The Model Thinker is a great overview of how modeling techniques can be employed to understand complexity, predict and discover unexpected and unintended consequences, as well as explore possible outcomes for potential scenarios. The author Scott E. Page uses his expertise in Social and Political Sciences to present examples in these domains to motivate the models, making it very accessible to most readers. This easy read is well organized with 29 short chapters covering 353 pages of content with 26 pages of notes. In addition, the manuscript is very well researched with more than 380 citations. For those who wish to expand their knowledge on modeling from a variety of perspectives, or for those who are looking for a gentle introduction into how modeling is used to make decisions, this book is perfect. It would be a great book for undergraduates in Mathematics, Computer Science, Statistics, Operations Research, Business, and Political Science to begin to build a repertoire of modeling frameworks and where they may be applied in the real world. This work is extremely accessible to anyone with a modest mathematical background as the most complicated mathematical tools presented only requires one to understand the idea of logarithms.
The book begins with several chapters motivating why one would want to pursue modeling for decision making and why using multiple models is often preferred. Here, Page presents the REDCAPE scheme for uses of models: Reason, Explain, Design, Communicate, Act, Predict, Explore (REDCAPE). These motivations are illustrated using very interesting examples, such as why your friends have more friends on average and why Lehman Brothers was allowed to fail but AIG was not. Then the author argues why using multiple models may be advantageous, citing the Condorcet Jury Theorem, The Diversity Prediction Theorem and the Model Error Decomposition Theorem. These are presented in very simple terms that most people interested in modeling would easily understand. The chapters early on also stress the ‘one-to-many’ approach to modeling, giving examples of how models from one discipline may be applied analogously to another, for example, how ecological models may be applied to the social sciences and how network models can be applied to business as well as politics. This section also covers the challenges of modeling human behavior in which most of the examples are oriented.
The Model Thinker covers modeling paradigms such as linear models, network models, diffusion, local interaction models, Markov models, systems dynamics, threshold and learning models. As well as these modeling paradigms the aspect of working with uncertainty is presented in a simplified manner using the Normal distribution and long tailed distributions. Ideas such as long-run behavior, equilibrium, path dependence as well as many concepts from Game Theory are presented. As there are many topics covered, none of the material presented would be considered in depth. Examples range from using Shapley values to determine the value of individuals in a coalition, to diffusion models for product adoption, to random walks for market efficiency, to Markov models for student engagement and boredom, to using Rugged Landscape models for evolutionary trait systems.
The broad coverage is used well by the author as he stresses the notion of being a ‘many model thinker,’ which advocates for individuals to not use a single model perspective as many different modeling paradigms should be used to understand the problem confronted. This ‘many model thinker’ approach really shines in the last chapter that considers the Opioid epidemic. A Multi-Armed Bandit model illustrates how one can consider the approval of Opioids for use in the general population, a Markov chain model is presented to consider how people transition between the states of No Pain, Opioid User, and Addict, followed by a Systems Dynamics model of paths that lead people to heroin addiction. The three models illustrate how complicated the problem is across many different criteria. This chapter also provides models for economic inequality looking at growth model for wage equilibrium for educated and uneducated labor, preferential attachment models, Markov models, sorting models and several others to illustrate how and why inequality exists from a variety of perspectives.
This book is a terrific choice for anyone wanting to learn more about modeling. While many disciplines use various aspects of modeling, this book is quite enlightening about the breadth of where models can be used across a wide variety of disciplines. Many domain specific researchers may have a grasp of the modeling paradigms commonly used in their context, and this work may allow them to bring other modeling paradigms to bear on their problems. This book is also recommended for those who develop and employ models on a regular basis as the breadth may spur innovation into improving current models.
There are a few limitations to the book that one should be aware of. One is that the focus on breadth comes at the expense of depth. This book may not be appropriate as a sole textbook for a course in modeling as algorithms are explained from an intuitive perspective versus giving specific techniques to model. However, it could accompany a more rigorous modeling textbook to provide a path for students easy access to modeling principles and their applications. Similarly, for researchers looking for a ‘cookbook’ of algorithms, this may not be a sufficient resource. However, the extensive references do provide resources that may contain the algorithms to employ these methods. Another limitation is that the book is written with a focus toward understanding versus rigor. Many modelers may read some of the explanations and wonder if statements could be oversimplified or missing technical rigor. While the work does cover statistical learning methods, one modeling framework that seems to have minimal coverage is deep learning. This may be due to the fact that deep learning is more algorithmic in focus and is inherently not intuitive.
Overall, this book is highly recommended for readers, from those who are simply curious about models and their applications to real world applications, to those who are actively applying modeling in their daily lives. Again the broad coverage of modeling and principles of modeling are sure to give most readers insights into models that may be unfamiliar. The easy-to-read prose makes for an enjoyable reading experience. And the short chapters on engaging topics makes one want to read many chapters in a single sitting.
Virginia Commonwealth University