Prevent worldwide disasters with pilotless airplanes by learning from statistics

This article is a translation of ‘Prevent worldwide disasters with pilotless airplanes by learning from statistics’By Theo Kocken, Free University ; in the Financieel Dagblad, 05 December 2018, The Netherlands.

Sometimes you hear people draw conclusions based on numbers. Then I have to think of the wisdom of the American writer and entrepreneur Mark Twain : ‘Lies, damn lies and statistics’; even when true, the conclusions drawn from them often are not.

Take the statement that flying without pilots is safer. Aviation expert Hans Heerkens of Twente University put, in the context of a news item about unmanned cargo (freight) flights in Kenia, that, in 70% of all accidents, pilots are the cause that determine if an accident happens or not. When removed from the cockpit, the pilot can be replaced by a computer that holds over 100 years collective flying experience; experience a pilot can never accumulate (NOS broadcast 27 November 2018). So, flying with computers rather than pilots would cause less accidents.

Let’s assume the 70% is correct. The precise percentage is not relevant for the point made here. What is relevant, is that this information presented above doesn’t say anything about the conclusion made: that flying without pilots is safer. This is because the conclusion is unfounded when based on this and could actually make the world a lot less safe.

Take the news of the recent crash of Lion Air in Indonesia, that took 189 lives. This crash was caused by malfunctioning of the new software installed. The pilots successfully managed to pull the aircraft straight 24 times, despite the software pushing its nose down. Eventually, the pilots lost control. The software and adjacent technology won from humans, who are capable of engaging multiple senses and solving problems creatively. (By the way, the hundred years of collective experience was present in this malfunctioning software).

Learning processes

That the other 30% of all accidents is therefore caused by software and that software too, can be faulty, is not the lesson to be learned here. The emphasis is, that despite several attempts to save the plane by the pilots, it still went wrong. And, that very often pilots are able to make corrections and that thanks to the pilots, an accident does not happen. Only looking at accidents that have materialised, produce hilarious learning processes and conclusions.

Take, for example, the following fictive, but not unrealistic equation. In 10.000 out of 100 mln flights somethinggoes wrong. That is 1 in 10.000 flights. One in 10 mln flights end in a fatal crash. Of the 10 fatal crashes in 100mln flights, 7 are (partially) caused by pilots. Violà, here we have the 70% pilot blunderers!
But of the 9990 flights that didn’t go so well but landed without harm, it is the pilots who overruled the system and saved the day. Unfortunately, in three cases they couldn’t, of which Lion Air is a recent example.

First conclusion: 70% percent of aircraft crashes are directly or indirectly caused by pilots, in absolute numbers, 7. Second conclusion: pilots prevented 9990 disasters. So, pilots have ‘caused’ 7 disasters and prevented 9990 disasters, which is approximately 1500 times as many.

This article isn’t about numerical example exercises. It is about the fact that you can not say that flying without pilots is safer based on the statistic that seven out of ten accidents are caused by pilots’ actions, without knowing in how many near-misses the pilots actions have lead to a safe landing.

It reminds me of the story of the planes in WW II that returned from battle. The cabins were always riddled by bullets, but the wings and cockpit were never hit. So, the cabins were strengthened, as cabins were obviously a target causing danger to the plane. What was failed to realise is that planes hit in the cockpit or the wings, never returned as these planes had become uncontrollable and crashed. ‘Survivalship bias’ this is called in the field of risk management. During WW II the wrong conclusions were drawn from the planes that returned. And in the case of pilotless planes, wrong conclusions are drawn from a few fatal accidents.

Naturally, the number of victims in an aircraft will be reduced when there are no people on board. Statistically, it is probably safe to say that no human lives on board an aircraft will be lost, no passenger or crew victims to mourn, when flying aircraft completely unmanned. But far more unconformable is a prediction of what will be the numbers of victims on ground, when we just leave pilots out of the cockpit based on the ‘7 out of 10’ statistic.

Many people, understandably, struggle with statistics. But saying something about safety of transport based on statistics is dangerous when pitfalls related to statistics are not very well understood. Fortunately, the aviation-industry understands the strong combination of humans and technology very well. The industry realises that much can be prevented by people’s actions where computers cannot (yet) solve the problems as computers do not understand their own flaws. Unfortunately, the pilots of Lion Air couldn’t win this time, despite their attempts. The multiple times that pilots have succeeded, will not make it to the news, but they do improve flight safety.

I would not enter a flight without pilots. Not because I do not like innovations, but because I understand the statistics behind real disasters and near-misses. And for who is going to Kenia for their holidays, you may want to look up a few times more often…

Author: Theo Kocken, Free University Amsterdam. Email: t.p.kocken@nu.nl
Published by Financeel Dagblad 05 December 2018. Reactions to: expert@fd.nl
Translated by Margriet Bredewold.

Leave a Reply

Your e-mail address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.