In a country that’s increasingly polarized, where partisan politics shape perception, the question of how to create a compelling argument — rooted in facts — that can change minds isn’t simply academic. It’s a question that’s chipping away at the foundations of American democracy.

The cruel joke at the center of this is the role that new technologies have played in creating the current political environment and the fracturing of political thought.

Now a new company that’s emerging from the latest batch of the Y Combinator accelerator is hoping to write a better punchline — using technology to solve the problems that technology has created.

Swayable was founded by three former physicists to help craft political messages that actually inform and persuade rather than simply incite and propagandize.

Formed initially out of frustration with the country’s response to climate change, Swayable’s three founders — James Slezak, Valerie Coffman and Lyel Resner — hope to bring the experimental rigor of scientific testing to the efficacy of political messaging.

BIRMINGHAM, AL – DECEMBER 12: Voters wait in line to cast their ballot at a polling station set up in the St. Thomas Episcopal Church on December 12, 2017 in Birmingham, Alabama. Alabama voters are casting their ballot for either Republican Roy Moore or his Democratic challenger Doug Jones in a special election to decide who will replace Attorney General Jeff Sessions in the U.S. Senate. (Photo by Joe Raedle/Getty Images)

For Slezak, the company’s chief executive, the importance of crafting a persuasive political message was driven home during his tenure as the vice president and chief of operations at NYT Global.

“What really stood out for me at the Times was that they do a lot of critical reporting on big issues, but it just wasn’t getting through to people on the other side of the filter bubble,” Slezak says. “It was a scary time coming out of the election of 2016 and a lot of people — myself included — wanted to contribute to get the world back on track and the political system more functioning. We’re trapped in these increasingly divided filter bubbles but there was just no obvious way to solve that.”

So Slezak tapped his network and enlisted some friends from his days as a research physicist to come up with a solution to the messaging problem.

And it is a problem. Last year a study by Stanford professor David Brockman and UC Berkeley political scientist Joshua Kalla revealed that all of the work political campaigns put into advertising, canvassing, direct mailing and phone calls simply doesn’t work.

“Campaigns probably need to get more creative and think more outside the box,” Brockman told Vox in an interview. “Whatever box they are working within now doesn’t usually produce results.”

Slezak and his team took Brockman’s advice — and his work — to heart.

“The way to have an important impact on a situation where the messages were getting off track was to use science,” Slezak says.

Working with Coffman, the former chief science officer of 3D printing and computer aided design company Xometry, and Resner, who was serving as a fellow at the Civic Data Lab at the Massachusetts Institute of Technology, Slezak came up with an approach that most social and data scientists know all too well.

What we’ve done is productize randomized control trials,” says Slezak. “It’s a set of experiment designs that have been used by social scientists for a long time… It’s also exactly what we would have done in a physics experiment.”

Marshaling resources on sites like Fiverr and Amazon Mechanical Turk, Swayable basically A/B tests political messaging to see which ones resonate and are most likely to persuade an audience.

Users send in at least four different versions of the message (typically a video) that they want to test. They’re logged into the Swayable platform and from there, the service goes to work. Swayable recruits its control and sample groups from sites like Facebook, Fiverr and Amazon Mechanical Turk and shows them the video.

To ensure the integrity of the results, Swayable tests to determine that its study groups are actually from the demographics that customers want to target. At the end of the survey — within 12 to 24 hours — Swayable comes back to a customer with the results of the study.

Using statistics, data and information targeting, the company can determine which messages are most likely to change someone’s mind, Slezak says.