4 EXPERT SYSTEMS Acta Structilia Desember 1995 THE FIFTH GENERATION - MORE THAN A GENEALOGICAL PURSUIT MARIUS M. MULLER 1 1 Old Mutual Properties*,?. 0. Box 11315, Johannesburg 2000, South Africa • Disclaimer : The views expressed in this article, are those held by the author and in no way reflect the views of his colleagues or employer. Abstract This article describes the nature of Ex-pert Systems and outlines their possible use in the field of building economics. Their usefulness in terms of quantification of uncertainty is also examined. It was found that the benefits of developing Ex-pert Systems are tremendous, but that the success of Expert Systems would be determined by their acceptance by the general public. Keywords: Artificial Intelligence, Building Economics, Expert Systems, Quantification of Uncertainty. 1. INTRODUCTION Sir Clive Sinclair, referring to Artificial Intelligence, stated that " ... ·Fifth Generation' [ computing] is the greatest battle-ground of the century. lfwe lose we are out of the game" (Brandon, 1984). Bowen and Edwards (1985) emphatically admit that no evidence exists to prove that there is a need for this paradigm shift and also that the hunt for a new paradigm emanates almost entirely from an academic pursuit of knowledge. Does this in any way lessen the credibility of Sir Clive's statement that "the 'Fifth Generation' is the greatest battle-ground of the century"? So far the Japanese government has invested US $500 million to this cause and the British government £200 million (Brandon, 1984). This paper, by way of describing the nature of Expert Systems, outlining their possible use in the field of building economics and by commenting on their usefulness in the quantification of uncertainty, will help to decide whether the 'battle' is worth fighting. Acta Structilia 2. NATURE OF EXPERT SYSTEMS December 1995 Expert Systems are a specialised form of Artificial Intelligence which are still in their embryonic stage (Brandon, 1990). The text that follows will describe the nature of Expert Systems. 2.1 Definitions: (Addis, 1984; Brandon, 1983; Fiegenbaum, 1984; Goodall, 1985) Many authors have put forward definitions for Expert Systems. They all include terms which refer to the capturing of human expertise within a system which allows interaction with a person of lesser expertise, so that the latter will be able to solve significant problems by using the knowledge of the former. Naylor (1983) extends this definition by sa)�ng that the capability of the system to justify its own line of reasoning in a manner intelligible to an enquirer is a further desirable characteristic. Buchanan and Duda (1983) encapsulate the fuller concept of an Expert System by defining it as a computer program that provides expert-level solutions or advice to problems and is heuristic (self learning), transparent and flexible. 2.2 Development As mentioned earlier, Expert Systems are specialised forms of Artificial Intelligence (Brandon, 1990). Bowen and Erwin (1991), citing Newton (1986) and Steels (1988), propose two types of Expert Systems, viz. first generation Expert Systems which are founded largely on heuristic reasoning based on rules and second generation Expert Systems which incorporate the above concept as well as deep reasoning based on a model of the __ problem domain, i.e. a self-learning capacity. Jackson (1986), however, proposes a three stage development. Firstly, the Classical period ( l 960's) characterised by "game playing heuristic" models. Secondly, the Romantic (mid 1960's to mid l 970's) involving the data capturing of specific and general principles. Lastly, the Modem period (mid l 970's to date) which leans towards self-learning knowledge based Expert Systems. 2.3 Character The purpose of Expert Systems, among other things, is to increase the understanding of the future, serve as a reference point and to provide a "skeleton for professionals to flesh out"(Brandon, et al 1987). This in tum would lead to further benefits such as enhancement and improvement of expertise, increased consistency in decision making and a speedier decision making process (Brandon, 1988). In addition to this, Bowen and Edwards (1985) see a replication and union of expertise and an improvement in documentation. Most Expert Systems at present are successful in areas involving surface knowledge, i.e. domain dependent knowledge (Basden, 1983), whereas the real success of Expert Systems lies in the area of deep reasoning, which is domain independent (Bowen and Erwin, 1991 ). Expert Systems are by nature very competent. They are able to replicate, although to a limited extent at present, the thinking process of humans (Winfield, 1982). Winfield (1982) and Brandon (1990) contend that Expert