@article{JMAI4455,
author = {Kyle Leduc-McNiven and Ryan T. Dion and Shamir N. Mukhi and Robert D. McLeod and Marcia R. Friesen},
title = {Machine learning and serious games: opportunities and requirements for detection of mild cognitive impairment},
journal = {Journal of Medical Artificial Intelligence},
volume = {1},
number = {0},
year = {2018},
keywords = {},
abstract = {This perspective paper presents a simple serious game on a mobile platform (Smartphone game). The game has the integrated capability to track a person’s play by storing player metadata on start time, end time, and moves within the game. These data can be analyzed to infer cognitive processes of strategy learning, retention, and recall over a brief period of time for potential future applications in pre-symptomatic assessment of mild cognitive impairment (MCI). Through machine learning (ML), the data are demonstrated to be of utility in providing a “cognitive fingerprint” of play. The ML methods used to classify play use synthetic data generated by robots (bots), ranging from bots playing perfectly to bots playing with various degrees of impairment. The findings include guidance on the volume of data required, as well as the features deemed effective for ML classification of various degrees of bot impairment. The work illustrates several significant considerations when applying ML to simple serious games and the data they can generate.},
issn = {2617-2496}, url = {https://jmai.amegroups.org/article/view/4455}
}