Mechanistic Models and the Explanatory Limits of Machine Learning (original) (raw)

II-Mechanistic Explanation: Its Scope and Limits

Aristotelian Society Supplementary Volume, 2013

This paper explores the question of whether all or most explanations in biology are, or ideally should be, 'mechanistic'. I begin by providing an account of mechanistic explanation, making use of the interventionist ideas about causation I have developed elsewhere. This account emphasizes the way in which mechanistic explanations, at least in the biological sciences, integrate difference-making and spatio-temporal information, and exhibit what I call fine-tunedness of organization. I also emphasize the role played by modularity conditions in mechanistic explanation. I will then argue, in agreement with John Dupré, that, given this account, it is plausible that many biological systems require explanations that are relatively nonmechanical or depart from expectations one associates with the behaviour of machines.

Systems Biology and Mechanistic Explanation

The Routledge Handbook of Mechanisms and Mechanical Philosophy, 2018

We address the question of whether and to what extent explanatory and modelling strategies in systems biology are mechanistic. After showing how dynamic mathematical models are actually required for mechanistic explanations of complex systems, we caution readers against expecting all systems biology to be about mechanistic explanations. Instead, the aim may be to generate topological explanations that are not standardly mechanistic, or to arrive at design principles that explain system organization and behaviour in general, but not specific mechanisms. These abstraction strategies serve various aims, including prediction and control, that are central to understanding the epistemic diversity of systems biology.

Systems biology and the integration of mechanistic explanation and mathematical explanation

The paper discusses how systems biology is working toward complex accounts that integrate explanation in terms of mechanisms and explanation by mathematical models—which some philosophers have viewed as rival models of explanation. Systems biology is an integrative approach, and it strongly relies on mathematical modeling. Philosophical accounts of mechanisms capture integrative in the sense of multilevel and multifield explanations, yet accounts of mechanistic explanation (as the analysis of a whole in terms of its structural parts and their qualitative interactions) have failed to address how a mathematical model could contribute to such explanations. I discuss how mathematical equations can be explanatorily relevant. Several cases from systems biology are discussed to illustrate the interplay between mechanistic research and mathematical modeling, and I point to questions about qualitative phenomena (rather than the explanation of quantitative details), where quantitative models are still indispensable to the explanation. Systems biology shows that a broader philosophical conception of mechanisms is needed, which takes into account functional-dynamical aspects, interaction in complex networks with feedback loops, system-wide functional properties such as distributed functionality and robustness, and a mechanism’s ability to respond to perturbations (beyond its actual operation). I offer general conclusions for philosophical accounts of explanation.

Explanation in Biology: An Enquiry into the Diversity of Explanatory Patterns in the Life Sciences

2015

Despite the philosophical clash between deductive-nomological and 6 mechanistic accounts of explanation, in scientific practice, both approaches are 7 required in order to achieve more complete explanations and guide the discovery 8 process. I defend this thesis by discussing the case of mathematical models in 9 systems biology. Not only such models complement the mechanistic explanations 10 of molecular biology by accounting for poorly understood aspects of biologi11 cal phenomena, they can also reveal unsuspected ‘black boxes’ in mechanistic 12 explanations, thus prompting their revision while providing new insights about the 13 causal-mechanistic structure of the world. 14

Explanatory models in neuroscience: Part 1 – taking mechanistic abstraction seriously

2021

Despite the recent success of neural network models in mimicking animal performance on visual perceptual tasks, critics worry that these models fail to illuminate brain function. We take it that a central approach to explanation in systems neuroscience is that of mechanistic modeling, where understanding the system is taken to require fleshing out the parts, organization, and activities of a system, and how those give rise to behaviors of interest. However, it remains somewhat controversial what it means for a model to describe a mechanism, and whether neural network models qualify as explanatory. We argue that certain kinds of neural network models are actually good examples of mechanistic models, when the right notion of mechanistic mapping is deployed. Building on existing work on model-to-mechanism mapping (3M), we describe criteria delineating such a notion, which we call 3M++. These criteria require us, first, to identify a level of description that is both abstract but detail...

(2015) Explanation in Biology: An Introduction

2015

Explanation in biology has long been characterized as being different from explanation in other scientific disciplines, in particular from explanation in physics. One of the reasons was the existence in biology of explanation types that were unheard of in the physical sciences: teleological and functional explanations, historical and evolutionary explanations. More recently, owing in part to the rise of molecular biology, biological explanations have been depicted as mechanisms. This profusion of explanatory patterns is typical of biology. The aim of the present volume Explanation in Biology. An Enquiry into the Diversity of Explanatory Patterns in the Life Sciences is to shed some new light on the diversity of explanation models in biology. In this introductory chapter, we recall the general philosophical context of scientific explanation as it has unfolded in the past seven decades, and highlight the specific issues that models of explanation have faced in biology. We then show how the different essays gathered in this collective volume tackle aspects of this important debate.

New Mechanistic Explanation and the Need for Explanatory Constraints

Scientific Composition and Metaphysical Ground, 2016

This paper critiques the new mechanistic explanatory program on grounds that, even when applied to the kinds of examples that it was originally designed to treat, it does not distinguish correct explanations from those that blunder. First, I offer a systematization of the explanatory account, one according to which explanations are mechanistic models that satisfy three desiderata: they must 1) represent causal relations, 2) describe the proper parts, and 3) depict the system at the right ‘level.’ Second, I argue that even the most developed attempts to fulfill these desiderata fall short by failing to appropriately constrain explanatorily apt mechanistic models.

The Completeness of Mechanistic Explanations

The purpose of the paper is to provide methodological guidelines for evaluating mechanistic explanations meant to complement previously elaborated interventionist norms. According to current accounts, a satisfactory mechanistic explanation should include all of the relevant features of the mechanism, its component entities and activities, their properties and their organization, as well as exhibit productive continuity. It is not specified, however, how this kind of mechanistic completeness can be demonstrated. I argue that parameter sufficiency inferences based on mathematical model simulations of known mechanisms is used to determine whether a mechanism capable of producing the phenomenon of interest can be constructed from mechanistic components organized, acting, and having the properties described in the mechanistic explanation.