Clearing up fuzzy logic - Forum - Column

March 1991 marked a significant anniversary for the computer industry. In that month, 40 short years ago, commercial computing was born with the installation of the Univac I computer at the U.S. Census Bureau.

Since then, tremendous technological advances have been achieved. The evolution of computer speeds and applications have outpaced even the wildest imaginings of its inventors. Yet, the industry seems to lag behind the desires of corporate management and users to deliver comprehensive and cost-effective information systems.

Year after year, and survey after survey, there are concrete indications that business management considers computerized applications to be too expensive to build, too difficult to maintain and too limited in scope.

The primary bottleneck is software. While today's hardware speeds, even on desktop computers, nearly boggle the mind, software technology remains unable to allow for automation of many critical applications. One major reason is that the most commonly-used computer languages (Cobol and newer fourth generation languages) are able to accommodate only discrete logic. (That is, "If A is greater than B, do C," or "If A is equal to B, do C.")

Discrete logic, however, comes up short when trying to automate business applications where the thought process of the expert decision-maker cannot be expressed exclusively in discrete terms. Expert system languages, which employ discrete logic in their rules, suffer a similar fate. They are still far too restricted in their ability to emulate expert decisions, where a response depends on judgment and "shades of grey."

But hope is on the horizon. A technology called fuzzy logic is already being successfully applied to the automation of such physical devices as cameras, elevators and subway cars. Japan, in particular, has used fuzzy logic to achieve spectacular success in building advanced automated physical systems. However, the potential of fuzzy logic is just beginning to be explored for its application in business information systems in the United States and the world.

MIMICS HUMAN DECISION-MAKING

Fuzzy logic is a technology that allows for the introduction of subjective terms and linguistic variables into computer language. By allowing the use of such expressions as "high," "very high" and "extremely high," for example, fuzzy logic offers the first real opportunity to develop applications in which the computer can emulate the way in which people actually make decisions, rather than restricting users to defining problems and solutions in discrete terms that a computer would understand.

According to Dr. Lofti Zadeh, a professor at the University of California at Berkeley who is credited with coining the term fuzzy logic, "99.9% of all human reasoning is approximate in nature." While some professionals may quarrel with the extent of that assertion, it seems clear that Dr. Zadeh is largely correct.

And therein rests the underlying reason for the difficulty that is encountered in building management information systems (MIS) to meet today's business requirements. Traditional computing languages simply cannot accommodate approximate reasoning.

When I entered the nascent field of computer software in 1964 working with the Univac I, the nature of the "business problem" was very different from what is experienced today. Then, the challenge was to convert punched card/tabulating equipment applications that had programmable plug boards to general-purpose programmable computers with stored programs. These tabulating applications were, by nature, very restricted. Essentially, they dealt only with sorting, data reduction, field selection, summarization and reporting. Therefore, their conversion to general-purpose computers was relatively straightforward.

Even in the '70s and '80s, businesses concentrated on automating business applications that were clerical and administrative in nature and for which the logic was well-known and generally discrete. Computer languages like Cobol, Fortran and Basic, in addition to higher-level fourth generation languages, were still, for the most part, adequate to meet the discrete needs of the business user.

But the next stage--the next era of real progress--will come from being able to capture and automate business applications in which the decisions require judgment and human reasoning. Fuzzy logic will improve the quality of the applications that have already been automated. But its greatest contribution will be in allowing us to build systems that, until now, could not be automated at all.

Fuzzy logic will allow insurance underwriters to build "rules" for risk assessment and enable the medical profession to weigh the differing opinions of multiple experts in forming diagnoses. It will let police department accept descriptions of criminal from multiple witnesses and "process" those descriptions from the witnesses' points of view. It will let corporate financial experts weigh the relative merits of mergers and acquisitions based on a combination of objective and subjective merits, rather than on discrete variables alone.

Until now, we have been limited to expressing business problems in high/low, on/off and either/or terms. But fuzzy logic's ability to accommodate approximate human reasoning will change that. Without question, fuzzy logic represents a major breakthrough for the computer industry and, consequently, for businesses of almost every kind.

Goetz is president of Goetz Associates, a management consulting firm located in Teaneck, N.J. He received two of the first patents ever issued by the U.S. Patent Office.

COPYRIGHT 1992 Wiesner Publications, Inc.
COPYRIGHT 2004 Gale Group