June 16, 1997
Wall Street Journal. Copyright Dow Jones & Co. Inc.
Cached from http://www.wsj.com/public/current/articles/SB865871697413819000.htm
Even the best technology won't be used if it is too complicated.
Take, for instance, that clock on the face of a videocassette recorder, which has become a parable on the problems with high-tech products. The clock is undeniably useful, but even many technologically savvy people don't bother to stop the thing from blinking "12:00." Why? Because setting it isn't intuitive on most VCRs.
The upshot: Technology often isn't used to its fullest potential. And that's especially true with computers.
So, research labs at universities and companies world-wide are searching for new ways for people to interact with their computers, with the goal of making the machines smarter and more user-friendly.
The result: a slew of advances, from software agents that watch how users operate and then anticipate their needs, to wearable computers that track users' every action during the day, to speech-recognition packages that let a person talk to the computer instead of typing in commands.
"There's a change in the [industry's] frame of mind that says the computer should come more than halfway to meet the consumer's needs," says Paul Saffo, a director at the Institute for the Future, a nonprofit research foundation in Menlo Park, Calif.
Some of the more extreme examples are still confined to the labs. But the front-runners of this new technology are already making their way to the public, and have the potential to drastically change the way people use computers.
Speech recognition, for instance, is an area scientists have been working on for more than 20 years. Their goal: finding a less artificial way for people to interact with computers. "The most natural way for a human to ask for something is by voice," says Don Nielson, a vice president at SRI International, a nonprofit research institute in Menlo Park, Calif.
Early speech recognition wasn't very effective: The software could only recognize up to a few hundred words over the phone, for example. Now a number of companies have introduced products that dramatically raise the bar for the technology.
One of the leaders is SRI, which formed Nuance Communications in January 1995 to make commercial products based on what SRI learned in the laboratory. So far, Nuance's most successful speech-recognition product is a voice recognizer for discount broker Charles Schwab Corp. The recognizer is connected to a phone system at Schwab so that when customers call in, they can find out how a particular stock or mutual fund is performing. (You can't place orders with the system, however.)
When you call Charles Schwab and enter the code to check a stock price, the voice-recognition system kicks in and asks what stock or fund you're looking for. Then, using a complex mathematical formula, it breaks down your request and analyzes it.
A few years ago, such systems understood only a few hundred sounds, but this system is programmed to recognize 14,000 discrete stocks and funds -- with over a million variations thereof, so a caller could say IBM or even Big Blue for International Business Machines Corp. The program can understand dialects from all over North America, including the distinctive accents of Brooklyn, Boston, and Texas.
The program is accurate about 95% of the time, according to Mr. Nielson. The program doesn't usually misunderstand words, says Mr. Nielson; the most common problem is that the system won't have a stock or fund in its database. If that happens, the system will ask the caller to repeat the name twice, and if that's unsuccessful, transfer him or her to a broker.
Other advanced voice-recognition work is going on elsewhere. BBN Corp. of Cambridge, Mass., has a number of products in development, including software that allows you to talk to your computer to surf the Web, instead of typing in those convoluted URL addresses. And IBM, in Armonk, N. Y., installs a simplified voice-recognition program in many of its home PCs. The software lets you dictate documents to your computer (albeit very slowly), and to locate and open files with voice commands.
Microsoft Corp. is working on different ways to let people interact with computers more naturally. For example, the Redmond, Wash., company incorporated animated characters called "wizards" to assist people in its Microsoft Office '97 software package.
If you type in a question, an animated character such as Clippet -- a paper clip with eyeballs -- appears and offers on-screen tips. Let's say you type in, "How do I print sideways?" Clippet appears and asks questions in dialogue balloons until it zeroes in on what you want. In the printing example, Clippet might ask whether you want to know about page orientation, margins or layout. If you click on the button next to page layout, it will ask whether you prefer landscape or portrait mode. By clicking on landscape mode, the page will print sideways.
