I am currently re-reading “The Black Swan: the impact of the highly improbable” by Nassim Taleb. In chapter 8, he writes about the problem of silent evidence, otherwise known as the survivorship bias. To illustrate this bias, ask yourself: what is the average salary for an actor? Odds are that you will overestimate the figure. Why? Because you are likely thinking of successful actors rather than the entire cohort. There are plenty of aspiring actors working in the Cheesecake factory who you never hear about.
The important feature of the survivorship bias is that we can be led to falsely believe that an entity ranking among the survivors has a special property which is not shared by the unheard of failures.
One possible property that we can be mistakenly led to believe a survivor possesses is that of “antifragility”. This concept was introduced in Taleb’s other book “Antifragile: things that gain from disorder”. Loosely speaking, something is antifragile if it gets better when subjected to shocks. Over time, we should find that such things will win out. When we select survivors in a cohort then, we should find that the antifragile are among them. It would be in error though to reason that being among the survivors implies antifragility: Antifragile entities are more likely to be among the survivors, but being among the survivors is no guarantee of antifragility.