terms, using the computer monitor to represent information.
Absolutely. Yeah, that's exactly the way they're thought of. The idea of the desktop interface with its windows and icons and so forth is, of course, an electronic version of a physical desk, which is a little piece of space, and I put stuff here and stuff there. And people think of a tree of files, I mean, a tree is an object in space, and they think of a file tree or hierarchy as a kind of quasi-spatial structure.
And as far as we're concerned, that was the inevitable way to get started in information management. It's like when people began thinking about flying: The natural way to fly was to build an artificial bird, just as the natural way to organize information on a computer was to build an electronic office with a file cabinet and a desktop. But it turned out that an artificial bird was the wrong way to make to an airplane, and that airplanes have their own rules.
And we believe that an artificial office and artificial file cabinet and artificial garbage can and desktop and so forth was an obvious first guess but not the right way for information to be organized on a computer, and that a narrative or a time stream, which makes perfect sense intellectually or cognitively, may not work in a paper office, but it does work when we have the software to make it happen.
Applying a Narrative Sequence
Applying a Narrative Sequence to Healthcare
Can you think of a thorny problem that might logically be reduced to narrative sequence?
Surehealthcare. Here's how the vast majority of people deal with a doctor: The doctor sits down and he's got a folder, and that folder is a heterogeneous bunch of notes and X-ray images and copies of prescription forms and lab test results. It's absolutely clear that the first step in putting medical records online has got to be based on that folder.
I've got to be able to open that folder up andworking from the bottom because it's arranged chronologically, like a streamtake the oldest document and scan it in, take the second oldest document and scan it in, and so forth. Just feed each item in the folder into software.
And what do I want to have happen to it? I want it to turn into a stream because this is a chronological record of my encounters with this particular doctor, and it's heterogeneous. I've got to be able to deal with handwritten stuff, so I'll look at that in the form of images. And I've got to deal with X-ray images as well, and so forth. Anything more complicatedsay, translating everything in that folder into a certain database formatis never going to be launched because it's just too expensive. It's too hard, and it's too massive an issue.
If I deal with three doctors, or whatever my medical world consists of, I have access to three streams. I can shuffle them together, and that's a medical record. That's a first step, and in many ways I think it'll be a very useful step toward ending the discussion about formats and saying we'll use the current format and put it online in the obvious, natural, intuitive, neutral way.
But then there are questions about controlling access to this and how to make parts of it available. I may want one doctor to see the whole thing, but the insurance company may only need access to certain parts of it, and so forth. I can do that automatically by saying, "Permission all documents of this sort to the insurance company."
And that way, instead of preparing a separate database or a separate presentation for the insurance company, the insurance company just looks at my stream when it needs to deal with my case. The insurance company has access to everyone's stream, but only a limited view, only the parts I've chosen to permission to it.
So, ultimately, I control my medical records. That's another aspect of medical records that the online effort hasn't really grasped: Ultimately, my medical records belong to me. They're not the private property of the hospital or the insurance company. They're my records and control ultimately needs to be in my hands.
Why are we still stuck in the old ways?
What I see in the technology world is more continuation of the habit of going back to old solutions and making them more complicated and adding new features and squeezing them harder rather than considering the possibility of what can we do differently now that computers are so much more powerful than they used to be. Do we still have to do what we were doing in 1984 [when the Macintosh was introduced], or can we do things differently?
Technologists are by and large wedded to different varieties of spatial organizations. Whether it is flogging the hypertext model, or going back to the distributed object file system model that we heard so much about in the late 1980s and early 1990s, or talking about search engines and search engine technology, which has been a big issue throughout the 1990s. Or working in the conventional hierarchical file system in desktop GUI environments. Or asking questions like: "What has Apple been doing in the online context? What interesting new gimmicks can we add? How can we accommodate audio stuff, how can we accommodate video material? What new categories, what new special applications, what new gadgets and gizmos and features can we add to the new release?"
The problem is that nothing ever changes. Instead of reorganizing, instead of simplifying, instead of streamlining what we've got, we just add more paper to the pile and the pile totters higher and higher. I think the trends and currents that technologists seem to hit on today are retreads.
And there are more and more separate, hermetically sealed technical communities within the technology world, and it leaves the nontechnical user increasingly badly served. My desktop computer has all kinds of power that no one is doing anything with because the software hasn't changed in any basic way for almost two decades. And we get more and more elaborations and fancy features and technical details in applications that really don't do what I need to have done.
You can see this in the bad relationship between the public and technology. I think people are intimidated by technology, but they don't want to say it. I think there's very little enthusiasm or excitement out there in the world of users of new computer technology, and that's very different from what it was like five years ago.
The industry has managed to kill off a tremendous wave of excitement and enthusiasm and interest by its complacency and its unwillingness to broaden its focus and think about what actual users need from their computers as opposed to what great new features can be added to this already grossly overfeatured software package.
There are certainly some technologists who are interested in a design approach, who care about making things simpler and more powerful and better rather than just revisiting golden oldies again and again. But I think they're still a small minority, and I think it's too bad. I know it will change, but I don't think it's changed a lot yet. Desktop computers are still new, and we still don't know what role they're really going to play. They're like the airplane in the late 1920s, years before anything like a modern airplane design had emerged. That's where we're at in the evolution of desktop computing.
David Gelernter is chief scientist at Mirror Worlds Technologies Inc. and a professor of computer science at Yale University. He has written numerous books and textbooks on computer science, aesthetics and cultural history, and his articles have appeared in The Wall Street Journal, The New York Times, The Washington Post, Commentary, ARTNews, National Review, Time magazine and others.