Citation for the award recipient: Rogelio Oliva

The System Dynamics Society presents the Jay Wright Forrester Award as often as once annually for the best contribution to the field of system dynamics during the preceding 5 years. The committee considers papers, articles, books, or other written material published in English. The award includes a commemorative plaque and US$5000.

The 29th Jay Wright Forrester Award is presented to Professor Rogelio Oliva for his article: “Structural dominance analysis of large and stochastic models,” published in System Dynamics Review in 2016, volume 32(1), pages 26–51.

“Structure generates behavior” is a key principle of system dynamics. Feedback loops and the stocks within them are the basic building blocks of system structure; the behavior of a system is determined by the dominant feedback loops at a given point in time. Nonlinearities can cause different feedbacks to become dominant at different times, and these endogenous shifts in loop dominance often produce complex, unexpected dynamics. To recommend policies to improve problematic behaviors, we should know their source, so we can be guided to focus on a pertinent subset of the structure for designing policies for change.

However, it is often not easy to identify the dominant loops nor when and why shifts in dominance occur. A simple mathematical solution does not exist for high‐order systems with many nonlinear relationships; model experimentation and sensitivity testing have been the usual approach. But such methods are inevitably incomplete and depend on modeler skill. The chance of erring is quite high, especially for the novice.

Mathematical approaches to gain deeper understanding of model behavior using eigenvalue analysis and related methods have been proposed for use in system dynamics for some time. A brief sample includes the work of Forrester (1983), Kampmann (1996), Gonçalves (2009), Mojtahedzadeh (2011), and Kampmann and Oliva (2006). Notable also are the works in the 2013 virtual special issue of System Dynamics Review, “Methods for Identifying Structural Dominance,” edited by Jim Duggan and Rogelio Oliva. Eigenvalue elasticity analysis (EEA) provides a formal relationship between structure (meaning links, loops, and parameters) and the behavior modes. The approach uses Taylor expansion to linearize the system at every time step. The Jacobian of the linearized system is then used to derive the eigenvalues, which decompose the behavior of the linearized system into its various modes. These elements of EEA then make it possible to understand how each mode relates to the stocks in the model, to understand the importance of linkages between variables, and to show how loop dominance evolves over time.

Yet most of the literature illustrates EEA using small teaching models. Applications to realistic models, including models with stochastic elements, have been limited (see Duggan and Oliva, 2013). Building on the prior work, including Kampmann's (1996) algorithm to identify the shortest independent loop set in a model, Rogelio developed a rigorous approach to define and identify loop dominance. In the 2016 article for which this award is given, he applies these methods to larger, more realistic models including one with stochastic features. The article describes how the approach can work well even under these more challenging conditions, making it more broadly useful.

Rogelio finds solutions to the challenges of applying EEA to large or stochastic models in several innovative and practical ways, including an adroit use of graph theory. As he says of these methods: “They work! They generate results” (p. 47). This emphatic statement is indeed appropriate. The mathematical machinery takes one to the desired destination: a precise understanding of which elements of model structure are generating model behavior, and hence where model elements might be changed to improve behavior. Rogelio's work reaches the goal in a clear, comprehensive, and effective way. It takes us closer to the day when EEA will be robustly incorporated in our modeling software packages, allowing us to better analyze our models, find sensitive relationships and parameters requiring stronger empirical grounding or more detailed modeling, and identify high leverage points for effective policy interventions.

Rogelio is Bob and Kelly Jordan Professor of Business at the Texas A&M University. He also serves as Adjunct Professor at MIT's Zaragoza Logistics Center and as Research Affiliate at the MIT Center for Transportation & Logistics. He developed his technical acumen as a graduate of the Monterrey Institute of Technology in Mexico and received his PhD from MIT. He studied Soft System Methodology with Peter Checkland at Lancaster University in England. Not only is Rogelio an outstanding scholar, but he also finds time to serve the System Dynamics Society and broader scholarly community, including as President of the Society in 2010 and Executive Editor of the System Dynamics Review from 2012 to 2015.

On behalf of the System Dynamics Society, the award committee is delighted to present Rogelio Oliva with the 2019 Jay W. Forrester Award.

