Adam Kucharski

Who they are

Kucharski's biography on his own newsletter says he completed a PhD in applied mathematics and is now a professor at the London School of Hygiene and Tropical Medicine. His research focuses on health threats, including large-scale studies of social behaviour and immunity, and on statistical methodology and open-source software tools. The page states that he has published over 100 research papers, cited over 25,000 times, and has advised multiple governments and health agencies. His publisher, Profile Books, adds that he has worked on global outbreaks including Ebola, Zika and COVID-19, producing real-time analysis for governments.

He is also co-founder of WholeSum, a venture-capital-backed company developing methods for interpreting large volumes of open-ended text. His journalism has appeared in the Observer, the Financial Times and the New Statesman, among other outlets, and he has contributed to documentaries including BBC Horizon. He has spoken at TED, Google and the Royal Institution. These details come from his own page and his publisher's author page and were not checked against a third source.

Key papers

Kucharski is better known on this site for his books than for individual papers, and none of his research papers is drawn on here. Profile Books lists three:

Key ideas

The vault reads Proof from page photographs, and its hub note records these arguments. They are the book's claims, and the vault's audit of the book marks its history and legal material as sound while leaving some trial figures to be traced.

Proving and persuading are different jobs. Kucharski treats proof as two things at once: establishing what is true, and convincing other people of it. The second is often harder.

The 5 per cent rule is a historical accident. In chapter 4, which follows cohort and case-control studies, Gosset's small-sample statistics, Fisher's null hypothesis tests, Neyman's confidence intervals and the randomised trial, he argues that the fixed threshold for significance came from practical problems and merged what had been rival schools of thought. See Ronald Fisher.

Trials are a guide, not an oracle. A randomised trial estimates an average effect, cannot always be run, and does not always transfer to other settings. The evidence hierarchy is best used flexibly, alongside natural experiments. See Randomised Controlled Trial.

Evidence thresholds depend on the stakes. The book's conclusion is practical: how much evidence is enough varies with what is at stake, and acting on the best available knowledge beats waiting for certainty.

The commercial descendant. The vault's notes connect the chapter's randomised trial to the AB tests described by industry practitioners, and record a Bing example in the book. The numbers in that example match the ranking-bug experiment in Kohavi and colleagues (2012). See AB Testing and Ron Kohavi.

---