독후감) Superforecasting

 

This book written by philp E. Tetlock is about skills to improve forecasting. Language is often emotional and vague. Especially there are several times in this book, he use metaphor to people who are not good at forecasting as chimpanzee. He has strong opinions about being precise. So the text also carries feelings that the author push his obsession rather than delivering useful tools. Moreover, each chapter does not clearly list summary. To explain one small takeaway, the author used many explanations and examples repeatedly. It is a relief that on the last chapter, he made a simple summary.

Still, the author gives us considerable methods to develop ways of forecasting. Some methods are widely used in interviews and tests which I already knew. For example, Break down intractable problems into tractable sub-problems, outside/Inside view, and dragonfly eye view. However, there are also methodology which is new to me. This method is based on news. Set the deadline and percentage at that moment. If you find new evidence that may change your probability, adjust your first picked probability. Else, keep your number. Repeat this until the deadline. It seems easy and simple. Yet, all method listed on the book is about how to increase and decrease probabilistic percentage. How we can deal with the number is what we actually need.

For instance, lets say weather foreaster said tomorrow will be rainy with 60% chance. Should I bring the umbrella? What should I do with the percentage? When I took course called "Offshore structure" in my master, I learned about stochastic wave theory. In short, it is about how to find percentage of dangerous wave height coming to hit the offshore structures. The result is probability. One day I asked the professor that what we need to do after find the percentage. He answered that only experienced enginners can decide what we need to do next. So it was about time accumulated experience.

It is nice to be accurate in many ways. We better keep this mind and practice over and over. Then aftet countless fails, we can hunch what to do next wisely.