Using mental models to compare and integrate experts' understanding of sustainable agriculture

IQSS Data Science Services presents a talk by Michael Levy, PhD Candidate, Department of Environmental Science and Policy, University of California, Davis. This is a special Data Science Services event which marks the beginning of the new Data Science Talks Series. Michael will speak from 4pm to 5pm; a Q&A session and an open disussion will follow. Abstract: Mental models are directed graphs in which nodes and edges respectively represent perceived components of a system and the causal relationships between them. Here, I describe the use of mental models to understand variance in experts' understanding of a complex social-ecological system, identify opportunities for intervention, and test the consequences of environmental and policy scenarios. I elicit mental models of sustainable agriculture from 148 experts in a variety of professions and regions in California. I find that education and experience don't predict mental model size or density, but they are associated with more-integrated and less-hierarchical models. To examine the content of the models, I develop a semantic analysis tool that leverages internet search results to quantify the similarity of concepts identified by different experts and put all the models in the same terms. Perceived goals for sustainability are widely shared while strategies are more idiosyncratic. Some presumably important connections receive little attention from the experts. Finally, I estimate a consensus model that reflects the aggregate understanding of all the experts, which can be used to identify leverage points in the system and explore implications of policy interventions. I conclude by discussing opportunities and challenges in extracting mental models directly from textual sources, which will allow the integration of large amounts of disparate information in a highly tractable form.