If an agent that could learn were available now, I would want it to take over certain functions for me. For instance, it would be very helpful if it could scan every project schedule, note the changes, and distinguish the ones I had to pay attention to from the ones I didn’t. It would learn the criteria for what needed my attention: the size of the project, what other projects are dependent on it, the cause and the length of any delay. It would learn when a two-week slip could be ignored, and when such a slip indicates real trouble and I’d better look into it right away before it gets worse. It will take time to achieve this goal, partly because it’s difficult, as with an assistant, to find the right balance between initiative and routine. We don’t want to overdo it. If the built-in agent tries to be too smart and anticipates and confidently performs unrequested or undesired services, it will be annoying to users who are accustomed to having explicit control over their computers.
When you use an agent, you will be in a dialogue with a program that behaves to some degree like a person. It could be that the software mimics the behavior of a celebrity or a cartoon character as it assists you. An agent that takes on a personality provides a “social user interface.” A number of companies, including Microsoft, are developing agents with social-user-interface capabilities. Agents won’t replace the graphical-user-interface software, but, rather, will supplement it by providing a character of your choosing to assist you. The character will disappear when you get to the parts of the product you know very well. But if you hesitate or ask for help, the agent will reappear and offer assistance. You may even come to think of the agent as a collaborator, built right into the software. It will remember what you’re good at and what you’ve done in the past, and try to anticipate problems and suggest solutions. It will bring anything unusual to your attention. If you work on something for a few minutes and then decide to discard the revision, the agent might ask if you’re sure you want to throw the work away. Some of today’s software already does that. But if you were to work for two hours and then give an instruction to delete what you’d just done, the social interface would recognize that as unusual and possibly a serious mistake on your part. The agent would say, “You’ve worked on this for two hours. Are you really, really sure you want to delete it?”
Some people, hearing about softer software and social interface, find the idea of a humanized computer creepy. But I believe even they will come to like it, once they have tried it. We humans tend to anthropomorphize. Animated movies take advantage of this tendency. The Lion King is not very realistic, nor does it try to be. Anybody could distinguish little Simba from a live lion cub on film. When a car breaks down or a computer crashes, we are apt to yell at it, or curse it, or even ask why it let us down. We know better, of course, but still tend to treat inanimate objects as if they were alive and had free will. Researchers at universities and software companies are exploring how to make computer interfaces more effective, using this human tendency. In programs such as Microsoft Bob, they have demonstrated that people will treat mechanical agents that have personalities with a surprising degree of deference. It has also been found that users’ reactions differed depending on whether the agent’s voice was female or male. Recently we worked on a project that involved users rating their experience with a computer. When we had the computer the users had worked with ask for an evaluation of its performance, the responses tended to be positive. But when we had a second computer ask the same people to evaluate their encounters with the first machine, the people were significantly more critical. Their reluctance to criticize the first computer “to its face” suggested that they didn’t want to hurt its feelings, even though they knew it was only a machine. Social interfaces may not be suitable for all users or all situations, but I think that we’ll see lots of them in the future because they “humanize” computers.
We have a fairly clear idea of what sorts of navigation we’ll have on the highway. It’s less clear what we’ll be navigating through, but we can make some good guesses. Many applications available on the highway will be purely for fun. Pleasures will be as simple as playing bridge or a board game with your best friends, even though you are all in several different cities. Televised sporting events will offer you the opportunity to choose the camera angles, the replays, and even the commentators for your version. You’ll be able to listen to any song, anytime, anywhere, piped in from the world’s largest record store: the information highway. Perhaps you’ll hum a little tune of your own invention into a microphone, and then play it back to hear what it could sound like if orchestrated or performed by a rock group. Or you’ll watch Gone With the Wind with your own face and voice replacing that of Vivien Leigh or Clark Gable. Or see yourself walking down the runway at a fashion show, wearing the latest Paris creations adjusted to fit your body or the one you wish you had.
Users with curiosity will be mesmerized by the abundance of information. Want to know how a mechanical clock works? You’ll peer inside one from any vantage point and be able to ask questions. Eventually you may even be able to crawl around inside a clock, using a virtual-reality application. Or you’ll be able to assume the role of a heart surgeon or play the drums at a sold-out rock concert, thanks to the information highway’s ability to deliver rich simulations to home computers. Some of the choices on the highway will be supersets of today’s software, but the graphics and animation will be far, far better.
Other applications will be strictly practical. For example, when you go on vacation a home-management application will be able to turn down the heat, notify the post office to hold your mail and the newspaper carrier not to deliver the printed paper, cycle your indoor lighting so that it appears you are home, and automatically pay routine bills.
Still other applications will be completely serious. My dad broke his finger badly one weekend and went to the nearest emergency room, which happened to be Children’s Hospital in Seattle. They refused to do anything for him because he was a few decades too old. Had there been an information highway at the time, it would have saved him some trouble by telling him not to bother trying that hospital. An application, communicating on the highway, would have told him which nearby emergency rooms were in the best position to help him at that particular time.
If my dad were to break another finger a few years from now, he not only would be able to use an information highway application to find an appropriate hospital, he might even be able to register electronically with the hospital while driving there and avoid conventional paperwork entirely. The hospital’s computer would match his injury to a suitable doctor, who would be able to retrieve my father’s medical records from a server on the information highway. If the doctor called for an X ray, it would be stored in digital form on a server, available for immediate review by any authorized doctor or specialist throughout the hospital or the world. Comments made by anyone reviewing the X ray, whether oral or in text form, would be linked to Dad’s medical records. Afterward my father would be able to look at the X rays from home and listen to the professional commentary. He could share the X rays with his family: “Look at the size of that fracture! Listen to what the doctor said about it!”
Most of these applications, from checking a pizza menu to sharing centralized medical records, are already starting to appear on PCs. Interactive information sharing is quickly moving closer to becoming a part of everyday life. However, before that happens, a lot of pieces of the highway still have to be put in place.