Chapter 18 – Expert and decision support systems

Contents

18.1  Overview
18.2  Introduction -- Experts in the Loop
18.3  Expert Systems
18.3.1  Uses of Expert Systems
18.3.2  Architecture of an Expert System
18.3.3  Explanation Facility
18.3.4  Dialogue and UI Component
18.3.5  Examples of Four Expert Systems
18.3.5.1  Example 1: MYCIN
18.3.5.2  Example 2: PROSPECTOR
18.3.5.3  Example 3: DENDRAL
18.3.5.4  Example 4: XCON
18.3.6  Building an Expert System
18.3.7  Limitations of Expert Systems
18.4  Knowledge Acquisition
18.4.1  Knowledge Elicitation
18.4.1.1   Unstructured Interviews
18.4.1.2   Structured Interviews
18.4.1.3   Focused Discussions
18.4.1.4   Role Reversal
18.4.1.5   Think-aloud
18.4.2  Knowledge Representation
18.4.2.1  Expert System Shells
18.4.2.2  High-level Programming Languages
18.4.2.3  Ontologies
18.4.2.4  Selecting a Tool
18.5  Experts and Machine Learning
18.5.1  Knowledge Elicitation for ML
18.5.1.1  Acquiring Tacit Knowledge
18.5.1.2  Feature Selection
18.5.1.3  Expert Labelling
18.5.1.4  Iteration and Interaction
18.5.2  Algorithmic Choice, Validation and Explanation
18.6  Decision Support Systems
18.6.1  Visualisation
18.6.2  Data Management and Analysis
18.6.3  Visual Analytics
18.6.3.1  Visualisation in VA
18.6.3.2  Data Management and Analysis for VA
18.7  Stepping Back
18.7.1  Who Is It About?
18.7.2  Why Are We Doing It?
18.7.3  Wider Context
18.7.4  Cost--Benefit Balance
18.8  Summary

Glossary items referenced in this chapter

accuracy, automation bias, backward reasoning, Bayes Theorem, black-box machine learning, bootstrapping, certainty factors, chatbot, clustering, COMPAS, computer vision, concept learning, confidence rating, connectionist model, constraint satisfaction, constraints, cost–benefit, data reduction, decision support system, decision tree, declarative knowledge, deep neural network, DENDRAL, depth first search, dialogue, dialogue component, dialogue!mixed control, dialogue!system control, dialogue!user control, domain-independent knowledge, domain-specific knowledge, down-sampling, ECG , evidence-based medicine, expert knowledge, expert system, expert system!brittleness, expert system!hybrid, expert system!limitations, expert system!meta-knowledge, expert system!purpose, expert system!rule tracing, expert system!shell, expert system!verification, explainable AI, explanation component, extensional representation, false negative, false positive, forward reasoning, generate and test, genetic algorithm, genetic programming, ground truth, heuristic evaluation function, high-level programming languages, human labelling, hybrid, hybrid AI, hybrid architecture, hybrid expert system, hybrid expert systems, inductive learning, interactive visualisation, Jaccard similarity, knowledge acquisition, knowledge base, knowledge elicitation, knowledge elicitation!critiquing, knowledge elicitation!focused discussions, knowledge elicitation!post-task walkthrough, knowledge elicitation!role reversal, knowledge elicitation!seed questions, knowledge elicitation!structured interview, knowledge elicitation!teach-back interview, knowledge elicitation!think aloud, knowledge elicitation!twenty questions, knowledge elicitation!unstructured interview, knowledge engineer, knowledge representation, knowledge representation!by networks, Licklider, J.C.R., Lisp, logic, logic rules, machine learning, MYCIN, natural language processing, network visualisation, neural network, ontology, ontology editors, OPS5, overlearning, OWL, parallel coordinates, pattern recognition, physical constraints, Poplog, precision, probability, production rules, production system, Prolog, PROSPECTOR, Protégé, Python, Query-by-Browsing, random forest, rapid serial visualisation, RDF, recall, ROC, rule induction, scrutability, search space, search strategies, semantic network, semi-supervised learning, SHRDLU, similarity measure, simulated annealing, SQL, sub-symbolic systems, sufficient reason, symbolic machine learning, symbolic regression, synergistic interaction, synthetic data, tacit knowledge, uncertainty, unconscious bias, unsupervised clustering, unsupervised learning, user interface, visual analytics, visualisation, XCON