How we map the space of possibilities: flexibility and limits in how minds represent options

Description
Jonathan Phillips is an Associate Professor of Cognitive Science at Dartmouth College. His research falls in the intersection of psychology, philosophy, and linguistics, and fits most easily within the broader umbrella of cognitive science. Much of the work he has done relates in one way or another to the psychological representation of modality – the way our minds represent possibilities. He has also conducted research on theory of mind, causal reasoning, moral judgements, formal semantics, and happiness. He earned his Ph.D. in Psychology and Philosophy from Yale in 2015, and before joining the Cognitive Science Program at Dartmouth in 2019, he worked as a postdoctoral fellow in Harvard’s Moral Psychology Research Lab. He is currently the Principal Investigator for the PhilLab at Dartmouth College, which explores how people think about possibilities themselves, and how that influences the way they think more generally, from the language they use, to the causal and moral judgements they make. He was named a Rising Star by the Association for Psychological Science in 2021, among many other honors including the Stanton Award from the Society for Philosophy and Psychology, and Fellowship at “Human Abilities,” the Centre for Advanced Studies in the Humanities at Humboldt Universität.
Abstract
Much of high-level cognition depends on representations of what is possible. To hold someone responsible, we consider what else they could have done; to identify a cause, we consider what else might have happened; to make a decision, we consider what options are available to us. Yet we know surprisingly little about what these representations look like, or whether each kind of judgment builds its own representation of the relevant possibilities or whether they all draw on a common one. I’ll present a research program that makes empirical and theoretical progress on these questions. I’ll introduce a method for measuring the set of possibilities people consider in a specific situation, and show that the process of generating them can resemble foraging in a structured space. I’ll also show that a single estimate of the represented space of possibilities can be used to predict decision making, force judgments, causal reasoning, and blame, suggesting that they all recruit a common map of the space of possibilities. In addition, I’ll show that these representations are both highly flexible (shifting with small changes in context) and seem to be represented abstractly rather than estimated through simulation. New work shows how LLMs provide a useful comparison for what is distinctive about human thought about possibilities.
