CAREER: Supporting the Intelligibility of Context-Aware Applications

Carnegie Mellon University, Pittsburgh PA

Investigators

Abstract

The maturing discipline of ubiquitous computing (ubicomp) has largely ignored usability in lieu the more fundamental quest to simply develop and deploy compelling applications. In these projects usability and social analysis is often post-hoc, one-off, and not accreting into a uniform body of scientific knowledge. This project spotlights the core ubicomp usability challenge as "intelligibility." That is the mapping of mental models, the user's explanations and predictions of a system's expected behavior, to the actual behavior of context-aware ubicomp application. While the use of mental models has been applied to well-understood systems, its application to context-aware systems is novel and promising. It is critical to understand how mental models develop, what sorts of explanations will promote correct mental model formation and what feedback will support users in making correct predictions about application behavior. Broader Impact: The combination of the fundamental understanding gleaned from the proposed studies with the resulting widely disseminated toolkit, has the potential to enable a new class of context aware applications and related usability metrics. Improved usability of and trust in context-aware applications should reduce the chance of abandonment by making it clear how these applications behave. Easier to understand systems may engage a broader cross-section of end users, advancing the adoption of context-aware systems. The PI has demonstrated that ubicomp design and evaluation projects such as this engage undergraduates from groups under-represented in computer science research.