Semantics for the intuitive governance of autonomic user‐centered pervasive computing environments

Abstract

As Pervasive Computing environments become more richly populated with sensors and computing power it may be possible for the environment to observe and interpret user actions and pre‐emptively provide accurate support for those actions. This requires accurate inference of user intent. Though individual user acts may be inferable from environmental context sensing, the inference of intent for which appropriate support might be offered is more challenging. Numerous researchers have used probabilistic techniques such as Bayesian Analysis techniques to attempt such inference. Unlike a desktop environment, however, pervasive computing environments are extremely heterogeneous, so any heuristics used must be tailored to the physical environment and the users in question. Thus these techniques are only likely to be accurate if configured with accurate knowledge of routine user behavior. Many existing approaches attempt to learn such knowledge from operational data, but this often requires user involvement in training and expert configuration of probabilistic processing structure. In this paper we examine a complimentary approach where users are presented with an intuitive interface to support the direct configuration of probabilistic structure by users with appropriate knowledge. By treating an intent inference system for a particular pervasive computng environment as an autonomic system, we approach the problem as the design of an intuitive governance interface for this system. We then present the design and usability evaluation of this governance interface.