Teach or not teach? Impacts of Objective Explanation in Game-based Learning: Insights from Gaussian Hidden Markov Analysis

Peng Zhang, Hanyu Su, Jialing Zeng, Yuanyuan Zhang, Junjie ShangProceedings 2024 · DOI

TL;DR

This study investigates whether explicitly teaching learning objectives in game-based learning (GBL) improves student outcomes. Using Gaussian Hidden Markov Models (G-HMMs) to analyze student interaction data, they found that teacher explanations help students enter and maintain more efficient learning states, though overall spatial ability gains were similar across groups.

What Problem Does This Paper Address?

Game-based learning (GBL) is a popular educational tool, but its effectiveness hinges on how well students understand the learning objectives embedded within the games. While well-designed games can integrate objectives, it's unclear whether explicit instruction from a teacher is necessary or beneficial. This paper tackles the question of whether teachers explaining objectives in GBL enhances student learning outcomes and processes compared to students learning objectives solely through reading or implicit discovery within the game.

Methods

The core of this study's methodological innovation lies in its use of Gaussian Hidden Markov Models (G-HMMs) to analyze student learning processes. G-HMMs are a statistical technique that allows researchers to infer unobserved (hidden) states of a learner (e.g., "efficient learning," "struggling") from sequences of observable actions (e.g., game attempts, time spent, stars earned).

The researchers collected game log data, which served as the "observations" for the G-HMM. By modeling these sequences, they could then analyze the transitions between different learning states for students who received explicit objective explanations versus those who did not. This approach provides a dynamic, process-oriented view of learning, complementing traditional pre- and post-test outcome measures.

Datasets

The study used data from a single educational game, Cube Elimination, designed to improve spatial abilities. The game features three maps with 33 levels.

  • Student Interaction Data: Logged from the game, including attempt number, level ID, timestamp, star rating, time to complete, hint views, and probe clicks. This data was used as observations for the G-HMM analysis.
  • Spatial Ability Test: Ekstrom's (1976) Mental Folding Ability Test was administered as a pre-test and post-test to measure students' spatial cognition.

Key Insights and Ideas

  • The Power of Explicit Instruction: Even in a game-based environment, direct teacher explanation of learning objectives can significantly influence how students learn, guiding them towards more efficient engagement patterns.
  • State Transitions Matter: Analyzing the dynamics of student states (e.g., how quickly they enter an efficient state, how long they stay there) provides deeper insights than just looking at final outcomes.
  • Efficiency is Key: Teacher explanations seem to help students quickly adopt and maintain an "independent" or efficient learning state, suggesting a reduction in time spent in less productive states.

Key Results

  • RQ1: Similar Overall Gains: There were no statistically significant differences in mental folding abilities (measured by the pre- and post-tests) between the group that received teacher explanations and the control group that learned objectives through reading. This suggests that both approaches, while different in process, led to comparable improvements in spatial skills.
  • RQ2: Improved Learning States: The experimental group (with teacher explanations) showed a higher likelihood of maintaining their learning states, particularly the "independence" state (84% self-transition probability). This indicates that once students in this group entered an efficient learning mode, they were more likely to stay there. In contrast, the control group exhibited more dynamic state transitions, suggesting their learning states were more susceptible to immediate game challenges.

Limitations and Future Work

The authors acknowledge several limitations:

  • Single Game and Limited Cohort: The study used only one game and a specific group of fourth-grade students. This limits the generalizability of the findings.
  • Future Work: The authors suggest future research should:
  • Employ a wider variety of games.
  • Include learners from diverse grade levels and cultural backgrounds to improve generalizability.
  • Further explore the interplay between teacher explanations and game design elements.

Summary updated

BibTeX
@inproceedings{zhang2024teach,
  author = {Peng Zhang and Hanyu Su and Jialing Zeng and Yuanyuan Zhang and Junjie Shang},
  title = {{Teach or not teach? Impacts of Objective Explanation in Game-based Learning: Insights from Gaussian Hidden Markov Analysis}},
  booktitle = {Proceedings},
  year = {2024},
  doi = {10.22318/icls2024.997303},
}
Figure