CU-Boulder Professor Explains

January 10, 1997

If a computer has ever given you fits, University of Colorado at Boulder psychology professor Thomas Landauer will probably cheer you up.

He says it's not your fault.

Everyday examples of how computers have actually made work more difficult abound, Landauer says. But rather than blaming the person who designed the computer people often blame themselves.

"There's almost never something that should be called a human error," Landauer says. "If there's a human error made in operating the computer it's because the system wasn't designed right. A system that's properly designed prevents errors."

Landauer authored an award-winning book titled "The Trouble with Computers: Usefulness, Usability and Productivity" published by the MIT Press and recently released in paperback. The book examines a problem called the "productivity paradox," that is, despite a $4 trillion investment in computers between 1973 and 1992 overall productivity gains have been quite small.

He doesn't want to get rid of computers but he is insistent on wanting to make them more useful. Before coming to CU-Boulder in 1994, he spent 26 years at Bell Labs and Bellcore, the last 10 years as director of cognitive science research at the New Jersey laboratory serving the Baby Bell telephone companies.

"One of the major claims in my book is that the whole trouble with computers boils down mostly to a failure to evaluate them for the purposes they really are intended for," he says. "What you should care about with a computer is whether it lets you get things done more easily, more efficiently and with greater satisfaction."

While the efficiency gains resulting from the introduction of textile or agricultural machinery is measured in increases of hundreds or even thousands of a percent, the computer efficiency gain is well under 50 percent, Landauer says. That is not enough to justify the enormous amount of money and time spent on them.

"People have been buying computers without finding out whether they will do what they want them to do," he says. "They are marvelous and awe-inspiring and people therefore seem to feel they must be productive."

Between 1973 and 1992, the U.S. investment in computers has been about $4 trillion dollars, he says. But according to a variety of economic studies, the return on investment is about zero -- for every dollar invested a dollar was produced. Other types of investment did much better.

Some people have argued that the productivity problem ended after 1993 but the facts are not yet clear, he says, and probably not much more encouraging.

Horror stories about formatting and calculation problems and losing several days work with a single click of the mouse are commonplace. But people usually don't count that lost time when considering how well a computer is serving them, he says. They consider it a mistake, a fluke.

When those mistakes are counted against a computer's efficiency it makes a vast difference, he says.

"It's a well-established fact you can make mistakes on computers that take much longer to correct than any mistake you can make in any other kind of technology," he says. That's because the same power of the computer to help you when it's doing the right thing can really hurt you when it's doing the wrong thing.

Making computers that increase efficiency requires testing by real workers doing real jobs and then comparing those results to the way people used to do things, Landauer says. Engineers who build computers and software often don't understand how difficult it is to use their software, and it is the way the software interacts with the user that is the source of the problems.

The development process should involve testing and fixing and more testing and fixing until the system is easy and foolproof to operate. "That can be done and is being done, but it isn't done regularly," he says.

Landauer likes to tell the story of returning a bicycle part at a store, a transaction that used to be both simple and quick before the shopkeeper bought a computer. Now it takes about 20 minutes to properly enter information about the part into the computer and wait for a printout. The process is now slower, costlier, takes a lot of time to learn and generates lots more paper.

"I don't want to get rid of them, I just want to get them to be better," he says.

Give FeedbackSee More Photos View Photo