# The past is not a predictor of the future

Mark Rosenker is the acting chairman of the National Transport Safety Board. Below is an extract from the BBC’s article about the investigation of the problem that hit BA’s Boeing 777 that crash-landed at Heathrow in January 2008.

Mark Rosenker […] said: "With two of these rollback events occurring within a year, we believe that there is a high probability of something happening again."

The other event he refers to affected a Delta Air Lines flight cruising over Montana, USA in November 2008.

It is worrying that someone who one would expect to have a good understanding of risk uses such ludicrous logic for something as important as air safety.

The fact that the problem occurred twice in ten months is not a predictor of its likelihood of happening again in the future. Just as someone winning the lottery twice does not increase their likelihood of winning it a third time.

3 Responses to “The past is not a predictor of the future”

1. Francis on March 14th, 2009 00:16
2. Tom Viner on March 16th, 2009 08:29

I think your lottery analogy is wrong because we don’t know the probability in advance. I’d compare it to picking names out of a hat where you don’t know how many names other than yours are in the hat and you put the name right back in after picking it. In that situation your best guess of the probability would be P the number of times your name was picked divided by N the total number of picks.

The problem for the NTSB is that they only know when the game started if they know what’s causing the fault and when that part of the aircraft or way of operating was begun or combination there of. So they only know how many times their name was pulled out the hat not how many names there are in the hat.

Of course the fact that two events happened within a year doesn’t mean anything special, it’s just a way of suggesting that maybe N is revatively low.

3. slatfatf on March 18th, 2009 16:32

Your article raises interesting issues. He is using inductive reasoning to formulate what may occur in the past. You maths geeks are automatically focusing on the numbers but in fact it is a reasonable position given human nature to say that this might happen again in the future. He is not saying when or how often. Just that it has happened twice and so it is not unreasonable to extrapolate to it happening in the future.

Your argument seems to be based upon the idea that it is unreasonable for him to KNOW this. It is. But no more than it is reasonable for you to KNOW anything at all based upon inductive reasoning which is just about everything you do in your everyday life. You eat your sandwich in the supposed knowledge that it will be nutritious. It is arguable that you have this knowledge as it is based upon little more than past experience. If you bought a sandwich from the shop across the road and two of them caused near fatal food poisoning would you continue to eat there and if so would you be nervous and say that if you have been poisoned twice you would likely not recommend it to your friends. But according to your point about the guy in the article you would have no grounds at all to no recommend the sandwich shop.