�鶹ӰԺ

By Joe Arney

Robin Burke is very familiar with the original news recommender system.��

Before he was a respected expert in the recommender systems that are key to digital life—everything from suggesting the next song in a playlist, to suggesting an online dating partner—Burke wore many hats at the community newspaper his father ran in California. Doing odd jobs as a photographer, darkroom technician, reporter and proofreader, Burke got to see how editors would shape each edition and choose which stories got the most play.

“Now, all that context, and that control over priorities, has shifted to these platforms, whether it’s your social media feed or an app like Google News,” said Burke, professor and chair of the Information Science department. “That journalistic voice isn’t there anymore.”��

Burke is part of a team of researchers from a group of top schools, including Minnesota, Northwestern and Clemson, that is seeking to better understand how digital recommender systems are performing the tasks once left to professional editors. The team secured a $2 million grant from the National Science Foundation to build a platform for researchers eager to experiment with the artificial intelligence that powers news recommender systems.��

It could be game-changing technical infrastructure for academic researchers, who are locked out of the proprietary systems built and studied by tech and social media companies.��

“The people who do this kind of research in industry don’t publish very much about it, so we don't know exactly what's going on in terms of how their systems work, or how well they work,” Burke said. “So the people at Google News, for instance, can do these experiments, but people like me can’t.”

Hot off the digital presses

It's an urgent consideration because the way we get our news is changing—a 2022 Pew Research Center survey found one in 10 U.S. adults get their news on TikTok; for American adults under 30, it’s more like one in four.

“We have put all this control over the public square of journalistic discourse into the hands of companies that don’t have any transparency or accountability relative to what they’re doing,” he said. “I think that’s dangerous. And so, it’s important to think about what the alternatives might look like.”

Those alternatives are bigger than just how news is recommended. The business model governing recommendations is optimized to sell ads while keeping users on a platform. As part of his work, Burke hopes researchers can experiment with alternative incentives that reimagine how we engage with technology.��

That ties back to Burke’s main research thread, which concerns bias in recommender systems. His work aims to create “fairness-aware” algorithms that eliminate inequality around, for instance, gender and ethnicity—which is closely related to what he’s building through NSF.

“If a system only shows us the news stories of one group of people, we begin to think that is the whole universe of news we need to pay attention to,” he said.

Some of the early deliverables of the project may include a newsletter—delivered through a major national news outlet—that would deliver recommendations to readers through a daily news update, and eventually a mobile news app that will help researchers understand the effectiveness of recommender systems in this space.��

If those projects are successful, Burke’s team will apply for additional NSF funding to further build out a robust system that will eventually become self-funded through contributions from other researchers.��

“To do this right, you need scientific infrastructure—the same way they build space telescopes and supercolliders,” Burke said. “This grant is about creating a piece of research infrastructure where somebody can come to the project and say, ‘Here’s my experiment, here’s my code, I want to deliver recommendations to a thousand users over six months and see what happens.’”