‘Expert systems’ architecture:
Introduction:
In recent years, the systems design process has been witness to a tremendous leap within advanced system design and, in the next few years, certain professionals will be introduced to the enormous potential that ‘artificial intelligence’ (AI) in business can offer. This will be particularly true, as projects under development come to fruition.
One of the greatest limitations of current information processing systems is that their scope has been restricted to the fast and accurate processing of numeric and text data. Broadly speaking, this processing has involved numeric and algebraic functions on numbers, or various forms of insertion, deletion and retrieval of text. This process is controlled largely, by operating system memories and by programs working on the numeric and text data. The accountant added debits and credits to arrive at balances, the scientist performed statistical tests on survey data, or the office administrator inserted or deleted files in cabinets.
The doctor having a knowledge of diseases, comes to a diagnoses of an illness by reasoning from information given by the patient's symptoms and then prescribes medication on the basis of known characteristics of available drugs together with the patient's history. The solicitor advises the client on the likely outcome of litigation based on the facts of the particular case, an expert understanding of the law and a knowledge of the way courts work and interpret this law in practice. The accountant looks at various characteristics of a company's performance and makes a judgement as to the likely state of health of that company. All of these tasks involve some of the features for which computerised systems traditionally have been noted - performing text and numeric processing quickly and efficiently - but, the distinctive feature largely absent from a totally effective system has been 'reasoning'.
Ladies and gentlemen, reasoning is the movement from details of a particular case and knowledge of the general subject area surrounding that case to the derivation of conclusions. Moreover, the primary function of 'expert systems' will be to incorporate this reasoning by applying general rules in a knowledge base to aspects of a particular case under consideration. Such systems, by definition, will be of importance to a whole array of professions including medicine, law, accountancy and science.
In commencing this article, it would perhaps be relevant in defining exactly what an expert system is. The short answer is that it is certainly a computerised system that performs the role of an expert or carries out a task that requires expertise (Dowe, July 2007). In order to understand what an expert system is, then, it is worth paying some attention to the role of an expert and the nature, generally, of expertise. Ascertaining what types of experts and expertise there are in business and what benefits will accrue to an organisation when it develops such a system, are, equally, valid follow-on questions. If we are to design and implement such 'artificially designed intelligence' systems, then we need to understand fully how an expert in human form has traditionally operated. We must try to understand his mindset - in given situations - to develop a system that accommodates his or her needs.
I have an Irish dentist. She is very capable at what she does and very thorough and methodical in her approach. She is also a highly trained professional that utilises and accesses various forms of information, at times, through her databases. Without going into the finite details of what dentists can and do record you will perhaps appreciate the degree to which AI systems could help such practitioners. The analogy could easily be applied to doctors, medical consultants and scientists from varying branches.
Part 1:
AI systems (Artificial Intelligence) are systems that, to a large degree, eliminate ignorance. The basic premise underpinning such designs is that it removes the need for operator, human or clerical error. Artificial intelligence and how such systems are driven is based purely on a body of knowledge, at source, within a particular subject area. Legally, such systems could, by definition, be held-out to be contracting in their own right. This point is perhaps worth considering a little further.
AI should base its access facilities on a high level of security, particularly on who and who cannot gain access to AI. The premise underpinning AI is that information has been fed from authoritative sources and, as such, an outcome driven by an autonomous AI server should be deemed professionally correct and competent. If, at some stage, the system creates errors in driving out that information for advice purposes, that is later found to be wrong or even negligent then, I believe, a liability and obligation exists on an organisation as to why such outcomes were delivered in an incorrect manner.
To deliver AI, in a given situation, is not by any means an easy task but one, if correctly implemented, will have an enormous benefit on reducing costs in the long-run and on a company' or organisations efficiency and productivity. It will likely reduce wastage such as excess manning by delivering information at a touch of a button that previously, would have taken many hours, in given situations, to compile and disseminate. Business efficiency, ladies and gentlemen, is about how organisations gain a competitive edge, not by remaining stale and out-of-date. Although this may very well sound condescending, the chief purpose of any business is to deliver its objectives quickly and effectively, in reducing costs. The only real way that this can be improved, beyond current standards, is by introducing, through a high level skill base, the 'reasoning' frameworks that are, in many institutions, clearly absent. This not only applies to many small type organisations who could benefit hugely from AI, but is equally apparently absent within large UK organisations such as the NHS. How good could AI be for the NHS?
