Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Evolving the Computational Notebook: A Two-Dimensional Canvas for Enhanced Human-AI Interaction

Konstantin Grotov, Dmitry Botov

TL;DR

The paper tackles the rigidity of one-dimensional notebooks in supporting non-linear workflows and human-AI collaboration. It proposes Computational Canvas, a two-dimensional workspace with freely arrangeable executable cells, separate environments, and enhanced output management, implemented as a VS Code plugin. Key contributions include the forking-based separate environments for safe experimentation, detachable outputs for organized presentation, and an API-enabled interface for AI agents. The work demonstrates improved non-linear exploration and collaborative potential within IDEs, with practical implications for data science and AI-assisted development.

Abstract

Computational notebooks, while essential for data science, are limited by their one-dimensional interface, which poorly aligns with non-linear developer workflows and complicates collaboration and human-AI interaction. In this work, we focus on features of Computational Canvas, a novel two-dimensional interface that evolves notebooks to enhance data analysis and AI-assisted development within integrated development environments (IDEs). We present vital features, including freely arrangeable code cells, separate environments, and improved output management. These features are designed to facilitate intuitive organization, visual exploration, and natural collaboration with other users and AI agents. We also show the implementation of Computational Canvas with designed features as a Visual Studio Code plugin. By shifting from linear to two-dimensional spatial interfaces, we aim to significantly boost developers' productivity in data exploration, experimentation, and AI-assisted development, addressing the current limitations of traditional notebooks and fostering more flexible, collaborative data science workflows.

Evolving the Computational Notebook: A Two-Dimensional Canvas for Enhanced Human-AI Interaction

TL;DR

The paper tackles the rigidity of one-dimensional notebooks in supporting non-linear workflows and human-AI collaboration. It proposes Computational Canvas, a two-dimensional workspace with freely arrangeable executable cells, separate environments, and enhanced output management, implemented as a VS Code plugin. Key contributions include the forking-based separate environments for safe experimentation, detachable outputs for organized presentation, and an API-enabled interface for AI agents. The work demonstrates improved non-linear exploration and collaborative potential within IDEs, with practical implications for data science and AI-assisted development.

Abstract

Computational notebooks, while essential for data science, are limited by their one-dimensional interface, which poorly aligns with non-linear developer workflows and complicates collaboration and human-AI interaction. In this work, we focus on features of Computational Canvas, a novel two-dimensional interface that evolves notebooks to enhance data analysis and AI-assisted development within integrated development environments (IDEs). We present vital features, including freely arrangeable code cells, separate environments, and improved output management. These features are designed to facilitate intuitive organization, visual exploration, and natural collaboration with other users and AI agents. We also show the implementation of Computational Canvas with designed features as a Visual Studio Code plugin. By shifting from linear to two-dimensional spatial interfaces, we aim to significantly boost developers' productivity in data exploration, experimentation, and AI-assisted development, addressing the current limitations of traditional notebooks and fostering more flexible, collaborative data science workflows.

Paper Structure

This paper contains 9 sections, 2 figures.

Table of Contents

  1. Introduction
  2. Computational Canvas
  3. Cells Arrangement
  4. Rich Outputs Manipulation
  5. Separate Environments
  6. Collaborative Exploration
  7. AI assistance in Computational Canvas
  8. Canvas Integration in IDE
  9. Conclusion

Figures (2)

  • Figure 1: (a) Prototype interface of the computational canvas. Orange cells in the canvas are the executable cells, the grey cells are the outputs, and the colorful areas are the separate environments with independent runtime states. (b) Schematic representation of the forking mechanism while creating a new environment. (c) A schematic representation of a chat between a user and an AI assistant is in Computational Canvas. The agent creates the dedicated environment that executes the user request.
  • Figure 2: example of detached outputs in the computational canvas opened in Visual Studio Code. The left side showed the code, which was executed with different parameters. The right side showed the outputs, which were detached and grouped together.