Scratch for Sports: Athletic Drills as a Platform for Experiencing, Understanding, and Developing AI-Driven Apps

Vishesh Kumar, Marcelo WorsleyAAAI 2023 · DOI

TL;DR

This paper introduces "Scratch for Sports," a curriculum and set of tools that leverage athletic drills to teach AI concepts to young learners. By integrating AI-powered sports apps and Scratch extensions, the project aims to make AI education more engaging, culturally relevant, and accessible, particularly for students who might not traditionally engage with STEM fields.

What Problem Does This Paper Address?

The authors identify a critical need for broader AI education, especially for young people. They highlight the persistent challenges in accessing computing and AI education, particularly for underrepresented groups based on gender, race, and socioeconomic status. Traditional AI education often happens in computing-specific contexts, which can be a barrier. This paper proposes using sports, a domain with high intrinsic motivation for many youth, as a bridge to introduce AI concepts in a more engaging and accessible way, connecting abstract AI principles to tangible, physical activities.

Methods

The core of the paper's methodology lies in creating a curriculum that combines existing AI-powered sports applications with custom-built Scratch extensions.

  • AI-Powered Sports Apps: The project utilizes apps like HomeCourt.AI, which uses pose detection and IMU-based gesture detection to track athletic performance in drills like reaction time and dribbling. These apps provide immediate feedback and gamified experiences.
  • Scratch Extensions: The authors have developed Scratch extensions that expose the underlying AI algorithms (e.g., pose detection, object tracking) from these sports apps. This allows learners to directly interact with and modify these algorithms within the familiar Scratch visual programming environment.
  • Embodied Cognition: The approach emphasizes "body syntonicity," drawing inspiration from Seymour Papert's work, where learners connect computational concepts through their own physical movements. This is exemplified by activities involving smartwatches and Micro:bits for step counting and gesture recognition.
  • Interdisciplinary Curriculum Design: The curriculum is designed to be implemented across various settings, including physical education classes, library programs, and summer camps, catering to a wide range of learners from elementary to high school.

Datasets

Not discussed in the paper. The paper focuses on the application of AI techniques rather than the training of new models on specific datasets. The tools leverage existing AI capabilities within commercial apps and open-source libraries.

Key Insights and Ideas

  • Sports as a Powerful AI Education Gateway: The paper strongly suggests that sports provide a highly motivating and culturally relevant context for introducing complex AI concepts, making them more accessible and engaging for a broader audience.
  • Bridging the Gap Between Physical Activity and Computational Thinking: By linking physical drills with programming, the project fosters a unique form of embodied learning where abstract AI principles become concrete and understandable through movement.
  • Democratizing AI Development: Providing Scratch extensions for AI algorithms empowers young learners to move from being passive users of AI to active creators, enabling them to tinker, experiment, and develop their own AI-driven applications.
  • Ethical Considerations Through Practical Application: The curriculum naturally leads to discussions about the ethical implications of AI, such as privacy concerns with data tracking in sports, by grounding these abstract issues in real-world applications that students are actively using.
  • The Value of "Body Syntonicity" in AI Education: Connecting AI concepts to personal body movement, as Papert advocated for with Logo, proves effective in making AI more intuitive and relatable for learners.

Key Results

The paper presents preliminary findings from pilot implementations, highlighting:

  • High Engagement: Learners showed significant excitement and engagement with the gamified drills and the ability to program their own AI applications.
  • Conceptual Understanding: The activities fostered discussions about AI concepts like pose detection, gesture recognition, and data analysis, as well as the limitations and ethical considerations of these technologies.
  • Development of AI Literacy: Students were able to grasp the underlying AI principles and apply them to create their own simple AI-driven games and applications using Scratch.
  • Qualitative Feedback: Teachers reported positive experiences, noting the effectiveness of sports as a context for learning about AI and its societal impact.

Limitations and Future Work

  • Technical Reliability: The authors note lower reliability with Scratch programming on iPads, often opting for Chromebooks or laptops with webcams for programming activities.
  • Algorithm Transparency: While they infer the AI methods used in apps like HomeCourt, the proprietary nature of these algorithms means learners cannot directly inspect or modify them within the original apps.
  • Future Work:
    • Developing more robust Scratch blocks for differentiating people, tracking balls, and scoring objects to enable more complex match recording and shot tracking functionalities.
  • Formulating reliable fabrication techniques for embedding programmable sensors into sports balls that retain integrity and can be charged.
  • Developing more open-source and easily programmable ball-based AI learning activities.
  • Formalizing curricular tools and materials for broader dissemination, ensuring they are sensitive to learners' cultural backgrounds and critical discourses around technology.
  • Centering discussions on the equitable creation, usage, and impacts of AI technologies.

Summary updated

BibTeX
@inproceedings{kumar2023scratch,
  author = {Vishesh Kumar and Marcelo Worsley},
  title = {{Scratch for Sports: Athletic Drills as a Platform for Experiencing, Understanding, and Developing AI-Driven Apps}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {2023},
  doi = {10.1609/aaai.v37i13.26901},
}
Figure