Affiliation: OHSU Computer Science
Funding Period: 2006 - 2009
Funding Source: BAIC Intel
View related topics
Context-based prompting has the potential to provide effective, timely reminders in many situations, including medication management and coaching situations. However, whatever is not explicitly modeled (or detected) — such as the likelihood of inference error, changes in overall physical activity, gait, or cognitive function — can result in poor inference or render the current coaching strategy ineffective. Such contingencies lead to inappropriate system behavior. These issues cut across project boundaries affecting Medication Management, Coaching, and Unobtrusive Multi-Person Tracking. This project will examine ways of anticipating, detecting, and accommodating behavioral patterns not present in training data, integrate inference uncertainty into the models, develop automated assessment of prompt effectiveness, and model the cost of errors