William of Occam, Thomas Bayes, and the Fate of MH370

By James Fallows

Three perspectives worth mentioning this morning. 

1) Mysterious disappearances were once the norm. A reader who runs a tech firm writes:

It's interesting to remember that this sort of mysterious disappearance was completely normal until comparatively recently. I'm not sure, but I think theTitanic (1912) may have been the first shipping disaster that played out in the press in nearly real time. Before 1899, the first news of a ship lost at sea was likely to be no news at all.

2) Eponyms of logic, Occam and Bayes. Most people easily grasp (though often stray from) the logical concept known as Occam's razor. It's the idea that, other things being equal, the simpler explanation for an event is more likely to be true. For instance, early in the MH370 mystery, one popular conjecture was that something terrible had gone wrong that impaired the pilots' ability to control the plane. And another popular one was that those pilots had found a way to sneak up underneath another plane, hide in its "radar shadow," and then peel off undetectably and land at a secretly arranged rendezvous site in Pakistan or Iran. Knowing nothing else, by Occam's razor sheer complexity made the second far less likely to be true. 

I find that people have more trouble with the concept of Bayesian statistics or probability, or simply the name, even though the simplest version of its implications makes common sense. That simplest version is the idea that probability estimates can be continually improved and refined if they are adjusted to reflect past experience or new evidence.

Thus the image at the top of the page (taken from here, as is the example that follows).  Acoustically, the phrases "I want four candles" and "I want fork handles" are practically identical, and if you listened to a recording with no other info, it would be 50/50 which statement the speaker had in mind. But if you're hearing this in a candle store, the probability changes in one way -- having nothing to do with the actual sound -- and in a cutlery store it changes the other way. There's much more to the concept, but that is the main idea. [Mea culpa! I did not know about the "Two Ronnies" episode on Fork Handles, but now I do.]

My reason for bringing this up is to point toward an interesting short book I read last year, which is all about this history of Bayes's approach and its modern implications. (Plus, why it probably should have been named not for the English clergyman/mathematician  Thomas Bayes but for the French mathematician Pierre Simon Laplace.) The book is The Theory That Would Not Die, by Sharon Bertsch McGrayne.

3) Clear thinking about MH370. Here are two examples. One is an article by Les Abend, a long-time 777 pilot whose sane-sounding judgments I have praised before. On the CNN site today he gives a chronology that discards some wild implausibilities and explains how a mechanical problem could have led to the evidence we now have.  

The other is an update from Chris Goodfellow, who ten days ago offered the first plausible-seeming discussion of why mechanical/electrical error --rather than hijacking, terrorism, or suicide -- was the least-implausible explanation for what went wrong. His views got a churlish early dismissal from Slate and some TV pundits, but I think they have held up better than some other theories. Today he explains how the latest evidence affects his interpretation, and mentions the Les Abend post. (This is on Google+ rather than a normal blog so you may need to prowl around a little.)

Previous post                                                                       Next post

This article available online at:

http://www.theatlantic.com/technology/archive/2014/03/william-of-occam-thomas-bayes-and-the-fate-of-mh370/359551/