Thinking, Fast and Latent: Exploring Individual Differences in Daily Cognition through Computational Cognitive Modeling

Google map

Abstract: 

Recent innovations in cognitive assessment, especially those leveraging high-frequency, ambulatory digital tools, offer promising avenues for detecting subtle, early cognitive shifts within everyday contexts and demands. Integrating computational modeling further enhances these tools’ precision by disentangling core cognitive features from extraneous signals. In line with these advancements, this study examines the validity of a brief, smartphone-based adaptation of a visual working memory associative binding task (‘Color Shapes’), previously shown to identify preclinical Alzheimer’s disease-related cognitive impairment. A diverse sample of 68 U.S. adults (69% women, 81% White; age range 24-80 years, M = 49, SD = 14) participated in a within-person factorial design, completing 60 trials for each of 16 task variations on smartphones over an 8-day period. A drift diffusion model was fit to Color Shapes response time and accuracy data to isolate key features of the decision-making process. To optimize the task for smartphone-based cognitive assessments, we experimentally manipulated three task properties—study time, probability of change, and choice urgency—to examine their effects on computational features such as evidence accumulation rate, initial response bias, and decision-making caution. Additionally, we assessed the impact of test array size (whole display vs. single probe) on responses across conditions. Our findings suggest that an optimized version of the Color Shapes task may be well-suited for smartphone-based, repeated cognitive assessments in real-world settings, with computational modeling showing potential for enhanced sensitivity in detecting early indicators of Alzheimer’s disease risk. This approach has strong potential for developing tools that support proactive monitoring of cognitive health in at-risk populations.