Human Error

Days like today bring some more general reflections. Yesterday's tragic crash of presidential plane killed 96 people, among them the President of Poland.

But the World never goes to a stand still. As time goes by, the human nature is to learn from mistakes and improve the collective knowledge, wisdom and experience to avoid such catastrophes in future. Punishments (as Jeffrey Satinover points out in The Quantum Brain), together with rewards, are essential base-forming ingredients of the learning curve of our civilization.

Speaking of the learning curve - the rate of air crash related fatalities has dropped tenfold over the last thirty years. This by itself is a proof we learn and draw conclusions.

In the recent years there has been a far and wide reaching trend of implementing various computer systems aimed to improve flight safety. And the wisdom of the crowd is they are usually doing the opposite. Every now and then we read media reports of aviation accidents where computers are pointed as primary causes of accidents. I have had many discussions on that. Even not touching aviation, people often point car ABS systems as not being able to brake as efficient as a non-aided human driver. That may be true, but only in ideal, predictable conditions, when the driver is focused on his performance. And this is the point. Fatigue and distractions are concepts not known in software algorithms and electronic brains. So is mental pressure and audacity. In vast majority of situations ABS saves lives. Because we are not top performing race drivers, and we are "just" humans, meaning we happen to be distracted and tired and surprised with sudden and unexpected conditions.

The same principle applies to aviation safety. We are eager to discuss the potential contribution of fly - by - wire system failures, especially to the accidents, that have not been fully explained. Like the June 2009 AF447. Deep in our minds we tend to favor smart humans over dumb computers. But computer failures do not explain why the AF447 flew straight into an inter tropical convergent zone cumulonimbus build-out, while the fuel calculations potentially do. And here we touch another factor, frequently contributing to disasters, which is cost cutting, a direct derivative of greed.

Of course there are many well known accidents, that simply could not have been avoided simply because we did not fully understand the underlying conditions. Like the fatal series of Comet disasters of 1954 caused by previously unknown phenomena of metal fatigue. Or the loss of the Columbia space shuttle. But statistically, like it or not, majority (more than 50%) of aviation accidents are caused by human errors. At the same time the big decline in fatalities over the last thirty years has to be contributed to increased use of computers assisting humans in difficult tasks of piloting and navigating aircrafts. Computers are far from perfect though - we still learn and improve both the hardware and algorithms. But their unspoken strength is they always obey the rules, while humans repeatedly tend to bend them, as if they were not able to learn from the past mistakes.