Why Did the Bayesian Sink?

A man, a boat, a storm. The man, Mike Lynch, a ‘tech’ billionaire. The boat, a ‘super-yacht’. (All these words one did know ten years ago.) The storm, a ‘waterspout’. The boat was called Bayesian. Ah, well, our modern ‘tech’ people like to be scientifically literate. Musk with his ‘Tesla’. And now Lynch with his ‘Bayesian’.

There are three levels of interest about the sinking of the Bayesian.

The first thing that makes the story of the Bayesian interesting is obvious: we have a story of the rich in a yacht in a storm, a story which ends not tragically but sadly. Not tragically, because it was apparently uncaused by the acts of Lynch and his friends and family and their crew. Sadly, because it was unexpected: an ‘act of God’, as we might say, or simply a sign of the arbitrariness of the goddess Fortuna. It is very sad: no one can hear of the story without regretting the accident. The fate of the daughter, due to go to Oxford, is a particularly poignant element.

The second thing is the mystery of what exactly caused it: what caused this yacht to go down so spectacularly. There is something of a miniature Titanic about the story: the privileged going down as a consequence of a natural event. But there are other speculations: that, perhaps, it was caused by human error; or, even more beguilingly, that it was caused by natural events as caused by human error – and here we have to sadly notice the Guardian article which argued that the “climate crisis fuelled the storm that sank the yacht” – “say experts”, of course.

But I’d say that there is a third and deeper enigma about this story and it is found in the name of the boat. Bayesian. There is something sinister about that name. At first it made me think of the Shakespearian word Bezonian, and the Miltonian word Serbonian: associations of choosing your king carefully when talking to a king (political dangers), and of armies drowned in the depths (natural dangers). Some commentators have noticed the name, Bayesian. Charles Moore made a wry comment on it in his Spectator piece. (He was due to meet Lynch for lunch next month.) The Guardian observed that the boat was named after Thomas Bayes, clergyman responsible for a theory which inspired Lynch’s work. I wonder what proportion of its readers looked a bit further into the subject.

Well, the enigma, the oddity, here is that Lunch named his boat Bayesian but that the boat Bayesian behaved in what seems to have been a desperately unbayesian way. This is what I would suggest is an almost tragic irony: and why it is the name that drags the whole story to a darker, more tragic, possibility.

For the death of Lynch might have been tragic. The boat presented a challenge to the same YHWH – that’s God to me and you – who dealt briskly with the Nimrod who built the Tower of Babel. Pride comes before a fall, and all that. Lynch, in naming his boat Bayesian, appeared to be building the strangest of all babels, one which floated, like an ark, like the Titanic, but which was kept in the pleasurable waters of the Mediterranean, and which had the second tallest mast in the world. (75 metres on a 56 metre boat.) Building tall objects always risks angering a thunder god, does it not?. And Lynch called this Babel Bayesian.

Not even ‘Bayes’, signifying, “I admire a man”, but ‘Bayesian’, as if not only exalting but embodying the very theorem of Bayes.

What is the theorem of Bayes?

Well, here, almost immediately, I come to an abyss where knowledge appears to exist and I am not the master of it. But anyone with eyes to see can observe an irony.

Tom Chivers this year published Everything is Predictable: How Bayes’ Remarkable Theorem Explains the World. The tagline of the book is, “Can you predict the future? Yes of course you can!” (Also, and of some interest to us: “Why are conspiracy theories effective and how can you combat them?”) Well, of course, the owner of the Bayesian, alas, failed to predict the future.

Bayesian probability is highly technical. It depends on a formula:

P(AB) = (P(BA)(P(A))/(P(B))

or “the probability of A given B” is “the probability of B given A” multiplied by “the probability of A” all divided by “the probability of B”. This means very little to me, but the salient point seems to be that in the 18th Century Bayes originated one of two major ways of explaining probability. It is the lesser one, the minority report.

The majority report way of explaining probability, the non-Bayesian one, is objective. It ignores what people believe. It deals only in the facts, the ‘hard’ facts, we might say. This is sometimes called frequentist probability: it involves looking at how often something has happened in the past and then claiming to know what probability there is that it will happen again in the future. It treats probability as a property of the world.

Bayesian probability, if I understand it right, treats probability as a property of the mind. In other words, it deals not with hard facts, but with the world as affected by our beliefs. It includes subjectivity. Interestingly, it involves the view that our expectation of what will happen is an element in whether it happens or not. It is, I am told, not that common in science, but is important for risk assessment, and the statistical analysis of dynamic systems.

The question is, is probability a property of reality or a property of our awareness of reality? In other words, is it objective, or subjective: it is to do with our minds, or not?

Difficult questions. But it brings us back to an essential point which is that whatever view we take of science, and whether scientific probability should be objective or include subjectivity, there is no question that politics is where we find muddy realities, with no clarity, no ‘solutions’, only guesses, proposals, compromises and bits of force. Life is more like politics than science. Could Lynch predict the future? No. Obviously not. And even with his admiration for Bayesian logic and statistics, he was no wiser than any of us about what was in store for him.

“Science,” writes Chivers, “is explicitly about making predictions – hypotheses ¬– and testing them. The problem is that in science, we like to think that there is an objective truth out there, and the Bayesian model of perception is explicitly subjective. A probability estimate isn’t some fact about the world, but my best guess of the world, given the information I have.”

Lynch not only made his best guess, but pinned that standard to the mast by naming his super-yacht after a theorem about making best guesses. He claimed that probabilities were not properties of the world, and, like Nimrod, challenged YHWH to prove him wrong. He was proved wrong.

There is more to this story. A detail mentioned sometimes, but not always, is that Lynch was celebrating his acquittal in a vast U.S. fraud case: and that his co-defendant in the case, Stephen Chamberlain, was killed in Cambridgeshire on Saturday 17th: hit by a car near Ely while out running. One defendant was fatally struck on Saturday, and his death was announced late on Monday the 19th; after the sinking of the Bayesian on the same day, Monday, though Lynch’s death was only confirmed on Friday. What are the odds? Such a remarkable coincidence seems to have something about it: not conspiratorial, there is no need for that, but certainly tragic, as if it were destined. Lynch apparently told a friend that the acquittal, in June of this year, had given him a “second life”. Well!

The only other writer I have seen who has noticed all of this is Rowan Pelling of Perspective Media: I found her piece after writing most of the above. She was so surprised by the coincidence that she wrote to the novelist D.B.C. Pierre, who replied: “It is the stuff of novels, to be freshly acquitted along with a partner in a 10-year fraud trial and then each be nailed in a one-in-a-million act of God on the same day. But even Putin couldn’t pull that off, especially the water spout.” Pelling speculates that we are seeing the return of the pagan gods. Well, whoever: YHWH, Homeric gods, the goddess Fortuna. It is a sobering story.

Bayesian probabilities might be better than the other sort, but the boat called Bayesian still rests on the floor of the Mediterranean.

Dr. James Alexander is a Professor in the Department of Political Science at Bilkent University in Turkey.