IRS Computer Fiasco Is Tip Of Software-Glitch Iceberg

Computers are so difficult for so many people that some experts are starting to raise red flags. The problem, they say, is not the people; it's the machines and, especially, their software. And it's time to complain, loudly.

"Let's not take this anymore," says Charles Kreitzberg, president of Cognetics Corp., a Princeton, N.J., software-usability consulting firm. "Let's really raise our voices and say things have to change."

"It's time to get angry," adds Ben Shneiderman, head of the Human-Computer Interaction Laboratory at the University of Maryland in College Park. "We need to make the issues more subject to discussion."

The latest evidence of computer-design failure comes courtesy of the Internal Revenue Service (IRS). In testimony before Congress, the federal agency responsible for collecting taxes admitted that several pieces of its $3.4 billion modernization program "do not work in the real world." But its problems merely reflect what is happening in other corporations and government agencies.

The IRS problem "is a large-ticket failure, but I've seen several comparable failures," says Tom Landauer, professor of psychology at the University of Colorado and author of a 1995 book, "The Trouble with Computers." "The people who build these things are very smart. [But] people do not devote frequent enough effort to find out what the system is supposed to do."

As a result, the systems "work," but the operators have to be quasi-engineers to figure them out, he adds. Mr. Landauer estimates that at least during the 1980s and early '90s, hard-to-use computers reduced the nation's productivity gains by one to two percentage points a year - a huge drag. In only one year since 1987 has per-worker output officially risen more than 1 percent. But critics say the official numbers understate productivity.

About 35 to 40 percent of the programming efforts in an information-technology department are spent reconciling duplicate data sitting in various company databases, estimates the Gartner Group, a consulting firm in Stamford, Conn. One observer likens that to taking a car apart to change its oil.

There are several reasons government agencies and corporations get tangled up in their own computer modernization programs. One is the increasing complexity of the systems themselves. Another is that computers installed during the 1950s and '60s are now reaching an age when they have to be replaced. A third is that information-technology departments often struggle under budgets and time-tables that are much too tight.

"When you try to revamp the information-technology part of your company, it turns out to be expensive, time consuming, and more trouble than you could ever imagine," says Ray Bennett, an expert at telephone giant Ameritech in Chicago.

The real problem, argues Mr. Kreitzberg, is the failure to design software in a way that normal people can use. A big remedy would be to incorporate more real-world testing during software design, he adds. "We know how to solve the problem.... When you do these things you cut your costs enormously."

In the case of the IRS, its computer problems won't affect most taxpayers, says Arthur Gross, the tax agency's chief information officer. Its main systems continue to process 200 million returns a year.

The failures are in computers that deal with exceptional cases. Installed in the 1980s and '90s to automate the process, these computers bog down in cases involving several tax issues, such as under-reporting income one year and not filing a return another year. An IRS agent might have to use several databases and even two or more computers to research the case, then move to yet another machine to enter results. This poor communication among machines makes the system "dysfunctional," says Mr. Gross, who was appointed last year to fix it.

Instead of trying a massive overhaul, the agency is now gravitating toward a piecemeal approach, which it will propose to Congress this spring and summer. "That is not an overnight solution," Gross warns. "We have no dates or estimated costs."

We want to hear, did we miss an angle we should have covered? Should we come back to this topic? Or just give us a rating for this story. We want to hear from you.