Measuring Cognitive Abilities in the Wild: Validating a Population-Scale Game-Based Cognitive Assessment

TL;DR

This study introduces Skill Lab, a game-based cognitive assessment tool validated on a large population sample, demonstrating its speed, accuracy, and potential for widespread cognitive profiling outside traditional lab settings.

Contribution

We developed and validated a novel, engaging game-based platform for rapid, ecological assessment of multiple cognitive abilities at a population scale.

Findings

Models accurately predict eight cognitive abilities from in-game behavior.

Game-based assessments are five times faster than traditional methods.

The tool replicates known age-related cognitive decline patterns.

Abstract

Rapid individual cognitive phenotyping holds the potential to revolutionize domains as wide-ranging as personalized learning, employment practices, and precision psychiatry. Going beyond limitations imposed by traditional lab-based experiments, new efforts have been underway towards greater ecological validity and participant diversity to capture the full range of individual differences in cognitive abilities and behaviors across the general population. Building on this, we developed Skill Lab, a novel game-based tool that simultaneously assesses a broad suite of cognitive abilities while providing an engaging narrative. Skill Lab consists of six mini-games as well as 14 established cognitive ability tasks. Using a popular citizen science platform (N = 10725), we conducted a comprehensive validation in the wild of a game-based cognitive assessment suite. Based on the game and validation task data, we constructed reliable models to simultaneously predict eight cognitive abilities based on the users' in-game behavior. Follow-'-up validation tests revealed that the models can discriminate nuances contained within each separate cognitive ability as well as capture a shared main factor of generalized cognitive ability. Our game-based measures are five times faster to complete than the equivalent task-based measures and replicate previous findings on the decline of certain cognitive abilities with age in our large cross-sectional population sample (N = 6369). Taken together, our results demonstrate the feasibility of rapid in-the-wild systematic assessment of cognitive abilities as a promising first step towards population-scale benchmarking and individualized mental health diagnostics.