While AI activities in Hour of Code have grown significantly, researchers emphasize the importance of hands-on learning and ethical discussions to prepare students for the AI-driven future.
Research: What Can Youth Learn About in One Hour? Examining How Hour of Code Activities Address the Five Big Ideas of Artificial Intelligence. Image Credit: Gorodenkoff / Shutterstock
*Important notice: arXiv publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as definitive, used to guide development decisions, or treated as established information in the field of artificial intelligence research.
In an article submitted to the arXiv preprint* server, researchers at the University of Pennsylvania reviewed how code.org's Hour of Code (HoC) activities aligned with the five big ideas of artificial intelligence (AI), focusing on machine learning (ML) and societal impact. Most activities emphasized perception and ML, while topics like representation were underrepresented.
The study highlighted increased attention to critical computing aspects but noted limited hands-on engagement, with many activities relying more on "telling" than "doing." Researchers suggested designing future activities to cover a broader range of topics and incorporating more collaborative, unplugged approaches.
Related Work
Past work highlighted government, universities, and industry efforts to broaden participation in computing, with initiatives like HoC and Massachusetts Institute of Technology (MIT's) Day of AI expanding AI and ML education for K–12 students. HoC, launched in 2013, reached millions globally but previously lacked AI/ML-focused content, which became a central theme in 2023. Studies revealed most HoC activities emphasized conceptual engagement, often with limited opportunities for hands-on creativity or deeper ethical discussions.
Researchers underlined the need for AI education to balance technical skills with ethical considerations, aligning learning with students' interests and real-world applications. The paper also emphasized the importance of addressing the environmental impact of AI, highlighting challenges like energy consumption in training AI models, which are often overlooked.
HoC AI Analysis
The context of this study is HoC, a platform hosting hundreds of activities tailored for various grade levels, ranging from beginner to advanced, and covering topics such as AI, art, and computer science. These activities, contributed by Code.org partners like technology companies and educational organizations, often include teaser videos and instructional guides for classroom use.
New activities are added each year in preparation for Computer Science Education Week. For instance, the repository grew from 348 beginner activities in 2020 to 495 in 2021, catering primarily to middle and high school students.
For data collection, researchers focused on beginner HoC activities related to AI/ML for middle and high school youth. Using the HoC website's filtering features, they identified 542 beginner activities in December 2023, which increased to 557 by April 2024. Activities labeled as AI-related or mentioning AI/ML in their descriptions were included if their hyperlinks worked, resulting in a final dataset of 47 AI-related activities for analysis.
The researchers analyzed the content of these 47 activities using the five big ideas about AI and the five aspects of ML as content categories. They also coded societal impact topics and instructional methods, such as hands-on or explanatory approaches.
Two researchers completed all activities, taking notes on their alignment with the coding framework, and then collaboratively resolved disagreements through iterative discussions involving a fourth researcher. The small dataset and exploratory nature of the study led the authors to prioritize unanimous agreement in their coding over independent coder reliability.
AI Education Insights
The number of AI-related activities in HoC increased significantly from six in 2021 to 47 by April 2024. However, they accounted for only 6.82% of the 557 beginner activities for middle and high school learners. Of these, 28 activities explicitly labeled as AI-related addressed at least one of the five big ideas of AI.
At the same time, ten did not engage with any ideas despite being categorized as AI-related. An additional nine activities, not labeled as AI-related, were identified as involving AI concepts.
Many activities misrepresented AI by focusing on general programming concepts without discussing AI/ML principles, potentially misleading educators and learners. This lack of clear distinctions between AI/ML concepts and general computing risks undermining the quality of AI education.
Among the 37 AI-related activities, perception was the most frequently addressed idea (83.33%), followed by learning (75%), natural interaction (41.67%), societal impact (41.67%), and representation and reasoning (13.89%).
Activities often relied on videos or teacher explanations rather than hands-on engagement, limiting opportunities for learners to design or train AI models themselves. The study noted that hands-on approaches, such as those using tools like Teachable Machines, are effective but remain underutilized.
Only two activities, "Data Literacy in the AI era workshop" and "Build your chatbot in Python," integrated all five big ideas, requiring significant technical expertise and time.
These activities highlighted key aspects such as voice recognition, rule-based systems, and privacy concerns but demanded more resources than the standard one-hour format.
AI's Societal and Environmental Impact
Fifteen activities (41.67%) addressed the societal and environmental impact of AI/ML, focusing on privacy, harmful biases, surveillance, misinformation, ecological impact, and ethics. Eight activities discussed privacy issues, such as data collection in "AVATAR: Big Data & Digital Footprints." Harmful biases were highlighted in "Discover AI in Daily Life" and "Generation AI."
At the same time, surveillance and misinformation were explored in activities like "Face the Future" and "BOLT meets chat generative pre-trained transformer (ChatGPT)." Ethics of self-driving cars were examined through trolley problems in "AI with RVR+: Autonomous Vehicles."
While some activities touched on environmental impacts, they often adopted a technosolutionist perspective, emphasizing AI's potential to solve ecological challenges without addressing the significant energy consumption and carbon footprint of training AI models.
Conclusion
To sum up, this paper analyzed AI/ML activities in HoC for middle and high school students, noting their significant growth from 2021 to 2024.
Most activities focused on perception and learning, with limited emphasis on other AI concepts like representation and reasoning.
Societal impact received increased attention, but environmental challenges and energy consumption were often overlooked.
The researchers emphasized the need for better instructional tools and clearer definitions of AI/ML concepts, as well as more unplugged and collaborative approaches, to make these ideas accessible to novice learners.
Many activities lacked hands-on engagement, highlighting the need for better instructional tools and the broader inclusion of AI/ML ideas in education.
*Important notice: arXiv publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as definitive, used to guide development decisions, or treated as established information in the field of artificial intelligence research.
Journal reference:
- Preliminary scientific report.
Kafai, Y. B., et al. (2024). What Can Youth Learn About in One Hour? Examining How Hour of Code Activities Address the Five Big Ideas of Artificial Intelligence. ArXiv. DOI: 10.48550/arXiv.2412.11911, https://arxiv.org/abs/2412.11911