Expert Systems for Decision Support in Business Start-Up
探讨专家系统如何帮助创业者应对创业初期的复杂决策，整合法律、税务、融资等多领域知识，提高咨询效率。
Can Expert Systems Help in Founding a Business? The need for successful business start-ups is increasingly apparent these days, since new businesses create new jobs and support structural changes in the economy. But the mortality rate of young firms is high. In West Germany, three of four new firms die before their eighth year (Institut der deutschen Wirtschaft 1989). Empirical studies show that many of the problems faced by new firms occur because of poor planning in the start-up phase. In starting up a business the founder is confronted with many problems (e.g., contacting public authorities, understanding state regulations, obtaining financing, formulating a deed of partnership, choosing a legal status, choosing potential suppliers, planning a system of distribution, assessing the market), which may prove too much for him or her. In West Germany there is little academic training of entrepreneurs. Only a few entrepreneurship courses are offered in public institutions, so the entrepreneur has to rely on the support of business consultants, lawyers, tax advisers, banks, or regional development agencies. This is a very unsatisfactory situation for the founder for the following reasons: * the founder wants to have complete support for his start-up, he has to contact several different experts: bankers offer advice on the best possible mixture of public money and private bank loans, tax advisers choose the legal form of a business so as to minimize the tax burden, lawyers draw up the business contract, and business consultants assist in setting up a business plan for the first years. But all these decisions are inter-related. The problem is that all of these experts may give the best advice in their own field but neglect to understand the ramifications this advice has for a related area. For instance, a tax advisor may recommend the legal status for the new business which results in the least tax burden, yet fail to consider the implications this choice has for company law or finance. * It is often difficult for the founder to understand the recommendations of the consultants; for example, the meaning of particular clauses of a deed of partnership are very difficult to understand. (Fuller (1989) discusses how different views of the problem can cause communication barriers between the consultant and his client.) Expert systems may help overcome these problems and in doing so make business consulting more efficient. Expert systems are computer programs that simulate the reasoning of human experts in a certain domain (Chadwick and Hannah 1986). Expert systems differ from conventional programs in that the latter manipulate data while the former present the knowledge (expertise) of experts. Conventional programs work on well-defined problems which can be solved by a deterministic algorithm. Expert systems work on more complex problem domains. For example, obtaining financing via public subsidy programs is a complex problem because of the enormous number of programs and the many details which have to be considered. Expert systems use heuristic procedures which are often supported by the use of probabilities and rules of thumb. Expert systems not only produce a solution but also explain their results (Waterman 1986, Sviokla 1986, Kurbel 1989, Hart 1986). Expert systems also have a different structure than conventional programs (see table 1). The knowledge base contains the knowledge of the expert (facts, heuristics) in the shape of rules (e.g., If there exist special risks within the business, a legal form with limited liability should be chosen.). This knowledge is processed by the inference engine, which draws conclusions and produces solutions. The explanation facility describes to the user why particular decisions are made and explains the steps used to reach them. The dialogue between the user and the expert system is controlled by the user interface in the form of question menus (Schnupp and Leibrant 1986, Kurbel 1989, Barrett and Beerel 1988). …