Instruction as reasoning about multiple objectives

Abstract

This paper describes a three-tier architecture for general intelligent tutoring systems — at the top level are the general objectives and their reasoning that infer, control and explain the instruction strategies; at the middle level is the planning mechanism that resolves the needs for the general behaviour of teaching; and at the bottom level is the execution of methods that reflect the specific behaviour of teaching. We also present a knowledge representation for general instructional objectives and their reasoning mechanism. Two fundamental design issues are addressed: (a) the epistemological adequacy of the objectives — easy to interpret for human and machine alike; and (b) the generality of the objectives — independent of the subject matter. We argue that the proposed architecture and knowledge representation can support an “instructional objectives first” approach which help to develop ITSs efficiently and communicate the studies to other researchers.