Model-Development in Neuroscience: Simplicity and Generalizability in Mechanistic Explanations

神经科学中的解释通常是通过机制模型来传递的。然而，许多大脑机制包含大量不同的组成部分，并跨越几个层次的组织。因此，神经科学家经常面临一个选择问题：模型中应该包括哪些成分和水平？或者简单地说：在一组相互竞争的多层次机制模型中，哪个模型是最好的？为了解决这样的选择问题，神经科学家在其他标准中或多或少地使用明确的简单性和概括性措施。简单性针对这样的问题：为了机械地解释一个特定的行为或认知现象，应该包含多少层次？如何精确地描述这些不同层次的组成部分和相互作用？这种简单性的考虑与经典的曲线拟合程序有很大的不同。可泛化性涉及以下问题：一个特定的模型应该适用于多少现象？它应该泛化多少个体和物种？在开发多层次脑机制模型时，平衡有时相互冲突的简单性和概括性双重标准至关重要。这个项目的总体研究问题是关于规范和实践，通过这些规范和实践，简单性和概括性被应用为神经科学多层次机制模型发展的标准。通过在几个深入的案例研究的基础上确定这些标准是如何以及应该如何应用于当前的神经科学研究，该项目旨在阐明如何开发和选择最佳的多层次机制解释。因此，它不仅有助于改进神经科学中机械解释的哲学概念，而且还为当前的科学研究提供规范性指导。更具体地说，这个项目的目标有四个方面：(1)总的来说，发展一个改进的神经科学多层次机制解释的哲学解释，强调简单性和概括性作为解释充分性的标准。(2)阐明简单性作为评估多层次机制解释的解释充分性的标准。(3)阐明在机械解释实践中对概括性的改进理解。(4)最后，一个科学目标是阐明在未来神经科学研究中使用的模型开发和模型选择的规范指南。

Explanations in neuroscience are often delivered by models of mechanisms. However, many brain mechanisms contain a large number of distinct components and span several levels of organization. As a consequence, neuroscientists often face a choice problem: Which components and levels should be included in a model? Or simply: Which model is the best one among a set of rival multi-level mechanistic models? In order to resolve such choice problems, neuroscientists use more or less explicit simplicity and generalizability measures, among other criteria. Simplicity targets questions such as: How many levels should be included in order to mechanistically explain a particular behavioral or cognitive phenomenon? How precisely should the components and interactions of these various levels be described? Simplicity considerations of this kind are of a very different nature than classical curve-fitting procedures. Generalizability concerns questions such as: To how many phenomena should a particular model apply, and over how many individuals and species should it generalize? Balancing the sometimes conflicting dual criteria of simplicity and generalizability is of paramount importance when developing models of multi-level brain mechanisms.The overall research question of this project concerns the norms and practices by which simplicity and generalizability are applied as criteria for the development of models of multi-level mechanisms in neuroscience. By determining on the basis of several in-depth case studies how these criteria are, and should be, applied in current neuroscientific research, this project aims to clarify how the best multi-level mechanistic explanations are developed and selected. Thus, it is expected to not only contribute to an improved philosophical conception of mechanistic explanation in neuroscience, but to also deliver normative guidelines for current scientific research. More concretely, the aims of this project are fourfold: (1) Generally, to develop an improved philosophical account of multi-level mechanistic explanation in neuroscience, with an emphasis on simplicity and generalizability as criteria of explanatory adequacy. (2) To articulate an account of simplicity as a criterion for assessing the explanatory adequacy of multi-level mechanistic explanations. (3) To articulate an improved understanding of generalizability in mechanistic explanatory practice. (4) Finally, a scientific objective is to articulate normative guidelines for model-development and model-selection to be used in future neuroscientific research.