People care more about being right than avoiding mistakes, new study finds
Research challenges assumptions behind forecasting models, AI systems and predictive decision-making tools
Institute for Operations Research and the Management Sciences
BALTIMORE, June 23, 2026 – Conventional wisdom says the best predictions are the ones that minimize mistakes.
New research suggests that is not necessarily how people see it.
A study published in the INFORMS journal Management Science found that when people make or evaluate predictions, they care more about the possibility of being exactly right than about reducing the size of potential errors. In many cases, people prefer predictions that increase the chance of a perfect outcome, even if doing so creates a greater risk of larger mistakes.
The findings challenge a fundamental assumption underlying many forecasting systems, predictive analytics tools and artificial intelligence models, which are typically designed to minimize average error rather than maximize the likelihood of being precisely correct.
The study, “Understanding People’s Preferences for Predictions: People Prioritize Being Right Over Minimizing How Wrong They Are in Expectation,” was authored by Berkeley J. Dietvorst of the University of Chicago Booth School of Business.
“For many people, the goal of prediction is to be right, not just to minimize average error,” said Dietvorst. “This is news to those who build models for predictive analysis who often work to minimize gross inaccuracies instead of simply working to make the predictions perfect.”
The research suggests that people evaluate prediction errors in a way that differs sharply from how many experts and forecasting systems approach the problem.
Rather than responding proportionally to every increase in error, people appear to be highly sensitive to the difference between being exactly right and being slightly wrong. Once a prediction is already wrong, additional errors matter less than many forecasting models assume.
Dietvorst analyzed 16 studies involving a variety of prediction tasks and experimental designs. Across the studies, participants consistently demonstrated what researchers describe as “diminishing sensitivity to prediction error.”
In multiple experiments, participants were presented with probability distributions and asked to make predictions. They were significantly more likely to choose outcomes that maximized their chances of being exactly correct rather than outcomes that would reduce expected error overall.
Additional studies examined emotional reactions to prediction errors. The results showed that satisfaction dropped sharply when participants moved from being perfectly correct to slightly wrong. By comparison, larger increases in error produced much smaller declines in satisfaction.
The findings have implications far beyond forecasting.
Many predictive systems, including AI tools, recommendation engines, business forecasting models and decision-support technologies, are designed around statistical measures that prioritize minimizing average error. The research suggests that users may judge those systems using a very different standard.
“The results highlight the importance of understanding human objectives before designing predictive systems,” Dietvorst said. “If models are optimized for goals that differ from what people value, adoption and trust may suffer.”
As organizations increasingly rely on predictive technologies to guide decisions, the study raises an important question: Are today's forecasting systems optimized for what people actually want?
The findings suggest that the answer may depend on whether success is defined by being less wrong—or by being right.
Read the full study here.
About INFORMS and Management Science
INFORMS is the world’s largest association for professionals and students in operations research, AI, analytics, data science and related disciplines, serving as a global authority in advancing cutting-edge practices and fostering an interdisciplinary community of innovation. Management Science, a leading journal published by INFORMS, publishes research on decision sciences, behavioral economics, analytics, and quantitative methods that inform managerial and policy decisions. INFORMS empowers its community to improve organizational performance and drive data-driven decision-making through its journals, conferences and resources. Learn more at www.informs.org or @informs.
###
Contact
Rebecca Seel
INFORMS Public Affairs Specialist
(443) 757-3578
Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.