I was first introduced to AI by an Irish gentlemen whose pulse ticked business. Irish businessmen have a clear philosophy of creating effective systems that are in accordance with the objectives of their companies. This is not particularly true with how so many companies in the UK wish to progress. Why this is the case is a little baffling suffice to say that British short-termism around the world is often acknowledged as its main downfall. Cultural diversities in many Irish companies also aids the process of business development: notably within the field of logistics, IT and electronics. AI is as much about integrating diversities and professional competencies as it is to do with the overall delivery of such systems.
Part 2:
Organisations have experts and specialists in many areas. A business organisation will probably have general expertise in such areas as tax, accounts, marketing, production and personnel as well as specific expertise in the narrow area of its activity – for example, the design, production and retail of motor cars. As well as ‘high level’ expertise there will also be ‘low level’ expertise. An example of this might be the knowledge possessed by a clerk of the way that an organisation stores its records with cross-referencing and the handling of exceptions. In general terms however, the author of this site believes that a major step towards AI, is the necessity in today’s era of creating, as far as is possible, a paperless driven system. Added bureaucracy, by definition of acquiring effective streamlining, cannot aid expert systems. Anything that slows or retards the information flow is a bottleneck, unwanted by experts acquiring information from such systems.
Expert systems will require, at some early stage, an appraisal and justification of the costs to be incurred in such systems if the benefits envisaged to the organisation are to be realised at some future date. Expert systems and related architecture does, in broad terms, offer an array of benefits to the organisation. For example, the implementation of such systems could become a cost-effective consultancy system in aiding or even replacing the existing expertise within the organisation. Being computer based, expert systems perform consistently unlike ‘human experts’ who themselves may have ‘off days’. Such infrastructures are also likely to create a standard or benchmark against which human experts can be compared. An archive of special skills that the organisation currently possesses and on which it is dependent for its continued successful functioning: experts may leave or retire from that organisation.
Ladies and gentlemen, any area of expertise should involve the analysis of a complex set of conditions, and the application of these to a specifiable area of knowledge which is amenable to computerised representation. This might not be so easy, as expert systems tend to remove the necessity in applying an element of general knowledge or common sense. Invariably, any system that is robust – even the most technically driven systems – will, at some stage, create an element of conjecture. However, the differential between expert systems and systems that many organisations currently have, is not the same. Expert systems are designed and implemented so as producing outcomes in line with AI. The basic parameters, for instance, will be so markedly different, that future predictions and outcomes will, as a result, differ from how current organisational systems function. Expert systems represent such a paradigm shift in systems philosophy that the onus on such systems implementation in the future will, unquestionably, bring human experts more closer to the systems implementation process. This, for the blog author, is an incredibly exciting time for future systems development because the opinions of many senior professionals from an array of professions will be sought. The conceptual notion of any expert system is to deliver and aid the process for those who ultimately sanction and authorise decisions.
Part 3:
Because of the way that parameters are built into expert systems, such systems should be able to cope and handle uncertain or incomplete information. It may also be able to learn, that is, modify its knowledge base or inference engine.
Expert systems have been gradually developing and are currently being developed for a wide variety of areas of expertise in business. For example, systems which provide advice to employers on the dismissal of employees. Other notable developments worthy of mention include systems which aid auditors in providing an effective and complete audit of a company's accounts or quick and effective assessments of a company's health from various perspectives such as government agencies, the trade creditor, and the financier of loans.
Ladies and gentlemen, the creation of an expert system, particularly from an accounting perspective, can only evolve once an effective and fully tested management information system (MIS) has been installed and can, equally, be relied upon, as a 'decision support system'. Failure in any one of these parts would, without question, render any expert system unreliable. In essence, an expert system seeks to move into a different league but the prerequisites, fundamentally, are linked to company systems that have maintained a competitive edge in terms of economies of scale and how certain levels and types of information can be used for a variety of purposes.
In building an expert system we must look at those individuals who can compliment the construction phase with certain skills. This involves the creation of advanced databases because at the heart of such systems lies the inference engine built on domain specific facts within the knowledge base. Here, we are not talking about basic database operations but more complex matrix's that can act and serve in proliferating information more widely within the system. An expert system seeks to replicate the human expert in whatever area that might be. This means removing inadequacies, or anything that stifles productivity. Expert systems do not take prisoners.
...
© Copyright MarkDowe, 2007: all rights protected
