As data scientists, clarity in our messaging and communication is essential. Typically, we achieve this by creating engaging visualizations paired with well-crafted explanations. Visuals must be straightforward for audiences who may not be data-oriented, and the accompanying text should emphasize the key messages.
While various tools can assist in this process, LLMs (Large Language Models) are emerging as an invaluable option. Having an AI assistant that can analyze a chart and distill its main insights makes it easy to incorporate these findings into presentations or documents. Many AI platforms—such as ChatGPT, Claude, and Gemini—offer this feature.
Our approach uses programmatic APIs, but anyone can utilize this technology simply by taking a photo of a chart with their phone or uploading an image through an AI app or web interface.
Here is an example with chatgpt’s online interface (other interfaces work similarly):
- Copy the chart you want explained to your clipboard or save it as an image file
- Open a new chat session or the one you’re using
- Paste the chart directly or use the clip icon to upload the image file
- Add your instructions, such as: “Explain the attached chart” or “Summarize three key insights from the chart.”
In this example, I generated a simple scatter plot to show the relationship between penguin bill length and bill depth across three different penguin species.
Creating the chart requires some expertise (although LLMs can also assist with this step). Once the chart is ready, uploading it with a prompt like “Explain this chart” in ChatGPT yields the following output:
This scatter plot visualizes the relationship between two physical measurements of penguins: bill length (x-axis) and bill depth (y-axis). The points represent individual penguins, categorized by species (Adelie, Chinstrap, and Gentoo) and are color-coded accordingly:
- Red: Adelie penguins
- Green: Chinstrap penguins
- Blue: Gentoo penguins
From the plot:
- Species Clustering: Each species forms a distinct cluster.
- Adelie penguins have smaller bill lengths (around 35–45 mm) and generally deeper bills (17.5–21 mm).
- Chinstrap penguins have bill lengths ranging from 40–50 mm and slightly shallower bills, generally between 17.5 and 20 mm.
- Gentoo penguins have the longest bills, between 45–60 mm, but shallower bill depths, often below 18 mm.
- Species Differentiation: This chart suggests that bill length and depth can be useful for distinguishing between these penguin species, as each occupies a different region on the plot. Gentoo penguins, for example, have significantly longer bills but shallower depths compared to the other species.
This plot effectively highlights interspecies differences based on these two measurements, which could assist in species identification or studies on penguin morphology.
The explanation is solid, though if I were explaining the chart myself, I might add a few notes on variance and emphasize key differences in the slope between length and depth, which visibly varies by species.
As always, take the output with a grain of salt. Review everything carefully and make adjustments as needed—LLMs aren’t immune to mistakes.
Conclusions
Large Language Models (LLMs) are becoming essential tools in the data scientist’s arsenal, particularly for making data visualizations more accessible and impactful. Platforms like ChatGPT, Claude, and Gemini enable these AI assistants to bridge the gap between complex data visuals and clear, narrative explanations that resonate with non-technical audiences.
Using LLMs is straightforward—simply upload your visualization to your chosen interface and prompt for analysis. However, it’s important to view LLMs as complementary tools, not replacements for human expertise. While they excel at creating clear, structured explanations, they may overlook subtle patterns that an experienced data scientist would recognize. Success lies in combining the AI’s explanatory power with human insight, producing richer, more accurate data narratives and leveraging the efficiency that LLMs provide.
Want to learn more? Reach out for a consultation with one of our experts to discuss implementing AI in your organization or explore our data science solutions.