Training AI to detect disinformation

The best spreaders of disinformation don't just lie. They weave their information into convincing stories.

Mark Finlayson is training artificial intelligence to understand how those stories work, giving analysts powerful new tools to identify coordinated disinformation campaigns before they take hold.

Finlayson, who received the U.S. Presidential Early Career Award for Scientists and Engineers in 2025, trains AI systems to detect the narrative patterns that make disinformation campaigns effective. His approach goes beyond simple fact-checking. Instead, he focuses on how adversaries structure their messaging with storytelling techniques that make false information compelling.

With funding from the Department of Defense and Department of Homeland Security, Finlayson has developed tools that help intelligence analysts identify coordinated influence operations. His systems analyze how usernames hint at credibility; how cultural symbols carry different meanings across audiences; and how events get arranged within narratives to maximize persuasive impact.

One of his breakthroughs addresses a fundamental challenge in using AI to analyze stories: narratives often don’t present events in chronological order. To solve this problem, Finlayson created an algorithm that correctly extracts timelines from complex narratives — a capability that helps analysts track how disinformation evolves and spreads.

Rather than replacing human judgment, these AI tools find patterns and connections that would be nearly impossible to detect manually across thousands of social media accounts. The work brings computational precision to a critical national security challenge: protecting democratic discourse from sophisticated foreign manipulation.

“A good story can change someone’s mind more easily than facts alone. That’s precisely what makes narrative such a powerful tool for those spreading disinformation.”