Purpose: The goal of our project was to develop a method for tailoring computer interfaces for elders with varying degrees of cognitive impairment.

Methods: There has been a rapid growth in the use of computers and the Internet by older users. Older computer users appear to have higher individual differences in their computing capabilities, and in addition, are at risk for medically related perceptual, cognitive, and motor problems. Clear and simple interfaces with adaptive capabilities for assistance are especially critical for this group of users. Our research has focused on algorithms for inferring a user's cognitive abilities using monitoring data from computer keyboard, mouse interactions, and performance on cognitive computer games. These data are then used to adapt computer interfaces and to provide tailored hints and assistance when needed. We investigated these approaches in a pilot study with a 9 of elders in a senior residential facility. For a period of 3 weeks, we monitored all keyboard and mouse activity, including performance on the research version of our FreeCell computer game. Using our cognitive assessment algorithms, we modified game difficulty and interfaces in the form of hints.

Results: Our assessments of cognitive performance based on the monitoring of computer interactions corresponded with results of standard neuropsychological tests for the 9 participants (mean age 79.5 ± 8.5). Three of the participants were diagnosed with mild cognitive impairment. Average performance scores based on the computer monitoring for the two groups could be separated by a simple threshold. In addition, an even more powerful predictor of cognitive impairment seems to be the variability in performance scores, with high variability being associated with cognitive impairment.

Conclusions: Our approach in monitoring users' interactions with computers focused on both the detection of cognitive changes that can be used for health management as well as for adapting computer interfaces to individual users with varying cognitive capabilities. Although our cognitive metrics were appropriate for determining the optimal computer game difficulty, the feedback on our hints was mixed. Feedback was often seen as a distraction and interruption, even though the user may be struggling with the computer interface. More sophisticated user models are required to successfully tailor computer interfaces to the wide spectrum of elder users.