Seminar title: Games and modeling for human-environment problem solving | Abstract: When we look at the results of a migration modeling exercise, we expect to see clean, curved arrows showing predicted mobility from place to place. We expect this because this is also likely what our data on migration look like – cleanly delineated flows of people from one place to another over a particular time interval. However, it is challenging to reach this clean set of predictions. We demonstrate the MIDAS (Migration, Intensification, and Diversification as Adaptive Strategies) agent-based modeling framework, and outline the layers of assumptions a modeler must make in order to simulate mobility. We show how MIDAS operationalizes current theory in mobility (on pushes, pulls, and moorings; on capabilities and aspirations; and on place attachment), and the role of data beyond mobility (such as labor participation and wages) in narrowing the problem of equifinality (many different models generating similar predictions). Lastly, we discuss the limits to our capacity to validate model assumptions in the domains of climate and adaptation, where many things simply haven’t happened yet. These constraints reshape the importance of model users as participants in the model building process, and shift modeling goals from ‘try to predict the future’ to ‘try to understand the consequences of our assumptions.’