Stony Brook Researchers Develop Tool to Uncover How People Form—and Misform—Causal Beliefs

Does drinking coffee really boost your productivity? Does a cold winter disprove climate change? Humans are natural pattern-seekers, constantly drawing connections between cause and effect. But sometimes we connect dots that aren't actually connected-a phenomenon researchers call "causal illusions." Now, a team at Stony Brook University has developed a method to study these illusions at scale and understand where public beliefs diverge from scientific reality.

Klaus Mueller, a professor in the Department of Computer Science, and his PhD student Shahreen Salim Aunti have created Belief Miner, a crowdsourcing system that reveals what large groups of people believe about cause-and-effect relationships-and where those beliefs may be misleading. Their work, published at the ACM Conference on Computer-Supported Cooperative Work and Social Computing (CSCW 2024), demonstrates how researchers can efficiently map collective understanding of complex issues like climate change, healthcare decisions, and public policy.

"Understanding what large groups of people believe about cause and effect is really hard to do at scale," Salim Aunti explains. "Traditional approaches, like expert interviews or fixed surveys, either take a lot of time or flatten out the nuance in how people actually reason. We wanted a way to collect rich, structured causal knowledge from many people and compare those crowd-sourced beliefs directly to what domain experts know."

The system works by having participants interact with a web-based interface to express their causal beliefs-essentially mapping which factors they think cause which outcomes. In their climate change study, 101 participants from Amazon Mechanical Turk and Prolific built their own causal networks explaining what drives global warming. The researchers then compared these crowd-generated models against a validated expert reference network based on established climate science.

One finding stood out: a notable portion of participants confidently believed solar activity is a primary driver of current climate change-a persistent misconception that didn't align with expert consensus. "What made it interesting wasn't just that the misconception existed, but that participants often held it with high confidence," Salim Aunti notes. "That combination-confident and wrong-is exactly the kind of pattern Belief Miner is designed to surface."

The implications extend far beyond climate communication. In healthcare, the method could reveal why patients resist treatments by surfacing the causal beliefs driving those decisions. In misinformation research, it could pinpoint which false causal links are most widely held. In education, it could help science communicators design explanations that target exactly where public understanding breaks down.

The research team is now investigating whether large language models could make such surveys more scalable and nuanced. In a world where misinformation spreads rapidly and policy debates often hinge on public understanding of cause and effect, Belief Miner offers a systematic approach to revealing the invisible architecture of how we think-and where our thinking goes astray.

-By Yuganshu Jain