In Pursuit of Democracy

In pursuit of democracy turns Congressional Record mentions from 1880 onward into an animated historical argument. Scroll for the broad arc, watch raw counts give way to percentages, and open individual dots for selected speeches—while keeping the project's OCR and model-assisted classification limits in view.

What is In Pursuit of Democracy?

The essay begins with accumulation: each dot stands for five speeches or written remarks containing the word democracy. As the timeline advances, the same word changes jobs—first often naming the Democratic Party, later carrying arguments about voting rights, war, civil rights, corruption, foreign interference, and government power. The project then shifts from sheer volume to the share of congressional speech using the term, before letting visitors inspect selected remarks year by year. Its methods section is unusually candid. Historical text comes from Stanford's parsed Congressional Record, newer material comes through an open congressional parser, and a named Gemini model classifies passages around the first matching word. The author also says a random sample was checked and warns that the categories remain subjective.

What you can do there

• Scroll through a timeline of Congressional Record references
• Compare raw counts with yearly percentages
• Select dots to read curated speeches

Why we picked it

This is data storytelling with an honest sense of scale: the crowd of dots makes repetition tangible, while the selected speeches keep the chart from turning history into anonymous confetti. Just as important, the published prompt and limitations give readers enough machinery to question the result instead of treating a bright dot as ground truth.

How to get the most from it

Read the opening legend before scrolling so the dot brightness and five-speeches-per-dot rule stay clear. Pause when the display switches from totals to percentages; those views answer different questions. At the end, open several dots from different decades, then read the methods before drawing conclusions from the threat categories.

Good to know

The source material includes OCR text, and the threat labels come from a model-assisted classification that the author calls subjective despite a random manual check. Historical claims and quotations deserve their linked context. The piece also covers racism, incarceration, political violence, and recent partisan conflict; accessibility, motion controls, and mobile behavior were not verified.

Who made it, and when?

Alvin Chang is the credited creator or organization.