Work

Learner-Customized Algorithm Visualization Using Generative AI

Year: 2026
Type: conference-paper
Abstract: Existing algorithm visualization tools are often restricted to fixed and predefined algorithmic forms, making it difficult to represent new or customized algorithm structures and behaviors as desired by learners and instructors. This study introduces a generative AI–based visualization system that operates independently of specific algorithms, enabling users to explore and learn not only classical algorithms but also newly developed ones. The proposed system integrates a Large Language Model (LLM) with a Python-based execution module. The LLM generates a formal specification of an algorithm's execution process, which is then executed to ensure reproducibility and scalability, while the resulting states are visualized through graphs and tables. The overall architecture follows a three-layer structure-LLM, Execution, and Visualization-linked by a visualization contract that maintains consistency across learning outcomes. By combining generative AI with algorithm visualization, this system establishes a new paradigm for algorithm learning, allowing learners to move beyond passive observation and actively engage in exploring and designing computational processes alongside AI.
Language: English
Open Access status: hybrid
OpenAlex ID: W7151014796
FWCI: 4.651
Cites: 18
Cited by: 1
Related to: 10