Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ChatScratch: An AI-Augmented System Toward Autonomous Visual Programming Learning for Children Aged 6-12

Liuqing Chen, Shuhong Xiao, Yunnong Chen, Ruoyu Wu, Yaxuan Song, Lingyun Sun

TL;DR

ChatScratch targets autonomous visual programming learning for children aged 6–12 by addressing artist's block, asset creativity constraints, and coding guidance through an AI-augmented system. It combines an interactive storyboard with visual cues, drawing-based asset creation using Stable Diffusion with ControlNet, and a Scratch-specialized large language model for code assistance, all integrated into two synchronized interfaces. In a within-subject study with 24 children, ChatScratch increased visual richness, asset originality, and code quality, while supporting personally meaningful projects and maintaining learner ownership (retention and expansion of templates). The findings demonstrate that structured planning, creative asset generation, and specialized coding guidance can substantially improve autonomous Scratch learning, suggesting practical pathways for scalable, AI-enabled CT education in home and resource-limited contexts.

Abstract

As Computational Thinking (CT) continues to permeate younger age groups in K-12 education, established CT platforms such as Scratch face challenges in catering to these younger learners, particularly those in the elementary school (ages 6-12). Through formative investigation with Scratch experts, we uncover three key obstacles to children's autonomous Scratch learning: artist's block in project planning, bounded creativity in asset creation, and inadequate coding guidance during implementation. To address these barriers, we introduce ChatScratch, an AI-augmented system to facilitate autonomous programming learning for young children. ChatScratch employs structured interactive storyboards and visual cues to overcome artist's block, integrates digital drawing and advanced image generation technologies to elevate creativity, and leverages Scratch-specialized Large Language Models (LLMs) for professional coding guidance. Our study shows that, compared to Scratch, ChatScratch efficiently fosters autonomous programming learning, and contributes to the creation of high-quality, personally meaningful Scratch projects for children.

ChatScratch: An AI-Augmented System Toward Autonomous Visual Programming Learning for Children Aged 6-12

TL;DR

ChatScratch targets autonomous visual programming learning for children aged 6–12 by addressing artist's block, asset creativity constraints, and coding guidance through an AI-augmented system. It combines an interactive storyboard with visual cues, drawing-based asset creation using Stable Diffusion with ControlNet, and a Scratch-specialized large language model for code assistance, all integrated into two synchronized interfaces. In a within-subject study with 24 children, ChatScratch increased visual richness, asset originality, and code quality, while supporting personally meaningful projects and maintaining learner ownership (retention and expansion of templates). The findings demonstrate that structured planning, creative asset generation, and specialized coding guidance can substantially improve autonomous Scratch learning, suggesting practical pathways for scalable, AI-enabled CT education in home and resource-limited contexts.

Abstract

As Computational Thinking (CT) continues to permeate younger age groups in K-12 education, established CT platforms such as Scratch face challenges in catering to these younger learners, particularly those in the elementary school (ages 6-12). Through formative investigation with Scratch experts, we uncover three key obstacles to children's autonomous Scratch learning: artist's block in project planning, bounded creativity in asset creation, and inadequate coding guidance during implementation. To address these barriers, we introduce ChatScratch, an AI-augmented system to facilitate autonomous programming learning for young children. ChatScratch employs structured interactive storyboards and visual cues to overcome artist's block, integrates digital drawing and advanced image generation technologies to elevate creativity, and leverages Scratch-specialized Large Language Models (LLMs) for professional coding guidance. Our study shows that, compared to Scratch, ChatScratch efficiently fosters autonomous programming learning, and contributes to the creation of high-quality, personally meaningful Scratch projects for children.
Paper Structure (44 sections, 1 equation, 7 figures, 5 tables)

This paper contains 44 sections, 1 equation, 7 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Learning Computational Thinking through Programming
  4. Enhancing Creativity in Scratch Learning Environments
  5. Supporting Children's Personally Meaningful Projects through Child-AI Collaboration
  6. Formative Investigation
  7. Method
  8. Findings and Design Implications
  9. Artist's Block in Project Planning
  10. Bounded Creativity in Assets Creation
  11. Code Assistance as a Scaffold
  12. Support Personally Meaningful Projects
  13. Design Goals
  14. The design of ChatScratch
  15. System Overview
  16. ...and 29 more sections

Figures (7)

  • Figure 1: Overview of ChatScratch: in the project preparation phase, ChatScratch assists children in planning their project with an interactive storyboard (a.1), provides visual cues to overcome their artist's block (a.2), and enhances asset quality with image polish (a.3). During the coding phase, it facilitates coding Q&A via a voice interface (b.1), offers support through voice guide (b.2) and generates a foundational code template (b.3).
  • Figure 2: When using ChatScratch, the child first (a) plans and creates assets with an AI-empowered interactive storyboard (i.e., (a.1) storyline, (a.2) visual cues, (a.3) inspiration button, (a.4) voice button, (a.5) image polish button), then (b) programs with step-by-step code assistant (i.e., (b.1) dialog bubbles button, (b.2) "code help" button, (b.3) "unclear" button, (b.4) "clear" button, (b.5) the generated Scratch blocks).
  • Figure 3: Overview of the assets creation pipeline for ChatScratch. Children express their needs and sketch on the drawing board. Children can then obtain visual cues to detail and enrich their assets. The stable diffusion model and ControlNet were used to iteratively polish their doodles. All polished assets are automatically imported into the ChatScratch programming interface.
  • Figure 4: Overview of the code assisting pipeline for ChatScratch. Children can ask the code assistant which is supported with a Scratch-specialized large language model, to get step-by-step coding tips. In the first step, the assistant generates voice prompts on how to use related Scratch blocks. In the second step, the assistant generates Scratch block templates to provide a tangible starting point. To generate high-quality code templates, we use expert-selected samples and employ prompt engineering to train our large language models.
  • Figure 5: Child participants' behaviors recordings. A boy was interacting with ChatScratch.
  • ...and 2 more figures