A comparative analysis of methods for expert systems

A Multi-Component Framework for the Analysis and Design of Explainable Artificial Intelligence

The rapid growth of research in explainable artificial intelligence (XAI) follows on two substantial developments. First, the enormous application success of modern machine learning methods, especially deep and reinforcement learning, have created high expectations for industrial, commercial, and social value. Second, the emerging and growing concern for creating ethical and trusted AI systems, including compliance with regulatory principles to ensure transparency and trust. These two threads have created a kind of “perfect storm” of research activity, all motivated to create and deliver any set of tools and techniques to address the XAI demand. As some surveys of current XAI suggest, there is yet to appear a principled framework that respects the literature of explainability in the history of science and which provides a basis for the development of a framework for transparent XAI. We identify four foundational components, including the requirements for (1) explicit explanation knowledge representation, (2) delivery of alternative explanations, (3) adjusting explanations based on knowledge of the explainee, and (4) exploiting the advantage of interactive explanation. With those four components in mind, we intend to provide a strategic inventory of XAI requirements, demonstrate their connection to a basic history of XAI ideas, and then synthesize those ideas into a simple framework that can guide the design of AI systems that require XAI.

Computer expert system for the histopathologic diagnosis of salivary gland neoplasms

The design, development, and testing of a prototype interactive histopathologic expert system capable of diagnosing 15 types of primary salivary gland neoplasms is described. The system incorporates a multiple subprogram modular design and makes use of multiple reasoning methods including: data-driven and goal-directed rule-based reasoning, linear pattern recognition, and Bayesian classification. Its user interface incorporates both a "hypertext" context-sensitive information assistance facility and the video display of stored and digitized photomicrographic images. The system can report a differential diagnosis of its findings with assessment of its confidence in its diagnosis. The system's performance was evaluated in a series of tests. The results of a weighted kappa analysis of the system's diagnoses versus those of four oral pathologists for 20 salivary gland neoplasms indicated no statistical difference in diagnostic performance between the system and the human experts and each of the experts in relationship to the others (Wilcoxon rank sums test). A modified version of Turing's test of artificial intelligence demonstrated no statistically significant difference in the system's diagnoses versus the diagnosis of four human expert pathologists (Fisher's exact test). The knowledge and experience gained in the development and testing of the expert system described in this study have demonstrated the validity of histopathologic diagnostic expert systems in a selected area of oral pathology.

A methodology for evaluation of knowledge-based systems in medicine

Evaluation is critical to the development and successful integration of knowledge-based systems into their application environment. This is of particular importance in the medical domain--not only for reasons of safety and correctness, but also to reinforce the users' confidence in these systems. In this paper we describe an iterative, four-phased development evaluation cycle covering the following areas: (i) early prototype development, (ii) validity of the system, (iii) functionality of the system, and (iv) impact of the system.

Diagnosis of selected pulpal pathoses using an expert computer system

The diagnosis of dental pain that results from a pulpal pathosis may prove to be a confusing and complex issue for both dental students and experienced clinicians alike. A computerized diagnostic expert system, COMENDEX, was developed to aid in the clinical diagnosis of pulpal pathosis and to provide a rapid, accurate second opinion when a human consultant is not readily available. The vast majority of diagnostic expert systems use a single reasoning methodology as their inference mechanism. COMENDEX is a prototype hybrid expert system that combines and exploits the best aspects of both rule-based and statistical reasoning methodologies. The COMENDEX endodontic diagnostic system was tested to determine the validity and accuracy of its diagnoses by the use of a variant of Turing's test and weighted kappa statistic. The promising result obtained from the initial tests suggests that an expert system using this type of hybrid reasoning methodology is well suited to the area of endodontic diagnosis and may prove highly successful if expanded to include other problem areas in oral diagnosis.