Math Content Readability, Student Reading Ability, and Gaming the System

Nominated for Best Paper at EDM 2025, exploring how reading ability and content readability affect gaming-the-system behavior in adaptive math learning software.

Overview

This project is based on my first-author paper at the International Conference on Educational Data Mining (EDM) 2025, which was nominated for Best Paper.
Our research examined how math content readability and student reading ability interact to influence gaming the system behaviors in Carnegie Learning’s MATHia adaptive software.

Research Questions

  • Are students with lower predicted reading ability more likely to game the system?
  • Do readability characteristics of math problems predict gaming behavior?
  • Can improved readability interventions reduce gaming the system?

Methods

  • Readability Metrics: 32+ features including word count, sentence length, Shannon entropy, and Flesch Reading Ease.
  • Analysis: Combined observational + experimental data; statistical modeling and correlation analyses.

Key Findings

  • Students with lower predicted reading ability are significantly more likely to engage in gaming behaviors.
  • Lessons with higher readability challenges saw more gaming the system across both reader groups.
  • Experimental results showed that readability-enhanced lessons reduced gaming for stronger readers (non-ER), but not for students with reading difficulties.
  • Readability factors like word count, entropy, and Flesch Reading Ease emerged as key predictors of gaming.

Impact

This research bridges psychometrics, NLP, and learning analytics, showing how non-math factors like readability interact with engagement in adaptive math learning.

Citation:
Khanna, P., Mathieu, K., Norberg, K., Almoubayyed, H., & Fancsali, S. E. (2025). Math Content Readability, Student Reading Ability, and Behavior Associated with Gaming the System in Adaptive Learning Software. In Proceedings of the International Conference on Educational Data Mining (EDM 2025).