External Reading

Philip Tetlock’s bestseller is the definitive book on Superforecasting, offering a fascinating insight into the method’s used by superforecasters to outperform even professional intelligence analysts. With practical advice for business, international affairs or everyday life, Tetlock’s book will change how you evaluate uncertainty.

Philip Tetlock’s first book is filled with evidence about how badly most forecasting and judgement practice and performance is. His landmark work lays out how human experts are often no better at making predictions that the layperson and the lack of accountability that often comes with that.

Shortlisted for the 2024 Royal Society Science Book Prize, Everything is Predictable is a popular science book about the history and the art of good forecasting, telling a series of stories about the ways humans have found to do better forecasting. Accessible, well written and an educational, Tom Chiver’s book is the perfect introduction for novices.

Written by artificial intelligence research Elizer Yudkowsky, this is not a book to make you feel warm and fuzzy. It focuses on why intelligent people and capable institutions become trapped in systemic mediocrity, or worse. Yudkowsky argues that even when everyone is competent and well-intentioned, large systems, from governments to financial markets to scientific fields, often settle into stable but suboptimal equilibria.  Sophisticated organisations routinely fall into predictable traps of bad inference, bad incentives, and bad coordination.

Cassi can help with these problems. The book brutally and in a clear eyed way shows the world as it is; Cassi builds tools for how decision-making should be to break those traps.

Statistician David Spiegelhalter’s guide to understanding and managing probability in everyday life argues that probability does not exist. It is not a property of the world but an expression of our relationship with the world. He offers the best description of subjective probabilities and Brier scoring in publication. It is rigorous and accessible, but also offers the underlying statistics in sufficient detail to go deeper for readers who want to.

This is a great place to go deeper than Tetlock’s Superforecasting, and come back to it with more appreciation for the field Tetlock has done so much to advance, and Cassi is now taking to its logical conclusion.

In War & Chance, J. A. Friedman applies probabilistic thinking to the defence and security spheres. Drawing references from historical and experimental evidence, War & Chance provides new foundations for assessing uncertainty in foreign policy.

Filled with practical real world examples, this book highlights how we can classify data and objects. From voting to HIV screening, the authors layout out a persuasive argument in how far you can get with relatively sparse and/or naïve classifiers.

This work by Kahneman, Sibony and Sunstein is vital for really exploring the differences between biases and error, and why collective intelligence approaches can counteract these issues. They argue that noise in human judgement – the variations in people’s judgements of the same problem – is not only inefficient for businesses but that it can lead to gross injustice.

Brian Christian & Tom Griffiths’ book is a great, simple guide to the different elements in forecasting and computation of forecasting. By applying insights from computer algorithms to everyday lives, Christian & Griffiths seek to solve common decision-making problems.