The model thinker — book of Math or book of Popular Science?
Not only an event is hardly predicted, but also it is hardly explained. It involves multiple complex factors so it’s so difficult to find causes and pattern. Thus we have to use some tools to analyse an event/phenomenon. Models are the tools to do that. A model which is a theory that greatly simplified the reality and is precisely described. It is frequently described by graphics and mathematical formula. For instance, F=ma describe the relationship between force, mass and acceleration. Just one model does not allow you to explain all phenomena in this world. Sometimes, we have to use multiple theories to do cross-validation or give us new perspectives. “Multi-model” thinking refers to using multiple models to analyse a phenomenon/event. This book tells us the advantages of “multi-model thinking” as well as its limitations and introduces different model with daily-life examples.
The first 4 chapters introduce theoretical foundations of “multi-model thinking”. It includes “Condorcet Jury Theorem” and “Diversity prediction theorem” which justifies the way of multi-model thinking. The former tells us “if the chance of a people makes correct decision independently is greater than a half, then more people involves in the vote, the more correct collective decision will be”. The latter one means “on average, a multiple model way gives more accurate predictions than the single one”. Chapter 5 to 7 introduces common statistical tools including Normal distribution and Linear Regression. The book begins to introduce more model in chapter 8.
Can a company still run without you (or someone you hate)? Why it is possible that a second majority in parliament have less bargaining power than the minority when forming a cabinet? Chapter 9 introduces Last of the Bus (LOTB)value and Shapley value to explain the contribution and power of each member. Using “Six Degrees of Separation” introduced in chapter 10, you may well find that you need only few steps to connect Jensen Huang (the president of NVIDIA). “The gym room I usually go is so full. How to deal with it?” Don’t worry. Chapter 17 tells us you will eventually find the best time slot. However, there are some conditions we have to meet: if “the total number of users is invariant”, if “you get complete information”, if “you can make infinite decisions” and if “the cost of all time slots are the same”…
The books give many remarkable and easy-to-understand instances for each model. Moreover, the author is very responsible because he tell us the assumptions and limitations of each model. This makes the book better than the others. However, there are still room for improvement: first, it uses very bad notations. Sometimes it is very difficult to understand the meaning of the theory after looking at the equation. For instance, the equation of “model error decomposition theorem” is entirely non-comprehensible [1]. Second, the book involves many math (to many of us it is painful), How much can readers digest? Will these math contents discourage them? (Especially, some notation used very badly). As a book of popular science, it might not suitable for ordinary readers who hate math.
[1]: I am not sure it is the problem of the author or the problem of the publisher of translated version.
