Supporting Upper Elementary Students in Learning AI Concepts with Story-Driven Game-Based Learning

Anisha Gupta, Seung Y. Lee, Bradford W. Mott, Srijita Chakraburty, Krista D. Glazewski, Anne T. Ottenbreit-Leftwich, J. Adam Scribner, Cindy E. Hmelo-Silver, James C. LesterAAAI 2024 · DOI

Introduction

As Artificial Intelligence (AI) becomes increasingly integrated into daily life, the need for "AI literacy" in early education has grown. However, teaching abstract concepts like automated reasoning and planning to upper elementary students (ages 8-11) presents significant pedagogical challenges. Traditional instructional methods often fail to engage this age group or prove too technically complex. To address this gap, researchers developed PRIMARYAI, a story-driven, block-based game designed to make AI Planning accessible through a structured "Use-Modify-Create" (UMC) scaffolding framework.

The PRIMARYAI Environment and UMC Framework

PRIMARYAI is a Unity-based game that immerses students in a narrative about saving endangered penguins on a volcanic island. Students control a "RoboPenguin" to monitor wildlife, engaging with AI concepts through a block-based programming metaphor. This approach allows them to focus on the logic of AI representation and reasoning without the frustration of syntax errors.

The instructional design follows the UMC framework to manage cognitive load:

  1. Use: Students run a pre-built AI planning task to observe robot behavior.
  2. Modify: Students tweak parameters or structural blocks to change the robot's destination or goals.
  3. Create: Students are tasked with building a planning task from scratch.
The AI Planning Dashboard in PRIMARYAI, where students define goals and actions for the RoboPenguin.

As shown in the AI Planning Dashboard, students define "Initial States," "Possible Actions," and "Goal States" to guide the agent. This visual representation helps demystify how an AI agent selects a sequence of actions to achieve a specific objective.

Key Insights: Narrative and Cognitive Transitions

The study revealed that the narrative served as a powerful "hook," fostering situated problem-solving. Students identified strongly with the mission, which helped maintain engagement during difficult tasks. One student even noted, "I am the penguin now!"

Hana, the game's guide, provides a "Modify" task to scaffold the transition from observation to creation.

A critical finding was the cognitive load associated with transitions between UMC phases. While students easily handled simple parameter tweaks, moving to structural modifications—such as deleting blocks to change a destination—represented a significant hurdle. Mission 3, which introduced these changes, saw a spike in completion time, marking a clear shift in cognitive demand. Interestingly, students who spent more time exploring and "tinkering" during the initial "Use" phase tended to perform better in these complex later stages, suggesting that early exploration builds a necessary foundation for later problem-solving.

Results and Learning Outcomes

The implementation with 16 fourth-grade students yielded promising results:

  • Significant Learning Gains: Pre- and post-test scores showed a statistically significant improvement in understanding AI planning concepts (p < .01).
  • High Engagement: 93.75% of students successfully completed the initial phases. While the "Create" phase was the most challenging, 60% of those who reached the final missions were successful.
  • Active Collaboration: Qualitative analysis of video data showed high levels of active engagement, with students celebrating mission completions and collaborating with peers to solve logic puzzles.

Limitations and Future Directions

While successful, the study's small sample size in a single rural district limits generalizability. Technical issues, such as networking instability in remote locations, also highlighted the need for robust offline capabilities in educational software. Future iterations of PRIMARYAI aim to include adaptive AI-driven tutoring to provide real-time, tailored support during the challenging "Create" phase and expand the curriculum to cover Machine Learning and Computer Vision.

Summary updated

BibTeX
@inproceedings{gupta2024supporting,
  author = {Anisha Gupta and Seung Y. Lee and Bradford W. Mott and Srijita Chakraburty and Krista D. Glazewski and Anne T. Ottenbreit-Leftwich and J. Adam Scribner and Cindy E. Hmelo-Silver and James C. Lester},
  title = {{Supporting Upper Elementary Students in Learning AI Concepts with Story-Driven Game-Based Learning}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {2024},
  doi = {10.1609/aaai.v38i21.30354},
}
Figure