Sotheara Veng

Generative artificial intelligence (GenAI) is entering children’s lives, often before they learn how to use it. Computational thinking (CT) may offer a foundation for this learning, yet research integrating CT and GenAI typically involves older learners. This study examined the feasibility of integrated CT and GenAI lessons with 22 second graders at a Title I school in the Mid-Atlantic United States. The study included ten lessons on CT, GenAI, or their combination. CT scores on a validated paper-and-pencil measure increased statistically significantly from pre- to post-test with an effect size larger than similar CT interventions; observations showed students applied CT concepts when writing prompts and identifying errors generated by GenAI. In a GenAI prompting post-test asking students to recreate an image, students described image content in detail but rarely used directive prompt language. Still, ChatGPT largely produced images matching the target image and students’ intent, possibly creating an illusion of competence. Self-efficacy and intrinsic value were rated high across lessons, with some differences reflecting strengths of lessons in supporting student interests and participation. These findings offer evidence that CT can serve as a framework for GenAI literacy in early elementary grades and that the combined approach can produce gains in each. Lessons learned from implementation present opportunities for improvement to the studied program and others.

Computational thinking (CT) skills are essential for navigating daily life, but how best to teach these skills is still being explored. Educational programming games can provide engaging environments to develop CT, yet uncovering student struggles during gameplay remains challenging. This study analyzed 15 undergraduates with varying programming experience playing the CT game, Fox and Field, utilizing gameplay videos, heart rate data, and surveys. Results revealed common struggle points in understanding the character's perspective, degree angles, and debugging, with notable differences between novice and experienced players. Findings suggest support targeting abstraction, pattern recognition, and debugging skills to improve future game design and scaffolding.