Systems overcome the 'black box' effect in that they allow for interaction with the user by allowing the user to question their "chain of reasoning" and to override any assumptions that have been made. This transparency of Expert Systems establishes confidence in the user as it allows actual control over the system. Stockley (1987) maintains that with regard to Expert Systems it is important to realise the limitations of any Expert System when using it so as to avoid failure. Kidd and Weldbank (1984) 5 6 Acta Structilia Desember 1995 state that for an Expert System to be successful it must contain "the richest collection of information possible about the knowledge which the expert uses." 3. USE OF EXPERT SYSTEMS IN THE FIELD OF BUILDING ECONOMICS Brandon (1987) says that traditional cost models are found wanting due to their inherent nature. These inherent problems of inexplicability, unrelatedness and detenninism are well documented by Bowen and Edwards (1985), Beeston (I 987) and Bowen et al (1987). A paradigm shift was therefore needed to overcome these problems (Kuhn, 1970). In the context of this paper, a model is defined as a tool for evaluating, in terms of cost, different design options at a stage early enough to pennit design-to-cost control at the point where control is most needed, i.e. prior to sketch design (Newton, 1987). 3. 1 Purpose and Usage Brandon (1987) sees the Expert System cost model as being to the advantage of society, the client, the employer and the expert. He also maintains that the Expert System will be able to serve the building economist in his decision making process by fulfilling the role of a decision support system. This is echoed by Blanning (1984) and Hamilton (1987), the latter adding that the Expert System should act as a checklist for the practitioner. To be meaningful as cost models for use in building evaluation, Expert Systems should be applicable throughout the building procurement process (Bowen et al, 1987), should conform to a positive design/data/model/output interface (Raftery, 1987) and should be iterative. The iterative concept is supported by Higgin and Jessop (1964) and Morrison and Stevens (1980). 3 .2 Problems and Developments Gray (1987) says that due to the complexity and specialist nature of the building industry, Expert Systems have had limited success. This confirms Rittel's (1966) statement that construction problems are "wicked" and that human input cannot be excluded. According to Birrell (1980) these problems can be overcome. He proposes that there are rules by which construction work is executed. This would thus accommodate the development of a rule-based system. / Bowen et al (1987) has decided to follow Beeston's ( 1983) approach in opting to find a solution via the implementation of realistic models, as opposed to 'black box' models (Brandon, 1982). They, however, add that this resource based model must take variability and uncertainty into account and will be based on PERT/Network type modelling. Erwin and Bowen (1990) in reply to Gray's (1987) argument that networks cannot be accommodated by a rule-based system, contend that a script-based knowledge representation system can overcome this problem. A major problem with decision making however, is the fact that different experts have subjective views as to which solution is best (Brandon, et al 1988). Mathur and Heah (1990) approach this problem with the solution of a multi-domain system which negotiates compromises. The solution is realised by developing a "conflict solving multiple domain system". Expert Systems will play an important role in the field of building economics in the future. It must be remembered that these models are only in their infancy stage and tremendous development still needs to occur. It is the author's opinion that ways should be found to favourably introduce Acta Struc:tilia December 1995 Expert Systems into the industry. This can be achieved by making the systems more user-fiiendly and by structuring them along lines which are familiar to practitioners in the industry. 4. THE USEFULNESS OF EXPERT SYSTEMS IN THE SPECIFICATION OF UNCERTAINTY. Existing cost models according to Higgin and Jessop ( 1965), Harris and Wijesundera ( 1986) and Bowen and Erwin (1989) lack the ability to handle uncertainty satisfactorily. Uncertainty may be defined as a state of knowledge about the variable inputs to an analysis (Marshall, 1988) and is brought about due to the objectives, constraints and consequences of possible actions not being precisely known (after Bellman and Zadeh, 1970). 