Amy J. Ko

Stefania Druga, Amy J. Ko

Creative coding platforms like Scratch have democratized programming for children, yet translating imaginative ideas into functional code remains a significant hurdle for many young learners. While AI copilots assist adult programmers, few tools target children in block-based environments. Building on prior research \cite{druga_how_2021,druga2023ai, druga2023scratch}, we present Cognimates Scratch Copilot: an AI-powered assistant integrated into a Scratch-like environment, providing real-time support for ideation, code generation, debugging, and asset creation. This paper details the system architecture and findings from an exploratory qualitative evaluation with 18 international children (ages 7--12). Our analysis reveals how the AI Copilot supported key creative coding processes, particularly aiding ideation and debugging. Crucially, it also highlights how children actively negotiated the use of AI, demonstrating strong agency by adapting or rejecting suggestions to maintain creative control. Interactions surfaced design tensions between providing helpful scaffolding and fostering independent problem-solving, as well as learning opportunities arising from navigating AI limitations and errors. Findings indicate Cognimates Scratch Copilot's potential to enhance creative self-efficacy and engagement. Based on these insights, we propose initial design guidelines for AI coding assistants that prioritize youth agency and critical interaction alongside supportive scaffolding.

Stefania Druga, Amy J. Ko

What role can AI play in supporting and constraining creative coding by families? To investigate these questions, we built a Wizard of Oz platform to help families engage in creative coding in partnership with a researcher-operated AI Friend. We designed a 3 week series of programming activities with ten children, 7 to 12 years old, and nine parents. Using a creative self efficacy lens, we observe that families found it easier to generate game ideas when prompted with questions by AI Friend; parents played a unique role in guiding children in more complex programming tasks when the AI Friend failed to help, and children were more encouraged to write code for novel ideas using the AI friend help. These findings suggest that AI supported platforms should highlight unique family AI interactions focused on children's agency and creative self-efficacy.

Maryam Arab, Thomas D LaToza, Amy J Ko

Knowledge sharing plays a crucial role throughout all software application development activities. When programmers learn and share through media like Stack overflow, GitHub, Meetups, videos, discussion forums, wikis, and blogs, every developer benefits. However, there is one kind of knowledge that developers share far less often: strategic knowledge for how to approach programming problems (e.g., how to debug server-side Python errors, how to resolve a merge conflict, how to evaluate the stability of an API one is considering for adoption). In this paper, we investigate the feasibility of developers articulating and sharing their strategic knowledge, and the use of these strategies to support other developers in their problem-solving. We specifically investigate challenges that developers face in articulating strategies in a form in which other developers can use to increase their productivity. To observe this, we simulated a knowledge-sharing platform, asking experts to articulate one of their own strategies and then asked the second set of developers to try to use the strategies and provide feedback on the strategies to authors. During the study, we asked both strategy authors and users to reflect on the challenges they faced. In analyzing the strategies authors created, the use of the strategies, the feedback that users provided to authors, and the difficulties that authors faced addressing this feedback, we found that developers can share strategic knowledge, but authoring strategies require substantial feedback from diverse audiences to be helpful to programmers with varying prior knowledge. Our results also raise challenging questions about how future work should support searching and browsing for strategies that support varying prior knowledge.