Three Starting Rules for Applying AI to Data Visualization

As an advocate, user, and creator of data visualizations, it’s impossible to ignore the tremendous opportunities and challenges that it brings. On one hand, the power of generative AI to create, organize, and contextualize vast amounts of information promises intense growth in computational and creative ability. On the other hand, as with many other technological leaps, it raises serious questions about information security, privacy, intellectual property, and the impact of continued automation on jobs that humans currently do.

This article isn’t meant to litigate those tradeoffs, and doing so would probably be difficult until we know more about the technology. What I will try to do here is to suggest some early insights about where the use cases are–and or not–in the area of data visualization.

When many of us approach AI, we do it from the standpoint of popular culture, in which we can find examples of AI that grows self-aware and destroys humanity in the Terminator series; or of sentient androids that are capable of human thinking like Data in Star Trek: The Next Generation. To continue the Starfleet allusion, perhaps the way to think of generative AI is more like the ship’s computer on the Enterprise, in which an AI can distill a vast amount of information in multiple iterations, and create a clear and actionable response.

So where can generative AI help us as data visualizers?

• In segmenting and contextualizing vast amounts of data. Applied to a large database of quantitative and even qualitative data, generative AI can help us segment data into actionable categories much faster than a human, or a team of humans, could do. It’s hard to place tens of thousands of text responses to a survey, for example, into mutually exclusive, collectively exhaustive categories consistently. But an AI can be applied to do this, either for the full database or a sample. This is especially useful in marketing and retail where understanding customer segments is essential. It’s useful to put the words and phrases that customers use to describe their experience in context.
• As an input into data visualization applications. Applied to a large database of quantitative and even qualitative data, generative AI can help us segment data into actionable categories much faster than a human, or a team of humans, could do. It’s hard to place tens of thousands of text responses to a survey into mutually exclusive, collectively exhaustive categories consistently. But an AI can be applied to do this, either for the full database or a sample. This is especially useful in marketing and retail where understanding customer segments is essential. It’s useful to put the words and phrases that customers use to describe their experience in context.
• As an input into data visualization applications. Generative AI is not yet in a place where it can create data visualizations itself. But the outputs it generates, in the form of matrices or other data, can be inputs into tools like Tableau, R, or others which can create useful visualizations.
• As a tool for creating background or supplementary art as part of data storytelling. Tools like Gemini and Adobe Firefly – which only uses images from its own stock database and the public domain to train its model–are useful for creating illustrative content to incorporate into greater data stories as background or supporting visuals. (The image accompanying this article is by Firefly, via a prompt to generate “a prism taking quantitative data and creating a spectrum of light.”)

In the class I teach on data visualization, I’m starting to hear from my students that their organizations are struggling with how to adopt generative AI as part of their workflow. Rather than jumping into a brand-new arena without a sense of how they will use it, it helps to use these potential use cases to find well-founded places to begin exploring how AI can help humans create more effective, clearer, and insightful data visualizations, with a greater goal of helping organizations understand the people they serve.