中文翻译：

---

第29届Jay Wright Forrester奖

获奖者的引文：Rogelio Oliva

系统动力学协会每年颁发一次Jay Wright Forrester奖，以表彰其在过去5年中对系统动力学领域的最佳贡献。委员会审议以英文发表的论文，文章，书籍或其他书面材料。该奖项包括一块纪念牌匾和5000美元的奖金。

第29届Jay Wright Forrester奖授予Rogelio Oliva教授，其文章为“大型和随机模型的结构优势分析”，发表于2016年《系统动力学评论》，第32（1）卷，第26-51页。

“结构产生行为”是系统动力学的关键原理。反馈回路及其中的库存是系统结构的基本构建模块；系统的行为由给定时间点的主要反馈回路确定。非线性会导致不同的反馈在不同的时间占主导地位，而环路支配性的这些内生变化通常会产生复杂的，意想不到的动态。要建议改善问题行为的政策，我们应该知道其来源，因此可以指导我们专注于设计变更策略的结构的相关子集。

但是，通常很难识别主导回路，也不容易确定何时以及为什么发生主导地位变化。对于具有许多非线性关系的高阶系统，不存在简单的数学解决方案。模型实验和灵敏度测试已成为通常的方法。但是这些方法不可避免地是不完整的，并且取决于建模者的技能。错误的机会非常高，特别是对于新手。

提出了使用特征值分析和相关方法对模型行为加深理解的数学方法，用于系统动力学已有一段时间。一个简短的样本包括Forrester（1983），Kampmann（1996），Gonçalves（2009），Mojtahedzadeh（2011）以及Kampmann和Oliva（2006）的著作。同样值得注意的是《系统动力学评论》 2013年虚拟特刊中的作品Jim Duggan和Rogelio Oliva编辑的《确定结构优势的方法》。特征值弹性分析（EEA）提供了结构（含义链接，循环和参数）与行为模式之间的形式关系。该方法使用泰勒展开在每个时间步长上使系统线性化。线性化系统的雅可比行列式然后用于导出特征值，该特征值将线性化系统的行为分解为各种模式。然后，EEA的这些要素使得有可能了解每种模式与模型中的股票之间的关系，了解变量之间联系的重要性，并显示循环支配地位如何随时间演变。

然而，大多数文献都使用小型教学模型说明了EEA。现实模型（包括具有随机元素的模型）的应用受到限制（请参阅Duggan和Oliva，2013年）。Rogelio在先前的工作（包括Kampmann（1996）的算法）中确定了模型中最短的独立回路集，在此基础上，Rogelio开发了一种严格的方法来定义和识别回路支配性。在获得该奖项的2016年文章中，他将这些方法应用于更大，更现实的模型，包括具有随机特征的模型。本文介绍了即使在这些更具挑战性的条件下，该方法也能很好地工作，使其用途更加广泛。

Rogelio通过多种创新和实用的方法（包括熟练使用图论）找到了将EEA应用于大型或随机模型的挑战的解决方案。正如他对这些方法所说：“它们有效！他们产生结果”（第47页）。这种强调的声明确实是适当的。数学机制将目标带到了期望的目的地：对模型结构的哪些元素正在生成模型行为的精确理解，以及在何处可以更改模型元素以改善行为的精确理解。Rogelio的工作以清晰，全面和有效的方式达到了目标。距离将EEA牢固地集成到我们的建模软件包中的日子越来越近，这使我们能够更好地分析我们的模型，找到需要更强的经验基础或更详细的建模的敏感关系和参数，

Rogelio是德克萨斯A＆M大学的Bob和Kelly Jordan商业教授。他还担任麻省理工学院萨拉戈萨物流中心的兼职教授，并担任麻省理工学院运输与物流中心的研究会员。他从墨西哥蒙特雷技术学院毕业后发展了自己的技术敏锐度，并获得了麻省理工学院的博士学位。他与英国兰开斯特大学的Peter Checkland一起学习了软系统方法论。Rogelio不仅是杰出的学者，而且他还抽出时间为系统动力学学会和更广泛的学术界服务，包括2010年担任该学会主席以及2012年至2015年担任《系统动力学评论》的执行编辑。

奖项委员会代表系统动力学协会高兴地向Rogelio Oliva颁发了2019年Jay W.Forrester奖。