Documentation Matters: Human-Centered AI System to Assist Data Science Code Documentation in Computational Notebooks
Project Overview
Generative AI is increasingly transforming education by streamlining processes and enhancing learning experiences, particularly through applications like the Themisto system, which automates documentation in data science. This system utilizes a combination of deep learning, query-based, and prompt-based approaches to improve the quality of computational narratives in Jupyter Notebooks, significantly reducing the time and effort required for documentation while boosting user satisfaction. The document emphasizes the critical role of human-AI collaboration, illustrating how AI can augment human capabilities rather than replace them. Furthermore, it highlights various applications of generative AI that facilitate collaboration and improve educational outcomes. However, it also acknowledges the challenges educators face in integrating these technologies into their teaching practices. Overall, the findings underscore the potential of generative AI to enhance educational processes and outcomes while advocating for strategies to overcome implementation barriers.
Key Applications
AI-Assisted Documentation and Reading Support
Context: Educational settings including data science education, collaborative projects among data science students and professionals, and children's reading programs where students interact with AI-driven conversational agents for reading support.
Implementation: AI tools assist users in generating documentation for code and processes within computational notebooks and provide interactive reading sessions with conversational agents. These tools utilize deep learning, query-based methods, and prompt-based approaches to enhance user experience and documentation quality.
Outcomes: Improved documentation quality, reduced time for documentation tasks, enhanced collaboration among data science teams, and effective communication patterns in children's reading compared to traditional human interaction.
Challenges: Limitations in generating context-sensitive documentation, reliance on user input for certain documentation types, varying levels of user comfort with AI assistance, and potential resistance to adopting new tools.
Implementation Barriers
Technical/Technological Barrier
Challenges in generating accurate context-sensitive documentation due to the complex nature of data science workflows, and complications in integrating AI technology into existing educational frameworks due to varying levels of technological infrastructure.
Proposed Solutions: Development of more sophisticated AI models that can better understand the context of code in data science applications. Additionally, invest in technological upgrades and provide training for educators to facilitate smoother integration of AI tools.
User Acceptance Barrier
Users may resist using AI-generated documentation if they prefer to write documentation themselves or find AI suggestions inadequate.
Proposed Solutions: Provide a hybrid approach that allows users to modify AI-generated suggestions to better fit their needs.
Cultural Barrier
Resistance from educators and students towards adopting AI tools due to fear of job displacement or skepticism about AI's effectiveness.
Proposed Solutions: Implement change management strategies, including workshops and demonstrations to showcase the benefits of AI.
Project Team
April Yi Wang
Researcher
Dakuo Wang
Researcher
Jaimie Drozdal
Researcher
Michael Muller
Researcher
Soya Park
Researcher
Justin D. Weisz
Researcher
Xuye Liu
Researcher
Lingfei Wu
Researcher
Casey Dugan
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: April Yi Wang, Dakuo Wang, Jaimie Drozdal, Michael Muller, Soya Park, Justin D. Weisz, Xuye Liu, Lingfei Wu, Casey Dugan
Source Publication: View Original PaperLink opens in a new window
Project Contact: Dr. Jianhua Yang
LLM Model Version: gpt-4o-mini-2024-07-18
Analysis Provider: Openai