Research Visualization with Claude: Code, Critique, and Iterate

Research Visualization with Claude: Code, Critique, and Iterate

Abstract: This workshop shows how to use Claude to generate, evaluate, and refine figures faster while sticking to the principles of visualization. The session is organized into three parts. The first part covers what Claude can and cannot do in visualization contexts, including where it saves time and where human judgment remains essential. In the second part, participants prompt Claude to generate and render figures directly as Artifacts with no Python environment needed, then critique the output against core visualization principles such as visual encoding, data-ink ratio, and audience appropriateness. In the third part, we iterate, refining the figure through targeted prompting and showing how the same prompt produces runnable Python code in Plotly Express and Altair for use in research workflows. Throughout the session, participants work with real datasets and leave with a repeatable prompting workflow they can apply to their own research.

Objectives:

Participants will learn to:

• Prompt Claude to generate and render figures directly as Artifacts with no setup required.
• Evaluate Claude-generated charts against core principles of effective visual communication.
• Iterate on chart design through targeted prompting while maintaining visualization best practices.
• Export Claude-generated code in Plotly Express and Altair for use in research notebooks, presentations, and web deployment.

Presenter: Ashwini Naik

Level: Introductory

Duration: 2 hours

Prerequisite: Familiarity with Python is helpful but not required. A UChicago Claude account is required. Participants are expected to bring a laptop for the hands-on portions of the workshop.