Assistive Technology (AT) helps people with cognitive impairment (CI) in their use of computers. Studies have found that AT systems are abandoned by CI users at shockingly high rates. One of the major causes of abandonment is an eventual misalignment with: (1) user goals and abilities and (2) the functionality delivered by the system. We use data stream-mining techniques to recognize when a user seems to learn (or forget) desired behaviors that are part of a plan for cognitive rehabilitation. In our case study, CI users are given an email system to aid their cognitive rehabilitation. When a CI user learns (or forgets) a specified behavior, the email system is adapted to meet the changing needs of a user. Our monitoring technology tries to recognize significant behavioral changes and then notify clinicians. The clinicians consider this information, along with other data, to determine when the email system should be adapted and what kind of adaptation should occur.

Dynamically interpreting and adapting therapy plans for individuals currently requires substantial effort from clinicians. With the cognitively impaired population increasing and the number of clinicians decreasing, the need for some automation in therapy analysis is critical. Methods such the ones we have proposed will be a critical factor in addressing the needs of the millions of people with cognitive disabilities.

Committee
Dr. Raj Sunderraman (chair)
Dr. Anu Bourgeois
Dr. Xiaolin Hu
Dr. William Robinson