Data Visualization: Understanding the medium

The Why

Every medium has range. A newspaper can run shallow, attention-grabbing stories, or publish reporting that digs into what's going on. A documentary can be edited to push your emotions and sell you a neat conclusion, or made in a way that helps you understand multiple perspectives. Most of what we encounter stays surface-level, but that's not a limitation of the medium itself.

The same is true for data visualization. The same dataset might be shown as a bar chart, a color map, or a visualization that surfaces patterns and prompts deeper questions. The medium allows for a universe of possibilities. How do you find your way through it?

The What

To navigate that space, we need a better understanding of how the medium works. Data visualization operates differently from forms like writing. It relies on perception as much as reasoning.

This means working with two kinds of bias. Visual perception comes with built-in shortcuts: we naturally compare lengths, spot outliers, and group by color before we've consciously thought about it. These perceptual biases can be harnessed to communicate quantities clearly, or they can mislead if used carelessly. Statistical processes introduce a different set of biases through sampling, aggregation, and summarization. These choices shape what the data says before it's ever drawn on screen.

Recognizing these biases gives us a more systematic way to think about visualization. Instead of picking chart types by convention, we can reason about what a given encoding reveals or hides.

The How

The talk looks at visualization and data as two separate components, then brings them together.

On the visualization side, we look at how visual encodings tap into perceptual biases to communicate quantities, and how our tendency to compare can be used as a deliberate lever in shaping what a visualization emphasizes.

On the data side, we look at how statistical summaries shape the story before any visual choice is made, and how understanding that process helps us show data more honestly.

We walk through a few pieces of my work, including A Visual Exploration of Indian Population, to see these ideas in practice.