Study: No-code AI agent creation boosts computational thinking skills

Researchers from a mixed-methods study examined computational thinking (CT) development among 93 pre-high school students during a five-day AI agent creation workshop using CocoFlow, a no-code platform. The study integrated pre-post assessments, behavioral logs, and interviews to measure CT gains and track learning trajectories. Results showed significant improvements in abstract thinking (d = 0.71) and algorithmic thinking (d = 0.70), with hierarchical regression identifying iterative testing engagement as a key predictor of increased self-efficacy (beta = 0.20, p = 0.05). The most striking finding was an Optimal Development Zone effect: students with moderate initial CT levels demonstrated substantially larger gains than both high-CT and low-CT peers (eta squared = 0.55). Qualitative analysis revealed that moderate-CT students developed adaptive expertise, while high-CT students tended to over-engineer their solutions and low-CT students struggled with task decomposition.

These findings challenge traditional linear assumptions about learning progression and have practical implications for educators and curriculum designers. The study provides strong evidence that no-code AI agent creation platforms like CocoFlow can effectively foster computational thinking in middle school settings. However, the differentiated outcomes by initial CT level suggest that one-size-fits-all approaches may be suboptimal. High-CT students may benefit from more open-ended challenges to prevent over-engineering, while low-CT students need explicit scaffolding for task decomposition. The moderate-CT group's disproportionate gains highlight the potential of AI creation as a sweet spot for developing CT in a typical classroom population. This research offers data-driven guidance for building more effective and equitable CS education programs.