Math Learning Environment with Game-Like Elements and Causal Modeling of User Data
Educational games intend to make learning more enjoyable, but at the potential cost of compromising learning efficiency. Therefore, instead of creating educational games, we create learning environment with game-like elements: the elements of games that are engaging. Our approach is to assess each g...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Others |
| Published: |
Digital WPI
2011
|
| Subjects: | |
| Online Access: | https://digitalcommons.wpi.edu/etd-theses/722 https://digitalcommons.wpi.edu/cgi/viewcontent.cgi?article=1721&context=etd-theses |
| Summary: | Educational games intend to make learning more enjoyable, but at the potential cost of compromising learning efficiency. Therefore, instead of creating educational games, we create learning environment with game-like elements: the elements of games that are engaging. Our approach is to assess each game-like element in terms of benefits such as enhancing engagement as well as its costs such as sensory or working memory overload, with a goal of maximizing both engagement and learning. We developed different four versions of a math tutor with different degree of being game-like such as adding narrative and visual feedback. Based on a study with 297 students, we found that students reported more satisfaction with more 'game-like' tutor but we were not able to detect any conclusive difference in learning among the different tutors. We collected student data of various types such as their attitude and enjoyment via surveys, performance within tutor via logging, and learning as measured by a pre/post-test. We created a causal model using software TETRAD and contrast the causal modeling approach to the results we achieve with traditional approaches such as correlation matrix and multiple regression. Relative to traditional approaches, we found that causal modeling did a better job at detecting and representing spurious association, and direct and indirect effects within variables. Causal model, augmented with domain knowledge about likely causal relationships, resulted in much more plausible and interpretable model. We propose a framework for blending exploratory results from causal modeling with randomized controlled studies to validate hypotheses. |
|---|