App Usage Predicts Cognitive Ability in Older Adults
Mitchell L. Gordon, Leon Gatys, Carlos Guestrin, Jeffrey P. Bigham, Andrew Trister, Kayur Patel · 2019 · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3290605.3300398
Summary
This paper investigates how older adults use smartphones differently from younger adults and whether those differences can be explained by cognitive function. The researchers collected three months of iPhone usage data from 84 healthy older adults (aged 61-76), logging 494,641 app launches and 186,968 phone unlocks. Each participant also completed the Cogstate Brief Battery (CBB), a neurocognitive assessment measuring psychomotor function/attention and learning/working memory. The study compared older adults' usage patterns against published data from younger users, finding that older adults use fewer apps (50.3 vs. 64.0 average), keep apps open 37% longer (121.2 vs. 88.6 seconds), send fewer messages (37.2 vs. 67 per day), use social networking apps half as frequently, and shift their usage earlier in the day. The researchers introduced the concept of "app dynamics" — analyzing sequential app usage within sessions through bi-gram analysis to reveal how users switch between apps (e.g., Camera to Photos, Messenger to Facebook). To isolate the role of cognitive ability, they conducted a quasi-experiment comparing "cognitively young" older adults (those scoring within one standard deviation of younger adult norms on CBB tests) against age-matched "cognitively normal" older adults.
Key findings
Cognitive ability explained a surprisingly large proportion of app usage differences between older and younger adults: 67% of the difference in number of apps used, 76% of the difference in app duration, and 79% of the difference in micro-usage breakpoint. A Gradient Boosted Trees classifier could predict whether an older adult was "cognitively young" with 83% ROC AUC for learning/working memory, using features including micro-usage breakpoint, app switch duration, and usage by time of day. Cognitively young older adults used more apps (59.5 vs. 47.9), kept them open for shorter durations (96.2 vs. 122.4 seconds), sent more messages (43.7 vs. 33.8 per day), and used their phones more at night (22.1% vs. 27.1% of usage). Notably, cognitively young older adults still differed from younger adults in time-of-day patterns, using phones more in the morning — suggesting this reflects lifestyle rather than cognitive factors. No significant effects of race, education, or employment were found after controlling for cognitive scores, strengthening the cognitive ability explanation.
Relevance
This research has two major implications for accessibility. First, it demonstrates that smartphones could serve as passive screening tools for cognitive decline by monitoring app usage patterns — potentially enabling early alerts to users or loved ones without requiring explicit testing. This non-invasive approach to cognitive health monitoring could be especially valuable given the growing older adult smartphone user base. Second, the finding that older adults are not a monolithic group — "cognitively young" older adults use phones much like younger people — challenges the common practice of designing simplified interfaces for all older users. Instead, the research suggests that adaptive interfaces should account for the full spectrum of cognitive function, providing simplified experiences only for those who need them rather than imposing them on all older adults. For developers, key design considerations include supporting longer app sessions, accommodating slower switching between apps, and recognizing that the number of apps used correlates with cognitive attention and curiosity.
Tags: older adults · cognitive decline · smartphone usage · mobile accessibility · digital biomarkers · machine learning · aging