By David Weinberger
For decades, when people thought of computers they thought of soulless machines that were instruments of conformity. There was good reason for this. That's exactly what computers were. Computers only really took off when they moved away from this type of rigidity. KM, unfortunately, runs the perpetual risk of forgetting that lesson.
Computers were born dumb. Even after we were past the days of punch cards, if you wanted your computer to process your data—for example, if you were applying for a mortgage—you had to lead it by the hand. Computers could only recognize that the string "2001" was the year and not the amount of your down payment by seeing that you had typed those numbers into a box that said "Year," thus giving the computer the metadata it needed to figure out what the heck you were talking about. Thus, our interactions with the computer were constrained to our pointing and grunting at it. Not very subtle.
We don't much like that way of behaving. We consider data entry to be the bottom rung of the computer career ladder precisely because filling in blanks is tedious. Why, it's the sort of thing computers should do! That's one reason why the personal computer took off, because with the PC came word processing. At last a computer wasn't forcing us to normalize our content, that is making us fit it into the assigned blanks. Instead, we could type in whatever we want and then assign it metadata by using key combinations that would say that this string of characters should be centered and that string should be italicized. At last we were dealing with content unconstrained (apparently) by the computer's need for metadata. The PC took off.
But, of course, word processors are all about metadata. The two basic strategies for dealing with it in word processing programs should teach us some lessons about what users want when it comes to dealing with information about information—and "information about information" is the content of KM.
Microsoft Word was originally designed around the idea of "declarative markup." That is, you defined the characteristics of the various types of paragraphs you want—body text, title, numbered list, etc.—and then simply flagged paragraphs as being of this or that type. Word calls these definitions "styles." (Interleaf called them "components." SGML calls them "elements.") There are tremendous advantages to using styles. You can update the style through a dialogue box and all instances update. You can apply a new style sheet (= template) and transform the entire look of a document. But, fewer than 10% of users employ styles, preferring the WordPerfect technique of selecting text and applying font changes, etc. Apparently, we want to stay as close to the content level as we can, avoiding the abstractions of metadata. Creating styles puts us right back into the old computer world of filling in forms, although in this case the form is a dialogue box. When we create and maintain styles, we feel like we're using a computer, not writing. (I, however, am a styles fanatic and have been known to create separate styles for two elements that are formatted alike but have different roles to play. This is the point at which being a Good Boy is indistinguishable from Wasting Time.)
If we resist using styles in part because it puts us back into the world of computers, the Web makes computers more invisible than ever before. Because a networked computer is a communication device more than an information processor, writing frequently becomes a direct relationship with the reader. Word processing is like making a speech: The audience is assumed to be passive. Writing for the Web—whether it's a Web page, a ’zine, or a line in a chat room—feels much more like a conversation. We expect replies. I am not typing into my computer; I am talking with someone. At its best, the Web makes my computer as invisible as the handset of a telephone when I am having a spirited conversation with a friend.
Now, how does KM work? In general—and, yes, this is a big generalization—it wants to put information and ideas into boxes so they can be found, shared and leveraged. It can either do this by requiring the user to fill in metadata or by trying to discern the metadata. The former puts us back into the pre-PC world of computers and conformity, which kills the conversations that generate the most worthwhile ideas. The latter is very difficult to do well . . . and, worse, extraordinarily easy to do really badly.
This is the Scylla and Charybdis of KM. There is no easy course. In my opinion, there are few places where getting users to fill in the metadata blanks will work, and there are always prices to pay: It costs users time, it limits the metadata to what the user thinks it should be, and it fuels resentment. The alternative leads to a messy KM environment with many wrong guesses made by the KM software. But the history of the computer--from instrument of conformity to Web communications tool--has taught us. that messiness has its own rewards. In short, when in doubt, get messy.
David Weinberger edits "The Journal of the Hyperlinked Organization" (www.hyperorg.com), e-mail [email protected]