Age? It's in the Game: An Exploratory Study on Detection of Cognitive Aging through Card Games

Abstract

We report on an exploratory study for assessing cognitive aging based on the acquisition and modeling of player data of commercial off-the-shelf games. To this end, candidates for digital biomarkers of cognitive performance were captured via FreeCell, from three distinctive age groups (18-25, 40-55, 65+) from 52 participants, playing for a total of 130 game rounds. Next, features were engineered and a machine learning model (Logistic Regression) was trained. We found features retained for the model to support theories on fluid intelligence and cognitive functions sensitive to cognitive aging. Performance metrics suggest that our model is successful in classifying young and older participants. However, classifying middle-aged players remains problematic. To conclude, this study suggests that commercial off-the-shelf games hold promise for the acquisition of digital biomarkers on cognitive aging and provides benchmark data for future studies. Nevertheless, as this is a first, exploratory study, further research is necessary.