New Free Tool Reveals 17 Years of How We Talk About Each Other

A research team at the University of Illinois has released a giant free tool that turns ordinary Reddit conversations into a searchable map of American attitudes. The collection, called ISAAC, holds more than 527 million English-language Reddit posts written between 2007 and 2023. Instead of grabbing everything, the team ran a human-checked filtering process to keep only posts that touch on six social categories: race, sexuality, age, disability, body weight and skin tone. They say less than 10 percent of what remains is junk or off-topic.

Why should you care? Because this is essentially a 17-year diary of how everyday people actually talk about each other — not what they tell pollsters. Each post was tagged by software for tone, emotion, moral outrage and more, and the team checked whether these tags line up with real-world events like protests and news spikes. That makes it useful for journalists explaining a cultural moment, companies trying to spot brand risk, or anyone curious whether online hostility is genuinely rising. And you don't need to be a programmer: ISAAC works through a point-and-click website, though it's also available for coders.

The most interesting — and uncomfortable — part is location. Software estimates each poster's home region, letting researchers connect online talk to local laws and public opinion. That's powerful for studying how a state's policies shape what people say. But it's also the catch: even though Reddit posts are already public, grouping them by guessed location and personal traits edges toward profiling. The estimates can be wrong, the labels are automated, and Reddit users skew young, male and English-speaking — so this is not a perfect mirror of America, just a very large window into it.