Project Awarded: $10,800
The human mind does not simply register every event as it happens; instead, it spontaneously summarizes experience by computing summary statistics or patterns. The choice to do so, and under what circumstances, is a design feature of our cognitive system. Here we aim to investigate spontaneous pattern computation and the intrinsic pressures that increase the likelihood of computing them. We take a novel approach by considering pattern computation as a general phenomenon, and leveraging insights from philosophical treatments (Dennett 1987; 1991). Disparate literatures have investigated two forms of patterns: 1) the ability to represent the mean of a set, such as the average size of a set of dots (Alvarez, 2011; Whitney & Yamanashi Leib, 2018) and 2) the ability to detect predictive relations among pairs or triplets of stimuli in space or time (Fiser & Aslin, 2001; Orbán, Fiser, Aslin, & Lengyel, 2008). Aim 1 is to evaluate whether there are common or distinct principles when computing means and contingencies in order to bridge these two literatures. Specifically, a key principle true of means is that they are compressions: the pattern is computed despite losing information about individual items (Alvarez, 2011; Whitney & Yamanashi Leib, 2018). We test whether this principle applies equally to contingency statistics. Aims 2 & 3 are to investigate what kinds of pressures increase our tendency to compute patterns, with a view to understanding the cognitive utility or motivation for pattern computation. We consider two possible (non-exclusive) motivations for computing patterns (Dennett, 1987): one is that we are hungry for patterns: we want to know about any patterns that exist. Another is that we are need-based compressors: we use patterns as an efficient way to deal with limits in our cognitive capacity. Even though compression does take place, a further prediction from the latter view has not been tested: that pattern computation is more likely to happen as it becomes harder to encode individual items. We probe two ways of making item encoding harder: 1) by making sensory input overwhelming by making it very fast; 2) by making items less predictable. We ask whether one or both factors increase pattern computation. If they do, it supports the notion that need-based compression is a motivator of computing patterns. On the other hand, the hungry-for-patterns view predicts that even when patterns are irrelevant, they should be automatically computed. Overall, our goal is to increase our understanding of the relative contribution of different pressures to compute patterns, with the broader aim of revealing core properties of a fundamental cognitive process in which we spontaneously engage every day.
Anna Leshinskaya, PhD. Postdoctoral researcher, Department of Psychology, University of Pennsylvania
Enoch Lambert, PhD. Postdoctoral researcher, Center for Cognitive Studies, Tufts University