Assessing regional to global climate change requires plausible scenarios of land use and land cover change (LULCC) at spatiotemporal scales appropriate to coupled land-atmosphere modeling. Land use/land cover, however, is an emergent outcome of complex and multiscale social and biophysical processes. Improving the resolution of LULCC necessitates increasing the resolution of social processes that are simplified or neglected at large scales. At regional scales simulating LULCC requires identifying relevant actors, and characterizing their decision-making processes and their concomitant dependence on biophysical systems and other actors. Agent-based models are a useful tool to represent these dynamics because they provide a flexible framework to represent multi-scale interactions, and heterogeneous information networks and sources. Allowing for variation in decision-making processes and information exchange through social networks creates multiple scenarios of LULCC that can capture our uncertainty in how LULCC may evolve over time. We examine how social learning and network structure affects LULCC in a diverse agricultural region. Use of agents allows us to dynamically downscale global market forcings from models such as the Global Change Assessment Model as a function of local social processes. Here, we present model results for an area in southwest Idaho, USA where seed crop production is of national and global significance, while a rapidly expanding metropolitan area is providing increased pressure on land use. The interaction of social learning with various social network structures highlights the importance of incorporating these dynamics to further our understanding of the heterogeneity and uncertainty associated with LULCC modeling.