Technology is an ever changing tool driven by compliance requirements as well as entity-centric needs to satisfy market demands. For compliance requirements, IT deployments tend to be reactionary rather than a continuous, proactive process. Consequently, IT compliance efforts are typically lacking constancy and conformity. To combat this tendency, IT planners should focus design and transition efforts on three time frames for meeting entity needs: the current state, the near-term state, and the long-term state of compliance requirements. Within this context, expert systems can be an invaluable tool to implement mandates that satisfy immediate needs and simultaneously position the entity to effectively meet the next potential compliance issue.



Expert System Development Activities 

IT usually pervades all organizational formations pursuing effective and efficient processing in response to compliance requirements, thus facilitating better decision-making through various information delivery mechanisms and offering opportunities for business model development that may lead to value creation as well as competitive advantages. To construct an expert compliance system, a knowledge engineer, performing a function similar to a system or business analyst, is typically needed. A designated knowledge engineer is responsible for defining issues in manageable terms, soliciting the knowledge, skills and abilities of experts, and translating these talents into electronically encoded formats. 

Expert system development is usually a four step process. It starts with the knowledge engineer obtaining an understanding of a particular judgment issue. It is followed by the acquisition of thought processes of experts in solving the issue. Next, a computer model is programmed to reproduce the adopted thought processes of defined situations; if a shell program is unavailable. Lastly, the system is tested and certified to ensure appropriate resulting decisions and usability. These steps are commonly known as: knowledge representation, knowledge acquisition, computational modeling, and model validation. 

Populating the Expert System 

Several methods exist for a knowledge engineer to obtain knowledge. One option is to go through textbooks and professional journals with the intent to extract definitions, axioms, and rules that apply to the issue. This type of knowledge acquisition is especially useful for teaching and reference situations because question-response paths are direct. However, how the question is posed to the expert system can lead to misleading results. Another method of acquiring knowledge is to ask human experts to explain their thought process and method for solving problem scenarios, sometimes referred to as verbal protocol analysis. Lastly, a human expert can enhance the information obtained from literary resources and often bring unpublished knowledge, gained through experience, to the decision process paths. As a result, this combinational knowledge makes human-based expert systems a valuable technology.

To incorporate human expert knowledge into a technology-based expert system, the right individuals must be identified and selected. Specialists tend to be trained in rather narrow domains and are best at solving problems within their defined domains. Assuming experts do exist and are willing to participate; good experts are those who are able to solve particular types of problem scenarios that most others cannot solve with the same efficiency and/or effectiveness. Additionally, considerable time can be saved in developing an expert compliance system if the knowledge engineer has experience in the area being modeled. 

After experts have been selected, the knowledge engineer must take the expert knowledge and transform it into a computational model. However, issues may arise because an expert discovers that they are unable to describe how a situational scenario is resolved. Typically, this is due to experts operating at a subconscious-level while performing some tasks to address a scenario. Considering the possibility of undefined steps generating misaligned logic paths in the inference engine, commonly, interdisciplinary teams of specialists must work in unison to formulate deductive reasoning processes for defined problems. 

To assist in assessing decisional acumen, most managers are under observation for situational responses impacting the entity. Therefore, information reliability is critical. During the final stage of preparation for deployment, an expert system has to be validated to ascertain reliability and scope of decisional processes. In the model validation step a knowledge engineer and/or IT assurance professional identifies errors, omissions and mistakes in the knowledge base. Furthermore, since the constructed system is designed to simulate an expert’s decision making process, it should be tested against opinions of subject matter experts. Lastly, if the system is later updated to keep the knowledge base current, model reevaluation is necessary to ensure continued decisional reliability. 

Summary 

From a technical perspective, the typical expert system can be divided into two essential parts: the knowledge base and the inference engine. The knowledge base contains the body of knowledge, or set of facts and relationships, obtained from the knowledge acquisition phase. The rules associated with a knowledge base tend to be heuristic and take the form of conditional statements. Whereas, the inference engine is a collection of computer routines that control the system paths through the knowledge base to enable recommendations. In addition, the inference engine serves as a bridge between the knowledge base and user. 

Methodologically, the knowledge engineer defines the ambit of issues that the purposed system will address because one logic path too broad may result in a system too difficult to manage and may generate a system crash. Contrastingly, the knowledge engineer must be careful not to limit an issue too much because a logic path too narrow will produce a system so rudimentary that results will be worthless.  Read Full Post