Learning from models about possibilities
Abstrat: Scientists often use models to make claims about possibilities – something “can evolve”, “might be sustained” or “possibly emerges”, to quote just some examples. But possibilities are of different kinds: epistemic possibilities concern what is possible given knowledge of the actual world, while alethic possibilities concern what is objectively possible – what might be the case but perhaps isn’t. What kind of possibility a modeler aims to learn, I argue, affects how the model is constructed and how it can be assessed. First, it affects the way in which model assumptions can be realistic: in epistemically possible modeling, models have actual targets, but in alethically possible modeling, models have merely possible, non-actual targets. This, secondly, affects how such models can be evidentially supported from observations about their targets. Thirdly, this distinction affects how such models feature in how-possibly explanations: some are just potential explanations in search of evidence, but others are fundamentally different from how-actual explanations. Finally, this distinction also affects how model ensembles should be interpreted – as collections of possible models of the same target, or as collections of models of possible targets. Thus, models are used to learn about different kinds of possibilities, and understanding these differences has relevant implications for modeling methodology.
Those unable to attend can stream the lecture from our YouTube page.