AI-Infused Collaborative Inquiry in Upper Elementary School: A Game-Based Learning Approach

Seung Y. Lee, Bradford W. Mott, Anne T. Ottenbreit-Leftwich, J. Adam Scribner, Sandra Taylor, Kyungjin Park, Jonathan P. Rowe, Krista D. Glazewski, Cindy E. Hmelo-Silver, James C. LesterAAAI 2021 · DOI

Introduction

Artificial Intelligence (AI) is rapidly transforming the global landscape, yet K-12 education—particularly at the elementary level—often lags behind. Most existing AI curricula target high school students who already possess foundational computer science knowledge. This paper introduces PRIMARYAI, a game-based learning (GBL) environment designed to bridge this gap for upper elementary students (ages 8-11). By integrating AI concepts into a compelling ecology narrative, the researchers aim to lower barriers for both young learners and their teachers.

The PRIMARYAI Environment

Developed using the Unity engine, PRIMARYAI immerses students in a "save the penguins" narrative. Students take on the role of ecologists on a New Zealand island investigating a decline in yellow-eyed penguins. To gather data, they deploy semi-autonomous robots disguised as baby penguins. This narrative framing transforms abstract AI concepts into concrete tools for scientific discovery.

Methods and Pedagogical Design

The system employs a combination of game-based learning and problem-based learning (PBL). The design process involved significant teacher co-design, ensuring the tool aligns with classroom constraints and the Next Generation Science Standards (NGSS).

The PRIMARYAI collaborative inquiry cycle within an elementary classroom.

As shown in Figure 1, the pedagogical loop moves from problem scenarios to hypothesis generation, followed by data collection and modeling within the game environment. This cycle concludes with small-group negotiation and the sharing of evidence-based solutions, mimicking real-world scientific inquiry.

AI-Infused Block-Based Programming

A core feature of PRIMARYAI is its AI-infused block-based programming interface. Using a Blockly-inspired system, students program their robots to perform complex tasks like image recognition and autonomous navigation.

Block-based logic for image classification and data recording.

Instead of focusing on syntax, students engage with the underlying logic of machine learning. For example, as seen in Figure 4, students can set an "image Label" by classifying a photo using a pre-trained or student-trained "animal classifier." They then use conditional logic (e.g., "If image Label = weasel") to trigger specific actions like recording a location. This approach scaffolds the "black box" of AI, making it accessible to 4th and 5th graders.

Key Insights and Results

The research highlights several critical findings:

  • AI as a Science Tool: Framing AI as a tool for ecological inquiry makes the subject relevant to students who might not be traditionally interested in coding.
  • Teacher Efficacy: Interviews with veteran STEM teachers revealed a significant "content knowledge gap." Teachers expressed a need for immersive professional development and "scripted but open-ended" curricula that provide security while allowing for student inquiry.
  • Lowering the Floor: Block-based programming acts as a cognitive scaffold, allowing students to focus on high-level logic (like training a classifier) without the burden of complex syntax.

Conclusion and Future Work

While initial teacher feedback is positive, future work will focus on quantitative pilot studies to measure student learning gains. The researchers also plan to explore more complex AI topics, such as perception and planning, and address critical issues like AI ethics and implicit bias within the immersive game world.

Summary updated

BibTeX
@inproceedings{lee2021ai,
  author = {Seung Y. Lee and Bradford W. Mott and Anne T. Ottenbreit-Leftwich and J. Adam Scribner and Sandra Taylor and Kyungjin Park and Jonathan P. Rowe and Krista D. Glazewski and Cindy E. Hmelo-Silver and James C. Lester},
  title = {{AI-Infused Collaborative Inquiry in Upper Elementary School: A Game-Based Learning Approach}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {2021},
  doi = {10.1609/aaai.v35i17.17836},
}
Figure