Comparative Analysis of Inference Under Uncertainty

When people make decisions under uncertainty or with incomplete information, people may rely on mental “fill‑ins” that are part of everyday intelligence but can sometimes lead to errors, like jumping to the wrong conclusion. This project asks why those “quick but wrong” choices happen and whether they arise from the same basic rules of intelligent behavior in humans and comparative animal models. Examining these processes in species that lack language is important for being able to understand the origins of reasoning and the principles that are independent of language abilities. Investigating how different species fill in missing information has potential for determining core principles of learning and reasoning under conditions of uncertainty. This understanding is important for fundamental brain science and relevant for artificial-intelligence (AI) engineers in building tools that guide smarter, safer decisions. The project also provides hands‑on research and data‑science training for high‑school, undergraduate, and graduate students, helping to prepare the next generation of scientists and AI‑literate STEM professionals. The research focuses on a phenomenon called the positive contingency bias, which is a kind of error made when making probabilistic inferences. The project focuses on a tendency to assume that a hidden part of a familiar scene is still present, even when that guess can be wrong. The team plans to develop matched computer‑based and conditioning tasks for humans and comparative animal models to test how different species infer the state of hidden cues as part of their natural intelligence. Planned experiments examine when the positive contingency bias grows stronger or weaker, whether it can cause learning about cues that are never directly seen (only imagined), and whether separate imagined features can be combined into new mental images that guide decision-making. The data analysis plan involves mixed‑effects and Bayesian approaches to compare learning and decision patterns across species, providing tests of associative‑learning and cognitive theories of reasoning under uncertainty. Finding the same bias in non‑linguistic animals would show that some reasoning “fallacies” and shortcuts are built‑in features of biological intelligence, which is relevant for basic brain science and improving the design of next‑generation AI and augmented‑intelligence systems that must operate safely and adaptively in uncertain real‑world environments. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria. NSF Award ID: 2523638 | Program: 01002627DB NSF RESEARCH & RELATED ACTIVIT | Principal Investigator: Aaron Blaisdell | Institution: University of California-Los Angeles, LOS ANGELES, CA | Award Amount: $594,592 View on NSF Award Search: https://www.nsf.gov/awardsearch/show-award/?AWD_ID=2523638 View on Research.gov: https://www.research.gov/awardapi-service/v1/awards/2523638.html