4.1 Types of Uncertainty According to Mamdani and Efstathiou (1985) no one technique exists that adequately treats all forms of uncertainty, viz. imperfect knowledge, intrinsic randomness, inherent indeterminancy and categorical uncertainty as is referred to by many authors (Bellman and Zadeh, 1970; Zadeh, 1979; Buchanan, 1982; Tong, 1982; Pang et al, 1986; Klir, 1987; Scott et al, 1988; Allwood, 1989 and Ng and Abramson, 1990). 4.2 Handling Uncertainty Traditionally, uncertainty was handled by way of techniques such as decision analysis, sensitivity analysis, input estimation, etc., but these were found to be lacking in their practical applicability (Marshall, 1988). Bowen and Edwards (1985) propose that to overcome problems of quantifying uncertainty, there needs to be the development of a model . which encapsulates Stochastic (random) simulation. Mathur (1982) also takes this approach and suggests that Expert Systems can find their true niche in the development of models dealing with stochastic variability and uncertainty. 4.3 Expert System-based Approaches to Uncertainty According to Erwin et al (1991), Expert Systems provide many built-in features for the handling of uncertainty, as well as functions which facilitate calculations with probabilities. Uncertainty features of Expert Systems can be classified into two main areas. Firstly, the ability to handle uncertainty of outcome. This concerns the aspect of conditional probability (Scott et al, 1988) and forward chaining (Erwin et al, 1991 ). Secondly, the ability to handle uncertainty of explanation. This refers to methods such as explanation description (Waterman, 1986), backward chaining (Holsapple and Whinston, 1987) and abduction (incomplete reasoning) (Giarratano and Riley, 1989). Therefore, according to Erwin et al ( 1991 ), "various Expert System based techniques for handling uncertainty exist" and they "appear suitable for use in a construction context". 7 8 Acta Structilia Desemher 1995 5. CONCLUSION It can thus be concluded that although empirical data for the need for Expert Systems does not largely exist, the benefits of developing Expert Systems are tremendous. The main benefit is that expertise will be unlocked for access by all thereby allowing more competition between "monopolising" professional bodies (Brandon, 1984). The success of Expert Systems, however, lies mainly in the acceptance by the general public of the ability of Expert Sy�ems. REFERENCES Allwood, R.J. 1989. Techniques and Applications of Expert Systems in the Construction Industry. Ellis Horwood Ltd., London. Basden, A. 1983. On the application of expert systems. International Journal of Man­ Machine Studies, 19, pp. 461-477. Beeston, D. 1987. A future for cost modelling, in Building Cost Modelling and Computers, edit P.S. Brandon. pp. 17-24, Spon, London. Bellman, R.E. and Zadeh, L.A. 1970. Decision-making in a fuzzy environment. Management Science, 17(4), pp. 141-164. Birrell, G.S. 1980. Construction planning - beyond the critical path. Journal of the Construction Division, American Society of Civil Engineers 106, C03, pp. 389-405. Blanning, R.W. 1984. Management Application of Expert Systems. Information & Management. Bowen, P.A. and Edwards, P.J. 1985. Cost modelling and price forecasting: practice and theory in perspective. Construction Management and Economics, Vol.3, pp. 199-215. Bowen, P.A. and Erwin, G.J. 1989. Expert systems in building design evaluation. South African Journal of Science, 85(4), April, pp. 251-257. Bowen, P.A. and Erwin, G.J. 1991. Second generation expert systems, machine learning, and the building procurement process. South African Journal of Science, Forthcoming. Bowen, P.A., Wolvaardt, J.S. and Taylor, R.G. 1987. Cost modelling: a process-modelling approach, in Building Cost Modelling and Computers, edit P.S. Brandon, pp. 387-395, Spon, London. Brandon, P.S., Basden, A., Hamilton, I. and Stockley, J. 1988. Application of Expert Systems to Quantity Surveying, RJCS (Quantity Surveyors Division). Brandon, P.S. 1984. Computers: Friend or Foe? Address to the 13th Triennial Conference, Royal Institution of Chartered Surveyors (Q.S.Division), April. Acta Structilia December 1995 Brandon, P.S. 1982. Building cost research - need for a paradigm shift, in Building Cost Techniques: New Directions (Ed. P.S. Brandon), E. & F.N. Span. London, pp. 5-13. Brandon, P.S. 1987. Models, mortality and the ghost in the machine, in Building Cost Modelling and Computers (Ed. P.S. Brandon), E.