Using ChatGPT-4 to Teach the Design of Data Visualizations

Abstract

Modern scientific communication revolves around data visualizations. While chemistry curricula include basic instruction on constructing data visualizations, the design of these visualizations is rarely taught. There are two main reasons for this lack of instruction: (i) most instructors were never taught design themselves and thus struggle to articulate design concepts, and (ii) refining data visualizations has traditionally required detailed knowledge of specialized software, which required excessive time to teach within existing curricula. This article reports my recent experience using the AI tool ChatGPT-4 to teach the design of data visualizations. ChatGPT-4 can process data, create, and display data visualizations─eliminating the need to teach specialized software. The article makes the case that this tool can help overcome the above barriers, by leveraging everyday language as the interface for creating data visualizations. The use of everyday language means that one need not possess a specialized design lexicon, but only understand the basics of design─which can be surprisingly easy to learn and teach. This article introduces basic graphic design principles, demonstrates using them with ChatGPT-4 to produce data visualizations, discusses the relative strengths and weaknesses of this approach, reports student perceptions of their experience with the tool, and touches on outcomes from teaching the design of data visualizations using this tool. The overall conclusion is that ChatGPT-4 offers an opportunity to provide meaningful instruction that moves beyond the mechanics of constructing data visualizations, focusing instead on designing effective visualizations. This approach helps prepare students to communicate their science more effectively.