Cost/benefit based adaptive dialog: case study using empirical medical practice norms and intelligent split menus

The notion of an adaptive user interface, one that accommodates user needs based on knowledge of the task at hand, is compelling but difficult to make practical. This paper examines models of the utility (as balancing of cost and benefit) in the initiation of task-specific dialog based on conditional probability of user goals in context. Illustrations in this paper are based on an empirical model of General Practice (GP) medicine as derived from a large database of GP/patient encounters. Application is explored with respect to generation of dynamic pick-lists (intelligent split menus) of diagnoses in the context of patient reason for encounter (RFE); medication warning dialogs are also briefly addressed. Simulated data entry demonstrates the effectiveness of various adaptation options. Findings include significantly superior accuracy per item of varying-length cost/benefit based pick-lists as compared to fixed length pick-lists. We conclude with discussion of the power of empirical cost/benefit based adaptation to build user engagement and with future directions toward application in clinical practice information systems.