Thus, Clippet understands questions even when they are worded somewhat fuzzily, because it looks for key words that tell it what the user is wondering.
"We have a huge group of people just doing pure research to understand how technology can be used to make things get done more easily," says Matthew Price, the product manager for Microsoft Office. Some users, though, say that the animated-character idea is a bit too cute for their taste, and that the information could be better dispensed through a plain-vanilla help menu.
Microsoft has also tried to make its latest Web browser, Internet Explorer 4.0, easier to use. The browser keeps track of every site you visit and then uses this information to try to anticipate the Web address you want to see.
Type in "yah" in the address line, for example, and the browser might search the names of your favorite sites and offer the option of linking to www.yahoo.com, the Yahoo! search engine.
The browser also makes getting to those engines a bit easier: Type in a question mark followed by a few words you want to search for, and a screen pops up offering listings for several different search engines that you can browse.
"We've tried to strip away the complexity that comes with navigating the Web," says Yusuf Mehdi of Microsoft's Internet division.
Another experiment in user-friendliness is being promoted by Molloy Group Inc. of Parsippany, N. J., which has developed a "cognitive processor" that helps software learn by experience. One use: a trouble-shooting system that Molloy is testing at Canon Inc. and three other companies.
The system aids corporate help desks by providing a database for techies to draw on. When an employee calls with a problem, the help-desk operator can type in a description and the software will spit out its diagnosis and likely remedies. Every time the computer successfully answers a query, it stores in its memory the steps it used to get there, thus learning to recognize different questions and answer them more effectively.
"It's very easy to interact with because it is so tolerant," says Bruce Molloy, the company's chief executive officer. "You can type incomplete or fuzzy statements, and the system will come back and ask you questions until it knows what the problem is."
In a test, the software was able to help with several specific queries, such as "How do I change the font?" For this, the program opens and guides you through the help menus in Microsoft Word. But sometimes the software is stymied by broader questions. "Why won't my printer work?" is met with the somewhat insulting suggestion that maybe the printer is turned off.
Of course, the ultimate in user-friendliness would be a situation where the computer seamlessly fits into any situation in everyday life. That, in a sense, has been the point of those personal digital assistants that can do everything from remembering phone numbers to receiving e-mail. But now several students at the Massachusetts Institute of Technology's Media Lab are taking the concept a step further. Or several steps.
The idea: build a portable computer that does everything a digital assistant can -- but that also has sensors to monitor the user's environment. So, if the user is, say, in a conversation, the computer could judge if it was appropriate to interrupt him or her with a phone message. Or if the user meets someone and later forgets his or her name, the computer could retrieve the name from memory.
"What your computer really should do is sense your context, and know if you're talking with someone, walking in a hallway or eating," says Thad Sparner, a graduate student who is finishing up his doctoral degree at MIT.
The finished product is a long way off. And the intermediate designs can look strange, to say the least. With bulky hard drives hanging in satchels from their shoulders and video cameras worn over one eye, these graduate students look a bit like Robocop. Still, six of them tote the computers on a regular basis, wandering around the MIT campus in Cambridge, Mass., and the surrounding area. The advanced features -- such as awareness of context -- are "one to five" years away, the students say. For now, the machines basically do what digital assistants can: store databases and keep users in constant communication with electronic and voice mail.
Bradley Rhodes, another MIT graduate student who is working on his doctoral degree in media arts and sciences, says he wears the contraption 12 to 16 hours a day. Mr. Rhodes, who admits he doesn't have a girlfriend at the moment, says his group is concerned about how people will react to wearable computers. That's why they have designed a tiny, one-handed keyboard that fits in a coat pocket, so they can type without people seeing what they're doing. And within the next five years, the students are hoping the wearable computer will have evolved into a much slimmer apparatus.
"We want it to be completely unobtrusive, so it can disappear into the sole of your shoe or something," says Mr. Sparner.
Ms. Quick is a staff reporter in The Wall Street Journal's New York bureau.