Unpuzzling tetris : exploring the mechanisms of expertise in a complex, dynamic task with simple machine learning models

Sibert, Catherine
Thumbnail Image
Other Contributors
Gray, Wayne D., 1950-
Fajen, Brett R.
Sims, Christopher Robert
Şimşek, Özgür
Issue Date
Cognitive science
Terms of Use
This electronic version is a licensed copy owned by Rensselaer Polytechnic Institute, Troy, NY. Copyright of original work retained by author.
Full Citation
This thesis aims to expand the study of expertise by closely examining a single task, in this case the video game Tetris. Many studies of expertise and skill focus on high level outcome measures, trying to predict overall performance development, or focus on the low level mechanisms of simple tasks that are incomparable to most real world tasks. Tetris provides an ideal middle ground, by being a complex and dynamic task that involves perceptual learning, decision making, motor skill, and long term planning, while also being simple enough to control in an experimental setting, and to model using basic machine learning principles. In this work, we first explore the range of expertise in Tetris and classify differences in how that skill is expressed. Second, we adapt existing machine learning models of Tetris to produce human-like behavior. Third, we add environmental capabilities and constraints to these models and observe the resulting changes to performance and behavior. Fourth, we use the models to evaluate low level components of the task, in this case individual placement decisions made during a game, to better understand how the final outcome is reached. Finally, we discuss how our results may be used to better study Tetris specifically, how to apply those ideas to similar real-time dynamic decision making tasks in general, and most broadly, how to approach the study of expertise in complex real world task environments.
December 2019
School of Humanities, Arts, and Social Sciences
Dept. of Cognitive Science
Rensselaer Polytechnic Institute, Troy, NY
Rensselaer Theses and Dissertations Online Collection
Restricted to current Rensselaer faculty, staff and students. Access inquiries may be directed to the Rensselaer Libraries.