Survivorship bias

Chart of observed damage patterns on World War II bombers. But only bombers that returned to base were observed. Note the areas that don't show damage.

Cogito ergo sum Logic and rhetoric

Key articles

Logical fallacy Syllogism Argument

General logic

Axiom Poisoning the well Nirvana fallacy Argumentum ad hysteria Viés de sobrevivência How do you know? Were you there?

Bad logic

Quote mining Common sense Argumentum ad cellarium Blaming the victim Didit fallacy گزاره‌ی صاحب‌نظر v - t - e

“”Once you are familiar with the idea of survivorship bias, you can start spotting it everywhere. For example, a gym might feature those who’ve toned up quickly as a result of going to their facilities but, of course, what they never show is those who signed up but achieved no more than a depleted bank account. Another example is when a company is tempted to copy the strategy of a rival. A policy of "radical transparency" that includes blunt public feedback may work for Netflix and firing the 10% most "under-performing" staff might have been effective at General Electric. But before adopting any of these policies where you work, you should be wary of only looking at their use in successful companies.

—Brendan Miller, "How ‘survivorship bias’ can cause you to make mistakes", BBC News.

Survivorship bias is a cognitive bias that occurs when focusing on entities that made it past a selection process, while overlooking those that didn't. This can lead to incorrect conclusions due to missing information. It's a specific type of selection bias. An example of survivorship bias in action would be to draw overly optimistic conclusions from analyzing the performance of businesses in a certain sector using data that excludes businesses which failed or never successfully started up.

A famous story about Diagoras of Melos, a 5th century BCE Greek atheist philosopher, shows him criticizing the efficacy of prayer by noting that people who prayed at sea and survived disasters were around to tell of it, while people who prayed and died during disasters were obviously no longer around to say anything.

An in-depth example

The following comes courtesy of You Are Not So Smart.

Suppose you're trying to help the military decide how best to arm their planes for future bombing runs. They let you look over the planes that made it back, and you note that some areas get shot heavily, while other areas hardly get shot at all. So, you should increase the armor on the areas that get shot, right?

Wrong! These are the planes that got shot and survived. It stands to reason that on some planes, the areas that were chewed to bits were not essential, whereas the areas where you don't see any damage did get shot on other planes, but those planes didn't survive. So those are the areas you reinforce. This was the brilliant deduction of Abraham Wald, a Hungarian-born Jewish statistician who fled Europe to work for the US military during World War II, which also goes to show that one shouldn't try to kill your best thinkers.

Other examples

One might look at sturdy old buildings and compare to modern constructions, which are often cheaply done and routinely leak, and assume that people back in the day knew how to build things to last. But you are ignoring all the terrible buildings that were either torn down or subject to acts of God; only the best built buildings would survive for centuries. Rome may have buildings that survived for millennia (with constant repairs), but virtually none of the original housing remains as anything more than a construction hazard.

Survivorship bias is also at play when considering the quality of artistic works throughout history. It's easy to look at Shakespeare and think that writers today are much less intelligent than they were in his day, but there were also plenty of writers of Shakespeare's day whose work wasn't as good, and so either didn't survive into the modern era or lacked the influence on Anglophone discourse that Shakespeare achieved. Used in this way, survivorship bias can lead to nostalgia for an imagined glorious past, hence the pop-culture terms 'nostalgia filter' or 'nostalgia goggles' which describe this effect.

Fans of foreign films may argue that their production tops those of Hollywood films, or that they are simply more exemplary of the "art of cinema". However, one must consider that films well-received in their home country are more likely to be ported for an international audience. Similarly, younger fans of classic rock music may only be exposed to the music that people who lived through the era liked; there are many hits that you won't find on today's classic rock radio, YouTube compilation, or your parent's record collection.

Survivorship bias can obscure the effects of workplace exposure upon health problems. When new employees with prior exposure to, say, asbestos or silica are inappropriately combined with employees without prior exposure, the apparent exposure effect is reduced. Employees exposed at prior jobs become ill sooner than previously unexposed employees, making exposure of the current employees appear negligible.

Michael Shermer of Scientific American cites survivor bias in books about business success and rags-to-riches stories. In one example, he observes that Steve Jobs dropped out of university to start up Apple Computer, which became a billion-dollar business, but that the failures made by those imitating Jobs' path to success were ignored. Also, Apple owes much of its success to Steve Wozniak, who did go to college for electrical engineering.

See also

Se você procura pelo artigo em Português, ver Viés de sobrevivência.