Expert systems and KM are a natural team
Like ballgames, peanuts and Cracker Jack
Although many knowledge management efforts have incorporated artificial intelligence into their processes, the potential synergy between the two fields has not yet been achieved. Areas in which artificial intelligence have benefited KM include help desks and loan application software, which are usually expert systems, autonomous agents and natural language search engines. However, the strengths of each discipline could be more effectively used by the other, to the benefit of both.
Knowledge management was raised as a key issue several years ago at the Third World Congress on Expert Systems, held in 1997 and sponsored by the International Society for Expert Systems. In a report on the meeting published by the American Association for Artificial Intelligence, Jay Liebowitz notes that presenters recognized the need for companies to use knowledge management techniques and to consider knowledge asset management. However, he points out that CEOs face challenges in measuring and valuing knowledge.
Expert systems are an ideal way to convert both tacit and explicit knowledge into a form that is available to many users, a key process in knowledge management. The implementation of expert systems involves systematic and well-established procedures for representing the knowledge of experts, a process referred to as knowledge engineering. Yet knowledge management efforts often struggle with eliciting and documenting tacit knowledge. At the same time, data warehouses, a mainstay of knowledge management, are a valuable source of information that may not always be used effectively in expert systems.
Many of the large, complex and expensive expert system authoring tools of the mid-1980s have fallen by the wayside. Those that remain are more modest in cost and relatively easy to use. Among them is Exsys, produced by the company of the same name which was introduced in 1983. Exsys has been used to develop thousands of expert systems, and the company plans to introduce a Web-enabled product called Corvid later this year. Exsys and Corvid are each priced at about $12,000.
In 1993, the Occupational Health and Safety Administration (OSHA) began developing Expert Advisors, a series of expert systems based on Exsys. The Advisors addresses health and safety issues in areas such as asbestos and fire safety. Impetus for the initial product originated from the Small Business Administration's Office of Chief Counsel for Advocacy, which urged OSHA to find new ways to help small businesses in interpreting government regulations. OSHA has worked with trade associations, unions and government agencies to respond to suggestions for expert systems in a variety of fields.
Users of Advisors answer questions about their work place, practices, materials and other topics. Advisors determines the hazards that are present and the OSHA regulations that apply. In addition, Advisors can generate legally sufficient plans of action for implementing appropriate procedures and handle administrative tasks such as preparing required letters.
But why use an expert system tool rather than just programming the decision tree?
"A key reason," says Edward Stern, facilitator for Expert Advisors, "is that an expert system shell allows us to lay out the logic of the regulations so anyone can see it clearly (without a lot of programming code). When we issue an Advisor, we need to get approval from a range of people, from technical staff to lawyers." Without a clear presentation of the underlying rules on which the system is based, it would be virtually impossible to get concurrence from the various parties. In addition, an expert system tool allows the subject matter experts, rather than programmers, to control the development of the product and to change it.
The series has attained the hoped-for goals by capturing the knowledge of the most experienced staff, some of whom have left OSHA.
"One of our senior epidemiologists is now at the Mine Safety and Health Administration," says Stern, "and an expert in lead is now at the Housing and Urban Development (HUD). Nevertheless, we have not lost their insights and understanding of the regulations."
The OSHA Expert Advisors program was one of 25 entries (from an initial group of 1,300 entries) selected in August as finalists in the Innovations in American Government awards program of the Ford Foundation and the Harvard JFK School of Government.
Selecting the right problem is essential for success. Dustin Huntington, president of Exsys, points out that the problem must be bounded and have real-world answers, but the answers do not need to be absolute.
"Expert systems can also handle uncertainty," Huntington notes, "and this capability distinguishes them from simple tree diagrams." For example, an expert system designed to help consumers select a camcorder could come up with the best match based on user input, even when the request was not a perfect match to any single product. It would not have to come to a dead-end and report that no product was a fit, as a simple database search might. That produces a much better emulation of the interaction that would occur with a human salesperson. Expert systems can, however, call up information from databases, which allows new information such as prices changes to be reflected in the decision without changing the rules.
A good application for large enterprises would be an expert system that advises employees on what their benefits will be when they retire. An expert system can integrate the rules for employee benefits with information from a database such as salary and length of employment. The availability of an expert system would save time for HR staff, which might be answering nearly identical questions over and over, and would improve employee access to that information.
Product configuration is also a good application. One of the earliest uses of expert systems was in medical diagnosis, and work continues in that arena.
Among the other currently available expert system development tools is XpertRule, from U.K.-based. Attar Software. XpertRule has been used in systems throughout the world to perform such functions as detection of in-store fraud in retail stores and guidance for contamination control at NASA. A companion product, XpertMiner, detects patterns in data and provides a graphical environment for developing the data mining process. XpertRule is priced at $1,995 for the development tool, with various options available for runtime licenses.
The Haley Enterprise offers Eclipse, a rule-based expert system, and Case-based Problem Resolution (CPR), a case-based tool. Both products are geared toward knowledge management solutions and have applications in customer relationship management (CRM). Eclipse, for example, has been incorporated into IBM's Visual Banker, a sales force automation solution that assists financial institutions in intelligent cross-selling and other CRM functions.