Four Easy Pieces: Development Systems for Knowledge-Based Generative Instruction

The Experimental Advanced Design Advisor (XAIDA) is a system for the development of computer-based maintenance training. XAIDA acquires knowledge of a device from a subject matter expert and applies common maintenance-training procedures to generate interactive training from the description. XAIDA provides maintenance training in four areas: the physical characteristics of a device, its theory of operation, operating and maintenance procedures, and troubleshooting. XAIDA relies on an instructional device known as a transaction shell, an instructional procedure applicable to particular instructional objectives of a specific type. XAIDA employs a different transaction shell for each of the four abovementioned areas, and each shell employs a knowledge structure appropriate to the shell. Semantic networks represent physical characteristics and procedures; causal reasoning schemes represent theory of operation; and fault trees represent troubleshooting. Each shell provides a browser that is used to present knowledge to the student and a practice environment that promotes skill acquisition under the guidance of an intelligent tutoring system. Subject-matter experts create device descriptions using a WYSIWYG knowledge acquisition system that makes for extremely efficient development. This paper describes each shell in detail, summarizes the research that has been done, and describes where XAIDA stands in relation to other knowledge-based authoring tools. RAPID DEVELOPMENT SYSTEMS FOR KNOWLEDGE-BASED GENERATIVE INSTRUCTION This paper describes a research and development effort aimed at making it easy for subjectmatter experts to develop Interactive Courseware (ICW). By “easy” we mean that courseware development should be as effortless as, say, spreadsheet development. We aim for development efficiency on the order of ten hours of development for each hour of instruction. Our ideal subject-matter expert is a maintenance technician with basic computer skills and no experience in ICW development. The ICW that we seek to produce draws on the best techniques available in the field, including conventional computer-based instruction, multimedia, exploratory environments, and intelligent tutoring systems. It adds to existing instructional media the same value that a knowledgeable instructor brings to instruction. Our approach is grounded in knowledge-based, generative instructional methods, and, in particular, Merrill’s (1993) notions of instructional transactions and transaction shells. A transaction is an instructional procedure for meeting a particular instructional objective such as being able to tie one’s shoe or answer questions about the characteristics of an automobile engine. A transaction shell is a generic form of a transaction. It might, for example, describe a procedure for instruction in step-by-step procedures in general or for associative learning in general. Transactions are produced by adding particular subject matter knowledge (e.g., the steps in tying one’s shoe or the characteristics of an automobile engine) to a transaction shell. Transaction shell theory is therefore founded on the hypothesis that one can factor subjectmatter knowledge and instructional procedures. We do not claim that the hypothesis is original with Merrill. However, his ideas were the immediate inspiration for XAIDA. Some sense of the ha l-0 01 97 33 6, v er si on 1 14 D ec 2 00 7 Author manuscript, published in "International Journal of Artificial Intelligence in Education (IJAIED) 10 (1999) 1-45"