What exactly are expert systems? What are the reasons for the recent surge of interest in these systems? How can these systems be applied to accounting and with what benefits? What are the names of some currently popular expert systems? Will these systems eventually make accountants obsolete? And, finally, are there potential problems in the adoption of these systems? This article attempts to answer these questions.

Definition and Description

Expert systems are a branch of artificial intelligence--a field of science and technology that has been developing over the past 30 years. They are essentially computer programs that emulate human thinking processes in problem-solving situations. The goal of expert systems is to arrive at the same results that a specific human mental process would produce, in a computer program.

Expert systems contain the following four components: knowledge base, inference engine, user interface and explanation facility. The knowledge base consists of facts in a specific topic area, and rules for using these facts. The rules are usually presented in the form of "if...then" statements. For example, a rule may state that "if it's raining, then put on the raincoat." Many of the rules are really "rules of thumb" developed by the human expert (or group of experts) from his or her experience in the area. These rules are called heuristics and are "extracted" from the expert by knowledge engineers who place them into the knowledge base.

The inference engine is a computer program that uses the information in the knowledge base. It drives the system by drawing an inference from relating user-supplied facts to a knowledge base rule and then proceeding to the next fact and rule combination. This process forms a chain in which the "then" part of one rule forms a link to the "if" part of the next rule, eventually reaching a conclusion. The following would be a typical example of a chain:

Rule 1: If cash collection and record keeping are not done separately, then close supervision is a necessary control.

Rule 2: If close supervision is necessary, then auditor inspection of the supervising process is needed.

Rule 3: If auditor inspection is needed, then increased substantive testing in the area is needed. (This is the conclusion).

The above is an example of forward chaining which reasons forward toward a conclusion. Backward chaining reasons backward from the goal to determine if the goal can be reached.

The user interface is the bridge between the inference engine and the user. It provides a communication facility between the two. Finally, the explanation facility helps the user understand why certain conclusions were reached.

Expert systems differ from traditional decision support systems (DSS) in that the latter mainly uses computation in algorithms (step-by-step formulas) while the former relies on rules of thumb and the "gut instincts" of the expert.

Recent Interest in Expert Systems

Over the past five years there has been a growing interest in expert systems. They are used to interpret medical test results, diagnose car problems and determine the causes of telephone line failures. Therefore, accounting academicians and practitioners have been seeking to apply this body of knowledge to accounting.

In addition, there has been a growing interest in applying expert systems to auditing due to the increading complexity of the audit environment, and increased competition between CPA firms.

Development of Expert Systems

There are two ways in which an expert system can be developed: custom development, or the use of an expert system shell. Custom development involves finding an expert in the given area, using the services of a knowledge engineer to extract the expertise and heuristics from this person, quantifying this information in a computer program, and finally, validating and testing the program.

Finding an expert in a given area of accounting is not a problem. Extracting the knowledge from the expert, however, is much more difficult. A key obstacle to understanding the expert's decision process is that the expert is often unable to explain how he reaches decisions. They are often reached intuitively and subconsciously. Accordingly, researchers have turned primarily to sciences such as psychology to help them extract this knowledge. A typical approach requires the expert to think out loud while solving problems. His words would be taped and then analyzed by the knowledge engineer.

Traditional programming languages such as FORTRAN and PASCAL have been used to program the expert's knowlege, but they are being replaced by specialized languages such as LISP and PROLOG, which were developed specifically for artificial intelligence applications.

Custom development of expert systems is an expensive and time consuming process. An alternative is the use of expert system shells-- commercial software packages that consist of a preprogrammed expert system with an empty knowledge base. Such a shell can perform different applications when the knowledge base is filled in differently by different users. This operation takes less time and costs less than custom development.

General Benefits and Applications

Many benefits can be derived from the implementation of an expert system in auditing. These include the following.

. Efficiency. Expert systems can significantly reduce the employee time required to plan and carry out an audit.

. Distribution of Expertise. Since human experts are in short supply, expert systems provide the means of using knowledge in a wide number of locations without actually having the human expert on the premises.

. Quality Control. Expert systems encourage consistent and uniform performance of professional tasks. Not only can expert systems provide guidance for various tasks, they can also help ensure that the key questions in performing a procedure have been addressed. This ensures quality control.

. Education and Training. A significant investment of time and money is spent on the training of an auditor. Expert systems that provide professional recommendations against which developing auditors could test their judgments and sharpen their skills, would significantly reduce the investment that need be made.

Specific Applications of Expert

Systems

Auditing. The auditing field is a prime candidate for the implementation of expert systems. Such systems can aid in the study and evaluation of internal control, perform analytical review procedures and interpret their significance, and relieve the problems of "standards overload" by assisting auditors with solutions to complex reporting and disclosure questions.

