Machine Learning Analysis of Engagement Behaviors in Older Adults With Dementia Playing Mobile Games: Exploratory Study
Bibliographic record
Abstract
Background: The prevalence of dementia is expected to rise with an aging population, necessitating accessible early detection methods. Serious games have emerged as potential cognitive screening tools. They provide not only an engaging platform for assessing cognitive function but also serve as valuable indicators of cognitive health through engagement levels observed during play. Objective: This study aims to examine the differences in engagement-related behaviors between older adults with and without dementia during serious gaming sessions. Further, it seeks to identify the key contributors that enhance the effectiveness of machine learning for dementia classification based on engagement-related behaviors. Methods: This was an exploratory proof-of-concept study. Over 8 weeks, 20 older adults, 6 of whom were living with dementia, were enrolled in a single-case design study. Participants played 1 of 4 "Vibrant Minds" serious games (Bejeweled, Whack-A-Mole, Mah-jong, and Word-Search) over 8 weeks (16 30-min sessions). Throughout the study, sessions were recorded to analyze engagement-related behaviors. This paper reports on the analysis of the engagement-related behaviors of 15 participants. The videos of these 15 participants (10 cognitively intact, 5 with dementia) were analyzed by 2 independent raters, individually annotating engagement-related behaviors at 15-second intervals using a coding system. This analysis resulted in 1774 data points categorized into 47 behavior codes, augmented by 54 additional features including personal characteristics, technical issues, and environmental factors. Each engagement-related behavior was compared between older adults living with dementia and older adults without dementia using the χ² test with a 2×2 contingency table with a significance level of .05. Codes underwent one-hot encoding and were processed using random forest classifiers to distinguish between participant groups. Results: Significant differences in 64% of engagement-related behaviors were found between groups, notably in torso movements, voice modulation, facial expressions, and concentration. Including engagement-related behaviors, environmental disturbances, technical issues, and personal characteristics resulted in the best model for classifying cases of dementia correctly, achieving an F1-score of 0.91 (95% CI 0.851-0.963) and an area under the receiver operating curve of 0.99 (95% CI 0.984-1.000). Conclusions: Key features distinguishing between older adults with and without dementia during serious gameplay included torso, voice, facial, and concentration behaviors, as well as age. The best performing machine learning model identified included features of engagement-related behavios, environmental disturbances, technical challenges, and personal attributes. Engagement-related behaviors observed during serious gaming offer crucial markers for identifying dementia. Machine learning models that incorporate these unique behavioral markers present a promising, noninvasive approach for early dementia screening in a variety of settings.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.007 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.000 | 0.001 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".