The methodological role of mechanistic-computational models in cognitive science

Jens Harbecke

doi:10.1007/s11229-020-02568-5

Abstract

This paper discusses the relevance of models for cognitive science that integrate mechanistic and computational aspects. Its main hypothesis is that a model of a cognitive system is satisfactory and explanatory to the extent that it bridges phenomena at multiple mechanistic levels, such that at least several of these mechanistic levels are shown to implement computational processes. The relevant parts of the computation must be mapped onto distinguishable entities and activities of the mechanism. The ideal is contrasted with two other accounts of modeling in cognitive science. The first has been presented by David Marr in combination with a distinction of “levels of computation”. The second builds on a hierarchy of “mechanistic levels” in the sense of Carl Craver. It is argued that neither of the two accounts secures satisfactory explanations of cognitive systems. The mechanistic-computational ideal can be thought of as resulting from a fusion of Marr’s and Craver’s ideals. It is defended as adequate and plausible in light of scientific practice, and certain metaphysical background assumptions are discussed.

Highlights

Researchers working in cognitive neuroscience have sometimes expressed different views on which research strategies optimize explanation, prediction, and understanding given limited research resources
The co-director of the Human Brain Project (HBP), Henry Markram, has expressed hopes “(...) to learn a great deal about brain function and dysfunction from accurate models of the brain. (...) There is no fundamental obstacle to modeling the brain and it is likely that we will have detailed models of mammalian brains, including that of man, in the near future.” (2006, p. 158) The view expressed here is that modeling brain structures allows for the simulation of cognitive functions, which in turn provides explanations of actual cognitive functions implemented by the human brain
In his book on vision, neuroscientist David Marr develops a normative ideal of explanation that is based on the contention that there are “[t]hree different levels at which an information processing device must be understood before it is understood completely.“ (1982, p. 24) In his view, the required levels of understanding are the “computational level”, the “representational and algorithmic level”, and the “hardware and implementation level”

Summary

Introduction

Researchers working in cognitive neuroscience have sometimes expressed different views on which research strategies optimize explanation, prediction, and understanding given limited research resources. To make the idea of the abstract plane representing a mechanistic hierarchy with several levels and a spatio-temporal dimension concrete, consider the recent neuroscientific research by Wallis (2012), Payzan-LeNestour et al (2013), Polanía et al (2014), Gluth et al (2017) and others on the mechanisms underlying value-based decision making in humans including boundary phenomena such as the attraction effect As these researchers have shown, the computational process of value-based decision making as applied by the human brain in large set of different environments is physically realized in several specialized brain regions including the orbitofrontal cortex, the nucleus accumbens and the amygdala (“level n”). The following sections develop in detail the adequacy of the MC-account of explanation

Levels of computation

Levels of mechanism

Redundancy

Conclusion