Impact of Generative AI-assisted programming on the computational thinking of high school students

Editorial reading guide · Independence

How can AI help my teenager learn programming while preserving the ability to reason through errors?

Examines the boundary between scaffolding and cognitive substitution. The AI-assisted class improved in overall computational thinking, particularly algorithmic modeling and abstraction/decomposition, but not pattern recognition/evaluation; interviews also described bypassing thinking. This supports asking what a child can explain and debug independently.

Study context

83 novice Python learners in two intact classes at one western Chinese high school; 42 intervention and 41 comparison students

Age / stage: Tenth grade; average age 15 years

Read with three questions

1. Which computational-thinking dimensions improved, and which did not?
2. When did rapid explanations scaffold understanding and when did students describe cognitive substitution?
3. What would my child need to demonstrate without assistance before treating a programming task as learned?

Keep the limits in view

Students were not individually randomized, and intact classes from one school may differ despite similar baselines. The eight-week intervention lacked detailed interaction logs and long-term transfer tests. Gains within this teaching arrangement do not establish that unrestricted home AI use improves thinking.