& F.N. Span, Ltd London, pp. 3-14. Brandon, P.S. 1990. Expert systems - after the hype is oYer, in Proceedings of the C.I.B. W- 55/65 International Symposium on Building Economics and Construction Management, Sydney, March, Vol. 2, pp. 314-344. Buchanan, B.G. 1982. New research on expert systems, in Machine Intelligence (Eds. Hayes, Michie and Pao), Ellis Horwood Ltd., Chichester, Vol. I 0, pp. 269-299. Erwin, G.J. and Bowen, P.A. 1990. Expert systems: script-based knowledge representation in building design evaluation, in Proceedings of the C.I.B. W-55/65 International Symposium on Building Economics and Construction Management, Sydney, March, Vol.2, pp. 462-472. Erwin G.J., Bowen, P.A. and Strez, H.A. 1991. The treatment of uncertainty in the building procurement process using expert system-based techniques, in Proceedings of the Conference on, Construction Project Modelling and Productivity and Fourth Yugoslav Symposium on Organisation and Management in Construction, Dubrovnik, April. Giarratano, J. and Riley, G. 1989. Expert Systems - Principles and Programming. PWS-Kent Publishing Company, Boston. Goodall, A. 1985. The Guide to Expert Systems. Learned Information (Europe) Ltd, England. Gray, C. 1986. 'Intelligent' construction time and cost analysis. Construction Management and Economics, Vol. 4, pp. 135-150. Gray, C. 1987. Expert system development for predicting time of construction at the early design stage, in Building Cost Modelling and Computers (Ed. P.S. Brandon), E. & F.N. Span Ltd., London, pp. 429-440. Hamilton, I. 1987. Developing expert systems for management applications, in Building Cost Modelling and Computers (Ed. P.S. Brandon) E. & F.N. Spon Ltd., London, pp. 441-451 . Harris, F.C., Wijesundera, A. 1986. Using Expert Systems in Construction. Construction Computing, July, 417. PSB. Higgin, G. and Jessop, N. 1964. Interdependence and Uncertainty. Tavistock Institute, London. Higgin, G. and Jessop, N. 1965. Interdependence and Uncertainty. Tavistock Institute, London. Holsapple, C.W. and Whinston, A.B. 1987. Business Expert Systems. Richard D. Irwin Inc., New York. Jackson, P. 1986. Introduction to Expert Systems. Addison-Wesley: London. 9 10 Acta Structi/ia · Desember 1995 Klir, G.J. 1987. Where do we stand on measures of uncertainty, ambiguity, fuzziness, and the like. Fuzzy Sets and Systems, Vol.24, pp. 141-160. Kidd, A. and Welbank, M. 1984. Knowledge acquisition. Expert Systems: State of the Art Report 12:7, Pergamon Infotech Limited, England, pp. 71-80. Kuhn, T. 1970. The Structure of Scientific Revolution (2nd Edit). University of Chicago Press. Mamdani, E.H. and Efstathiou, H.J. 1985. Higher-order logics for handling uncertainty in expert systems. International Journal of Man-Machine Studies, Vol. 22, pp. 283-293. Marshall, H. 1988. Techniques for Treating Uncertainty and Risk in the Economic Evaluation of Building Investments. National Institute for Standards and Technology, NIST Special Publication, No.757, September. Mathur, K. and Heah, K. 1990. Expert Systems in building design a multiple domains approach, in Proceedings of the C.I.B. W-55/65 International Symposium on Building Economics and Construction Management, Sydney, March, Vol. 2, pp. 514-524. Morrison, N. and Stevens, S. 1980. A construction cost data base. Chartered Quantity Surveyor, June, pp. 313-315. Naylor, C. 1983. Build Your Own Expert Systems. Sigma Technical Press, London. Newton, S. 1984. System for Setting on Economic Strategy for Building Design in Commercial Property: An Interim Report. Portsmouth Polytechnic, Department of Surveying. Newton S. 1987. Computers and cost modelling: What is the problem? In Building Cost Modelling and Computers, Edit P.S. Brandon, pp. 41-48 Spon, London. Ng, K. and Abramson, B. 1990. Uncertainty management in expert systems. IEEE Expert, April, pp. 29-48. Pang, G., Bigham, J. and Mamdani, E.H. 1987. Reasoning with uncertain information. IEEE Proceedings, Vol.134, Part D, No.4, pp. 231-237. Raftery, J.J. 1987. The state of cost/price modelling in the U.K. construction industry : a multi criteria approach, in Building Cost Modelling and Computers (Ed. P.S. Brandon). E. & F.N. Spon. London, pp. 49-71. Rittel, H. 1966. · The Universe of Design'. Paper, University of California. Scott, I., Gronow, S. and Rosser, B. 1988. The nature and use of uncertainty in property valuation expert systems. Journal of Valuation, Vol.7, pp. 218-247. Steels, L. 1988. Second generation expert systems, in Principles of Expert Systems, IEEE Press, pp. 381-389. Acta Structilia December 1995 Stockley, J. 1987. Knowledge acquisition for expert systems, in Building Cost Modelling and Computers (Ed. P.S. Brandon), E. & F.N. Spon Ltd., London, pp. 477-489. Tong, R.M. 1982. The representation of uncertainty in DISCIPLE. Aids Technical Note, TN1018-1 Waterman, D.A. 1986. A Guide to Expert Systems. Addison-Wesley Publishing Company, Reading, Massachusetts. Winfield, M.J. 1982. Expert Systems : an introduction for the layman. Computer Bulletin, Vol.7, December. Zadeh, L.A. 1979. A theory of approximate reasoning, in Machine Intelligence (Eds. Hayes, Michie and Mikulich), Ellis Horwood Ltd., Vol.9, pp. 149-194. 11