Dynamic mechanistic explanation: computational modeling of circadian rhythms as an exemplar for cognitive science

Abstract

We consider computational modeling in two fields: chronobiology and cognitive science. In circadian rhythm models, variables generally correspond to properties of parts and operations of the responsible mechanism. A computational model of this complex mechanism is grounded in empirical discoveries and contributes a more refined understanding of the dynamics of its behavior. In cognitive science, on the other hand, computational modelers typically advance de novo proposals for mechanisms to account for behavior. They offer indirect evidence that a proposed mechanism is adequate to produce particular behavioral data, but typically there is no direct empirical evidence for the hypothesized parts and operations. Models in these two fields differ in the extent of their empirical grounding, but they share the goal of achieving dynamic mechanistic explanation. That is, they augment a proposed mechanistic explanation with a computational model that enables exploration of the mechanism’s dynamics. Using exemplars from circadian rhythm research, we extract six specific contributions provided by computational models. We then examine cognitive science models to determine how well they make the same types of contributions. We suggest that the modeling approach used in circadian research may prove useful in cognitive science as researchers develop procedures for experimentally decomposing cognitive mechanisms into parts and operations and begin to understand their nonlinear interactions.

Introduction

Two widely accepted assumptions within cognitive science are that (1) the goal is to understand the mechanisms responsible for cognitive performances and (2) computational modeling is a major tool for understanding these mechanisms. The particular approaches to computational modeling adopted in cognitive science, moreover, have significantly affected the way in which cognitive mechanisms are understood. Unable to employ some of the more common methods for conducting research on mechanisms, cognitive scientists’ guiding ideas about mechanism have developed in conjunction with their styles of modeling. In particular, mental operations often are conceptualized as comparable to the processes employed in classical symbolic AI or neural network models. These models, in turn, have been interpreted by some as themselves intelligent systems since they employ the same type of operations as does the mind. For this paper, what is significant about these approaches to modeling is that they are constructed specifically to account for behavior and are evaluated by how well they do so—not by independent evidence that they describe actual operations in mental mechanisms.

Cognitive modeling has both been fruitful and subject to certain limitations. A good way of exploring this is to contrast it with a different approach, one that involves more direct investigation into the internal parts and operations of the mechanism responsible for a phenomenon and tailors modeling to this mechanism. To do this we will focus on the phenomenon of circadian rhythms in animals: the ability of the nervous system to regulate activities, including human cognitive activities, on an approximately twenty-four hour cycle. Circadian effects on cognition generally have been ignored in cognitive science, but whether or not that is a desirable state of affairs is not relevant here. Rather, our goal is to use the increasingly prominent role of computational modeling in circadian rhythm research as a different type of exemplar against which to view cognitive modeling. In circadian research, the models are not proposals regarding the basic architecture of circadian mechanisms; rather, they are used to better understand the functioning of a mechanism whose parts, operations, and organization already have been independently determined. In particular, circadian modelers probe how the mechanism’s organized parts and operations are orchestrated in real time to produce dynamic phenomena—what we have called dynamic mechanistic explanation (Bechtel & Abrahamsen, in press).

We begin with an overview of mechanistic explanation in general. We then develop the case of circadian rhythm research, where the architecture has been highly constrained by empirical inquiry into the physical mechanism and modeling is directed to understand the mechanism’s dynamics. We do this by examining in turn six different exemplars from the research literature on computational modeling of circadian rhythms. In all of these cases computational modeling was needed to understand the behavior of a complex mechanism involving nonlinearly interacting components. In examining their particulars, though, we draw out six more specific contributions of computational modeling. We then go through these six contributions again, this time presenting for each a cognitive model and querying to what extent it might make the same kind of contribution. This review of models also brings to light certain differences between cognitive scientists and circadian modelers in how they approach computational modeling.