There are several audit expert systems in existence today. Some have already reached the market while others are in the prototype stage. The Internal Control Monitor (TICOM) analyzes the flow of documents and effectiveness of controls, recommends additional controls, and points out weaknesses. Using TICOM, an auditor can review an internal control system through observation and interviews. The auditor first completes a traditional review of the system and then employs TICOM to model the system. TICOM is superior to traditional internal control evaluation methods because the evaluation is more extensive and the documentation of the system is more thorough.

AUDITOR is an expert system that assists the auditor in evaluating the adequacy of the allowance for bad debts. It contains approximately 25 primary rules in its database.

EDP-XPERT assists Computer Audit Specialists in judging the reliability of controls in advanced computer environments. AUDITPLANNER is used for the auditor's planning-stage materiality judgments. ARISC models an auditor's internal control evaluation process in the purchasing-accounts payable-cash payments cycle. Finally, Arthur Young's AY/Decision Support is currently in use for audit planning at Arthur Young.

Taxation. The application of expert systems to the taxation field is relatively new. Several public accounting firms are monitoring research activities in this area and the IRS has established an advisory group to examine applications in the compliance area.

Auditors in the field are principally concerned with tax accrual computations and give little attention to tax planning. Coopers & Lybrand has developed an expert system called Expert Tax to guide the auditor and tax accountant through the process of gathering the necessary information for tax accrual and tax planning. It has the ability to point out the importance of the information being requested by providing explanations for its questions. This system was developed by extracting the expertise of 30 senior tax and audit professionals. Its knowledge base contains over 3,000 rules.

CORPTAX aids the tax accountant in evaluating the tax consequences of stock redemptions under various IRS Code sections. TAXADVISOR is used for estate planning and contains 275 rules derived from standard estate planning books.

Management Accounting. Expert systems can be used in the areas of information systems development and technological implementation. These areas deal with the design and development of the overall management information systems and thus have direct implications for management accountants. The role of these accountants would be to develop the knowledge base of the expert system.

Education. Another area of use for expert systems would be the education and development of management accountants. The system can be used to train new personnel by teaching them the rules and reasoning used by experts in the field.

Since the system of internal control and management's credit-and- collection policies both fall within the realm of the management accountant's interests, the AUDITOR and TICOM expert system mentioned earlier should also be of great use.

Potential Problems of Expert Systems

One issue that concerns the accounting profession is the potential adverse effect expert systems could have on the CPA's image. One of the highly respected traits of the CPA is the ability to draw conclusions about subjective information. An expert system, however, reduces this process to a more objective and structured analysis by encoding the thinking in a program and then simply feeding it through the computer. Will this cause the public to view the CPA's work as less subjective and creative?

Another issue is the impact on quality control. In general, to lower the risk of undetected auditing errors, the auditor's work is subject to checking and review by colleagues. If an expert system's work is not reviewed by a human, will the risk of undetected errors become too high? And in the opposite situation, where the computer chooses one course of action and the human another, will the firm find itself in a difficult legal position should the human be proven wrong?

Finally, will humans rely so much on expert systems that they will never develop skills equal to those of the original human experts?

Expert systems have not been in use long enough to give definitive answers to these questions. More time must pass before we can determine if these fears have any basis in fact.

Obsolescence of the Human

Accountant

A question that many people, especially students and junior accountants, are asking today is: Will expert systems replace the human accountant? The answer is no. Accountants will always be needed to develop the knowledge bases and interact with the expert system in order to develop information that will be useful in the decision-making process. Machines, as opposed to humans, cannot think creatively. They require step-by-step logical instructions which they follow mechanically. They are only as good as the programmers and experts. In contrast, humans employ abstract reasoning and intuitive feelings to recognize relationships and patterns among variables. As explained previously, it is very difficult for knowledge engineers to capture this reasoning because the humans themselves often have difficulty explaining or understanding their own thinking processes.

In addition, there are often disagreements among the experts themselves as to the proper path to take, just as two doctors may disagree about a diagnosis. These disagreements tend to hinder the development of expert systems since it is unclear whose opinion to follow.

Another reason expert systems will not readily replace creative thinking is cost. There are three categories of costs involved in creating an expert system: 1. Knowledge extraction cost. Paying the expert and the knowledge engineer for their services. 2. Programming costs. Paying the programmer to encode the knowledge in computer language. 3. Maintenance costs. Correcting errors in the programs and updating the databases to reflect new information.

Each one of these costs is high and taken together they are huge. Accordingly, it is highly unlikely that such systems will ever completely replace human accountants.

The Future Outlook

A limiting factor during the early 1980s in the growth of expert systems was the unavailability of low cost, high powered personal computers with large primary and secondary memories. The introduction in 1988 of IBM's line of PS/2 machines together with the powerful new OS/2 operating system can only serve to accelerate the development and growth of expert systems. In addition, the availability of integrated expert system environments (already available) which integrate expert systems with spreadsheets and word processors, should whet the appetite of accounting professionals.

