### Toward adaptive dual expert and intelligent tutoring systems in medicine: a case study for spinal injuries diagnosis

John Halloran, Theodoros N. Arvanitis

Traditionally, a classical AI model of intelligence has informed the design of Expert Systems and Intelligent Tutoring Systems for medical applications. This paper\footnote{A revised version of this paper, entitled Toward dual adaptive expert and intelligent tutoring systems: case study in medical AI'' appears at the Proceedings of the International Conference on Cognitive Systems, 1996.} argues that such a model of intelligence produces educationally atypical results in that it intrinsically reduces the capacity of such systems to deliver an educational product -- the effective learning of a medical domain -- with human-like efficiency. Systems based on an adaptive model of intelligence, one which emphasizes intelligence as {\bf adaptation to dynamic environments} and as {\bf distributed}, can help (a) transform Intelligent Systems into more human environments, and (b) establish {\bf educational robustness} -- the reliability, durability, and utility of educational software in a given educational sphere. The paper provides evidence for this claim through the discussion of a case study: an adaptive dual system for teaching diagnosis of spinal injuries. The system contributes to the establishment of general principles for the design of adaptive educational software for medical diagnosis. Central to these are that an adaptive system functions as a cognitive model; and that the concept of the {\bf Intelligent Agent} is likely to be importantly implicated.