You have to feel sorry for computers. They are having a hard time living up to peoples' expectations.

Computers have been counted on to rescue failing manufacturing industries through a new burst of automation.

In the paperless factory of the future, engineers at computer terminals would design new products and transmit dimensions directly to robot-controlled machines. Other computers would reject faulty products, manage inventories, fill orders and bill customers electronically.

In the service economy, computers would usher in a new Age of Information, creating electronic pipelines to carry data, voices and pictures simultaneously between homes and libraries, stores, businesses and doctors' offices.

All this remains in the cards, but it is approaching at a slower, more uneven pace than enthusiasts for technology promised, expected or hoped.

The computer industry, no longer the exception, is facing the problems of overcapacity, shrinking profit margins and hypercompetition that dog older industries such as steel. Even International Business Machines Corp., the industry bellwether, is having trouble moving computers.

Researchers are looking anew at the computer revolution, asking some sobering questions: Has the promise of the technology been oversold? Is the information computers assemble expanding too rapidly for human operators to absorb? Do investments in technology really justify the costs?

Some answers come from a detailed case study about the automation of the U.S. metalworking industry in a book by sociologist Donald A. Hicks, published by the American Enterprise Institute.

His study looks at the way technology has spread through this industry, which includes both large corporations and tens of thousands of small businesses that cut, drill and shape metal for other manufacturers.

The automation of the metalworking industry has been surprisingly slow, Hicks reports. He quotes estimates that less than 4 percent of metal-cutting and metal-forming machine tools in the United States were computer-controlled as recently as 1983. This is true even though computer-controlled machine tools are considerably more reliable and efficient than those run by human operators.

One solution would be to speed up the introduction of modern computerized machine tools by giving companies even larger tax incentives to buy them. But Hicks argues that these companies have good reasons for not rushing to buy the newest technology. "It is not enough that technology offers a better way of doing something," "It is not enough that technology offers a better way of doing something." -- Sociologist Donald A. Hicks he said. The technology also must fit the peculiar requirements of the actual production arrangements within the workplace, he added. In other words, managers who take a realistic look at their businesses, their employes' skills, their customers' needs and their competition -- and then put off buying a costly new computerized lathe for the time being may well know what they're doing.

Another researcher, Martin Neil Baily, warns that those calculations are harder to make when technology is brought into the white-collar service sector.

Baily, a senior fellow at Brookings Institution, recalls his visit to a large company that has computerized its shipping operations. "Instead of clerks filling in the bills by hand and filing them in a large cabinet, everything is now computerized," he said in an article published by Bell Atlantic Co.

The result was a huge increase in the information readily available to company planners, including volumes of detail about geographic and seasonal shipping patterns. "What is this information worth to the company? They do not know," he concludes. They can't tell whether the value of these data exceeds the cost of collecting and processing them.

In the factory, the value of an investment in technology is usually a straightforward calculation, because the yards of textile or the accuracy of a spot weld can be directly measured or checked, Baily says.

The information that comes from new technology in the service sector is much harder to measure. When a new kind of diagnostic machine is rolled into a hospital, the doctors can get more detail about a patient's condition.

But there is no easy way to calculate the value of it. And because an insurance company is usually paying the bill, the patient isn't thinking hard about the costs and benefits of the additional tests.

Throughout the white-collar world, the cost of processing information is declining dramatically and the amount of information that can be assembled on a computer screen is growing just as fast. The bottleneck is understanding how to use this outpouring of information.

"It is not enough to show the dazzling potential of the technology; some value must be placed on the increased flow of information. It either reduces costs elsewhere or it attracts new customers," said Baily. But it can be justified only by what it accomplishes.

There is an inevitability about the increasing power of computers to process information. In 1970, the mainframe computers capable of making a million calculations a second cost $1 million each and filled several hundred square feet of office space. Six years ago, a computer with that same power cost $50,000 and could sit on a desk. By the year 2000, it will cost $30 and will fit inside a briefcase, scientists predict.

There is nothing inevitable about our ability to tap that power to solve the problems of companies, governments and people.