It is no longer a question of whether or not expert systems will be used in accounting. They are already here. And there is great potential for continued, rapid growth of these systems. Accounting professionals who wish to stay abreast of current developments and remain competitive should plan accordingly.

Ethics and Auditing with Expert

Systems--"Never the Twain

Shall Meet?"

Article I of the AICPA Code of Professional Conduct states: "In carrying out their responsibilities as professionals, members should exercise sensitive professional and moral judgements in all their activities" (emphasis added). It is important to distinguish that judgement is required at two levels--professional and ethical (moral).

But what is an expert system and of what relevance are ethics considerations? Expert systems are a branch of artificial intelligence in which computer programs are designed to emulate the thinking processes of human experts to solve problems in specialized areas. A 1987 AICPA study (An Introduction to Artificial Intelligence and Expert Systems) stated:"...they hold considerable immediate and long-term promise for the accounting profession." This area is considered by many to be the future of computer auditing. The accounting profession has always been receptive to advances in computer technology and the emergence of expert systems has not changed this, although the technology is just evolving and actual applications in use are very limited. However, expert systems have many potential applications for the accounting profession, such as engagement planning.

A primary component of an expert system is the knowledge base--it contains the facts and rules that are used when making decisions on specific topics. To date, the facts and rules in developing systems have focused on professional, and not on ethical, considerations, since the primary focus has been on solving technical problems.

The general public sees the accountant as someone of professional stature, integrity and morality. One could argue that the entire basis of the profession rests on its ethical behavior, and Article III of the Code of Professional conduct states: "To maintain and broaden public confidence, members should perform all professional responsibilities with the highest sense of integrity." It further states: "Integrity is measured in terms of what is right and just."

Although one could argue extensively about a proper definition of ethics, "what is right and just" will be the accepted definition here for ethical and moral judgment. How then, is the professional to provide assurance that the decisions of an expert system consider proper ethical judgments? One could argue that is the accountant's task, but if true expert systems evolve, system decisions may be considered final. Challenging a sophisticated computerized system decision based solely on ethics might well be perceived as antiquated thinking.

Additional problems are likely to occur in the way expert systems are developed. Without intending offense to any individual or profession, computer experts and academic researchers, who are the most likely developers of true expert systems, do not function under the same ethical concerns and constraints as members in public practice.

Related to this concern is the AICPA study's statement: "The nature of expert systems may make it economically feasible for nonaccounting enterprises to actively compete with the accounting profession in the future." Will these nonaccounting enterprises have the same high level of concern for ehtical standards as is currently evidenced by the accounting profession?

If the development and implementation of expert systems is viewed as inevitable, as it likely is, how then is the profession to move forward technologically while maintaining its ethical integrity? Perhaps the most straightforward answer is to become involved now in the development phase and to incorporate in any and all knowledge bases overriding ethical and moral rules.

Clearly, however, written rules cannot be the entire answer--our society has regularly evidenced its penchant for finding loopholes. However, all systems must constantly address the question of whether anything is improper in fact or appearance.

Interestingly enough, in Kipling's "Ballad of East and West" the twain shall never meet until the "great Judgement Seat." Although as a profession, divinity is not the goal or purpose of accountants, moral and ethical judgements are a constant and a hallmark. There is no reason for the profession not to embrace new technology as a tool for its operational efficiency, but there is reason in the case of expert systems for it to be concerned about, be involved in, and research possible ethical considerations and constraints in their development.

What's New

New Accounting Software Series

MAS 90 EVOLUTION/2 is a new series of accounting software designed to take full advantage of 286-and 386-based microcomputers and multi-user oriented operating systems. More than 100 new features have been incorporated in this system recently announced by State of the Art, Inc.

Some of the new features include: . New user interface with menu bars and pull-down windows; . Integrated backup and restore commands for all industry standard operating systems; . User-selectible screen color preference, keyboard handling and terminal/monitor presentation; . Support for all industry standard printer configurations including laser printers; . High level password security; . Faster response times for data entry programs; . Deferred report printing to eliminate printer contention; . Context-sensitive Help screens; . Expanded import/export capabilities; . Extensive list and search features; . More powerful custom report generators.

Source code and programming tools are integral to the new product, enabling VAR's and consultants to provide customization and additional industry specific features required by clients. MAS 90 EVOLUTION/2 applications include: General Ledger, Accounts Receivable, Accounts Payable, Payroll, Inventory Management, Sales Order Processing, Purchase Order Processing, Job Cost, and Bank Reconciliation. MAS 90 EVOLUTION/2 State Of The Art, Inc. 3545 Howard Way Costa Mesa, CA 92626-1418 415- 845-4716 . Requirements: Systems are grouped into three categories and are machine class specific. Group 1 consists of single user DOS, DOS Network and SCO Xenix versions. Group 2 consists of versions for Altos 286 and 386 computers. Group 3 will include several Unix-based applications. . Price: Most Group 1 modules are individually priced at $795. Job Cost is $995 and Bank Reconciliation is $195.