Conor McGowan

In 2025, 62 American black bear (Ursus americanus) experts from Canada, the United States, and Mexico participated in a rigorous elicitation process designed to capture their collective knowledge of the species. We compared their responses against empirical data sourced from 54,000 black bear detections, compiled from datasets donated by researchers across the species' range.
Expert-elicited data is increasingly used to fill gaps in empirical knowledge for wildlife research and conservation decision-making, but its reliability remains poorly understood. Using the black bear as an ideal study species, we evaluated the agreement between "best-case" expert-elicited data and "best-available" empirical data.
Our results show that expert elicitation can identify broad or qualitative trends—such as relative species densities across space—but remains unreliable as a standalone source of quantitative data. Experts were more accurate on well-studied aspects of bear ecology (e.g., the effects of protected areas, particularly at local spatial scales) than on under-studied aspects (e.g., the effects of fire, particularly at range-wide scales). This distinction is critical because expert elicitation is typically employed precisely when empirical knowledge is weak: for under-studied species, broad spatial extents, and unknown ecological relationships.
Our conclusions held even after attempts to control for bias, vary expert group size and composition, or adjust our data aggregation and analysis approaches. However, we demonstrate that expert knowledge retains value when integrated with sparse empirical data; for example, as informative Bayesian priors.
With the support of the bear research community, we reveal the limited utility of expert elicitation as a standalone tool for wildlife research and conservation decision-making, and provide recommendations for when and how to use it to augment, rather than replace, empirical data.