In this paper we aim to give an analysis and cognitive rationalization of a common practice or strategy of modeling in systems biology known as a middle-out modeling strategy. The strategy in the cases we look at is facilitated through the construction of what can be called mesoscopic models. Many models built in computational systems biology are mesoscopic (midsize) in scale. Such models lack the sufficient fidelity to serve as robust predictors of the behaviors of complex biological systems, one of the signature goals of the field. This puts some pressure on the field to provide reasons for why and how these practices are warranted despite not meeting the stated goals of the field. Using the results of ethnographic study of problem-solving practices in systems biology, we aim to examine the middle-out strategy and mesoscopic modeling in detail and to show that these practices are rational responses to complex problem solving tasks on cognitive grounds in particular. However making this claim requires us to update the standard notion of bounded rationality to take account of how human cognition is coupled to computation in these contexts. Our account fleshes out the idea that has been raised by some philosophers on the "hybrid" nature of computational modeling and simulation. What we call "coupling" both extends modelers' capacities to handle complex systems, but also produces various cognitive and computational constraints which need to be taken into account in any computational problem solving strategy seeking to maintain insight and control over the models produced.

中文翻译：

---

介观建模是处理复杂生物系统的一种认知策略。

在本文中，我们旨在对系统生物学中一种常见的建模实践或策略（即中间建模策略）进行分析和认知合理化。通过构建所谓的介观模型，可以简化我们所研究的案例中的策略。计算系统生物学中建立的许多模型都是介观的（中型）规模。这样的模型缺乏足够的保真度来充当复杂生物系统的行为的稳健预测器，这是该领域的标志性目标之一。尽管未达到该领域既定目标，但是这给该领域带来了一定压力，以说明为什么以及如何应保证这些做法的原因。利用民族志研究系统生物学中解决问题的实践的结果，我们的目的是详细研究中间淘汰策略和介观模型，并证明这些做法是对基于认知基础的复杂问题解决任务的合理响应。然而，提出这一要求要求我们更新有限理性的标准概念，以考虑到人类认知在这些情况下如何与计算耦合。我们的叙述充实了一些哲学家提出的关于计算建模和仿真的“混合”性质的想法。我们所谓的“耦合”不仅扩展了建模人员处理复杂系统的能力，而且还产生了各种认知和计算约束，在寻求保持对所生成模型的洞察力和控制力的任何计算问题解决策略中都需要考虑这些约束。