Some users will find it convenient to handwrite instructions to a computer, rather than speaking or typing them. Many companies, including Microsoft, have spent some years working on what we call “pen-based computers” capable of reading handwriting. I was overly optimistic about how quickly we would be able to create software that would recognize the handwriting of a broad range of people. The difficulties turned out to be quite subtle. When we tested the system ourselves it worked well, but new users continued to have trouble with it. We discovered we were unconsciously making our handwriting neater and more recognizable than usual. We were adapting to the machine rather than the other way around. Another time, when the team thought they had created a program that worked, they came proudly to demonstrate their achievement to me. It didn’t work at the demonstration. Everyone on the project happened to be right-handed, and the computer, which was programmed to look at the strokes in the writing, couldn’t interpret the very different ones in my left-handed penmanship. It turned out that getting a computer to recognize handwriting is as difficult as getting one to recognize speech. But I remain optimistic that as computer performance increases we’ll have computers able to do this too.
Whether you give the command by voice, in writing, or by pointing, the selections you’re going to want to make will involve more complicated choices than just which movie to watch, and you’ll want to be able to make them easily. Users won’t stand for being confused or frustrated or for having their time wasted. The highway’s software platform will have to make it almost infallibly easy to find information, even if users don’t know what they’re looking for. There will be lots of information. The highway will have access to everything in hundreds of libraries and to all types of merchandise.
One of the worries most often expressed about the highway concerns “information overload.” It is usually voiced by someone who imagines, rather aptly, that the fiber-optic cables of the information highway will be like enormous pipes spewing out large quantities of information.
Information overload is not unique to the highway, and it needn’t be a problem. We already cope with astonishing amounts of information by relying on an extensive infrastructure that has evolved to help us be selective—everything from library catalogs to movie reviews to the Yellow Pages to recommendations from friends. When people worry about the information-overload problem, ask them to consider how they choose what to read. When we visit a bookstore or a library we don’t worry about reading every volume. We get by without reading everything because there are navigational aids that point to information of interest and help us find the print material we want. These pointers include the corner newsstand, the Dewey decimal system in libraries, and book reviews in the local newspaper.
On the information highway, technology and editorial services will combine to offer a number of ways to help us find information. The ideal navigation system will be powerful, expose seemingly limitless information, and yet remain very easy to use. Software will offer queries, filters, spatial navigation, hyperlinks, and agents as the primary selection techniques.
One way to understand the different selection methods is to think of them metaphorically. Imagine specific information—a collection of facts, a breaking news story, a list of movies—all placed in an imaginary warehouse. A query does a search through every item in the warehouse to see if it meets some criterion you have established. A filter is a check on everything new that comes into the warehouse to see if it matches that criterion. Spatial navigation is a way you can walk around inside the warehouse checking on inventory by location. Perhaps the most intriguing approach, and the one that promises to be the easiest of all to use, will be to enlist the aid of a personal agent who will represent you on the highway. The agent will actually be software, but it will have a personality you’ll be able to talk to in one form or another. This will be like delegating an assistant to look at the inventory for you.
Here’s how the different systems will work. A query, as its name indicates, is a question. You will be able to ask a wide range of questions and get complete answers. If you can’t recall the name of a movie but you remember that it starred Spencer Tracy and Katharine Hepburn and that there is a scene in which he’s asking a lot of questions and she’s shivering, then you could type in a query that asks for all movies that match:"Spencer Tracy,” “Katharine Hepburn,” “cold,” and “questions” In reply, a server on the highway would list the 1957 romantic comedy Desk Set, in which Tracy quizzes a shivering Hepburn on a rooftop terrace in the middle of winter. You could watch the scene, watch the whole film, read the script, examine reviews of the movie, and read any comments that Tracy or Hepburn might have made publicly about the scene. If a dubbed or subtitled print had been made for release outside English-speaking countries, you could watch the foreign versions. They might be stored on servers in various countries but would be instantly available to you.
The system will accommodate straightforward queries such as “Show me all the articles that ran worldwide about the first test-tube baby,” or “List all the stores that carry two or more kinds of dog food and will deliver a case within sixty minutes to my home address,” or “Which of my relatives have I been out of touch with for more than three months?” It will also be able to deliver answers to much more complex queries. You might ask, “Which major city has the greatest percentage of the people who watch rock videos and regularly read about international trade?” Generally, queries won’t require much response time, because most of the questions are likely to have been asked before and the answers will already have been computed and stored.
You’ll also be able to set up “filters,” which are really just standing queries. Filters will work around the clock, watching for new information that matches an interest of yours, filtering out everything else. You will be able to program a filter to gather information on your particular interests, such as news about local sports teams or particular scientific discoveries. If the most important thing to you is the weather, your filter will put that at the top of your personalized newspaper. Some filters will be created automatically by your computer, based on its information about your background and areas of interest. Such a filter might alert me to an important event regarding a person or institution from my past: “Meteorite crashes into Lakeside School.” You will also be able to create an explicit filter. That will be an on-going request for something particular, such as “Wanted: 1990 Nissan Maxima for parts” or “Tell me about anybody selling memorabilia from the last World Cup” or “Is anyone around here looking for someone to bicycle with on Sunday afternoons, rain or shine?” The filter will keep looking until you call off the search. If a filter finds a potential Sunday bicycling companion, for instance, it will automatically check on any other information the person might have published on the network. It will try to answer the question “What’s he like?"—which is the first question you’d be likely to ask about a potential new friend.
Spatial navigation will be modeled on the way we locate information today. When we want to find out about some subject now, it’s natural to go to a labeled section of a library or bookstore. Newspapers have sports, real estate, and business sections where people “go” for certain kinds of news. In most newspapers, weather reports appear in the same general location day after day.