Wednesday, June 13, 2018

Climate Change as a Game of Russian Roulette


Ugo Bardi speaks at an event on climate change in Florence, on May 25th, 2018. Obviously, what I am holding is a toy, not a real gun. I was discussing guns as a metaphor for our penchant for doing dangerous things without exactly knowing what we are doing - for instance injecting large amounts of greenhouse gases into the atmosphere. With climate, we are playing a dangerous Russian Roulette game involving the whole of humankind (photo courtesy, Ilaria Perissi). 




In a science fiction story I read years ago, the protagonists live in the future and have forgotten what guns are. Then, someone finds a still working handgun and starts playing with it. As you may imagine, the results are tragic.

Now, let's make a small exercise in epistemology. Suppose that you are one of the characters of that sci-fi story. You have never seen a gun before and you would like to understand what it is and what it does. Basically, there are two ways of approaching the question: the scientific/reductionist way and the Bayesian/evolutionary way. Let me explain these concepts.

The scientific/reductionist way. You dismantle the unknown object and try to build a model of its inner workings. You note the mechanical system that makes a metal hammer hit on the chambers of a spinning drum. One of these chambers contains a brass cylinder partly filled with a mixture of chemicals that you can analyze. You find that the mechanical stress generated by the hammer may ignite the chemicals, producing high-pressure gases which might push an ogival chunk of some 100 g of lead through the front barrel at a speed of the order of 300 m/s. If you align the barrel with a human head, the effects of the chunk of lead passing through the brain would be hard to simulate, but they might involve serious damage. You conclude that it is, most likely, a weapon.

The Bayesian/evolutionary way. You examine the gun and try to build a probability estimate based on empirical tests. You note a small lever at the bottom and proceed to pull it, noting that it generates a clicking sound. You pull it a few more times: nothing happens as long as the hammer doesn't hit the loaded drum chamber (which you don't know about since you didn't dismantle the gun). Then, you conclude that it is probably a musical instrument.

The difference in this approach shows mostly if you use the gun to play the Russian Roulette (*) with one bullet in a six-chambers drum. Then, after five "clicks" the frequentist would tell you, "pull the trigger one more time and you are dead." But the Bayesian would say (**), "since you tried five times and nothing happened, then you are reasonably safe if you try once more."

Of course, these two viewpoints are extreme, there are plenty of intermediate ways to approach a problem, but they indicate how difficult it may be to deal with something unknown. And that's the big, big trouble with climate change. It is gigantic, enormous, complicated, and most likely dangerous. But we are like the characters of the science fiction story of the unknown gun: we have no direct past experience to rely upon. Without disparaging the Bayesian method, surely helpful in many cases, it may be a suicidal approach to use when dealing with something dangerous for which you have insufficient statistical data. That's the case of the Russian Roulette and also of climate change.

There follows the question: do people think Bayesian of Frequentist? It is a controversial subject but, personally, I'd say that it makes sense to say that most people think Bayesian. That may be the reason why humankind has such a cavalier attitude toward the danger of climate change. The statistics we have on climate for the recent past don't tell us anything about the possibility of a true catastrophe. So, we might be tempted use a Bayesian approach to conclude that we have no reason to be worried - and the more time goes by without catastrophes occurring, the more this conclusion seems to be reinforced. After all, haven't we pulled the trigger of this thing you call "gun" already five times? It has to be harmless.

Of course, the scientific/reductionist approach tells us otherwise when the climate system is analyzed and modelized. It tells us that the change may be extremely destructive - actually catastrophic. But that approach seems to be reserved for a small fraction of the population trained in the scientific method. There follows that humankind is playing the Russian roulette with the Earth's climate. And that might well end the way a Russian Roulette game must eventually end.



(*) During the past two decades, the number of suicides in the US has increased by some 25% and the most common method was using firearms. A peculiar form of suicide consists in playing the game of the Russian Roulette. There are no worldwide statistics but 10 years of records in Kentucky alone reveal 24 cases of people who killed themselves in this way. Clearly, it is not a form of mass entertainment, but it does happen. It is hard to say what goes on in the mind of the people who engage in this kind of game, but likely it has to do with the fascination we feel for guns. Nobody knows exactly how many small firearms exist in the world, but the number could be more than half a billion. In the US, there is about one gun per person although, of course, they are not evenly distributed. Some people take guns as true objects of worship and some seem to believe in the existence of a Gun God requiring human sacrifices - that may be the ultimate reason why some people play the Russian Roulette.

From the paper by Lisa Shields. "In one situation, a 22-year-old African American man used a 0.22 caliber revolver in a game with a friend. Each participant pulled the trigger on 2 occasions; the victim discharged the fatal bullet on his third attempt. The decedent was a university student, a member of the varsity football team, and was studying electrical engineering with a 3.0 GPA. Blood and urine toxicology screens yielded no ethanol or other drugs. In another circumstance, a 46-year-old divorced white man employed as a custodian engaged in Russian roulette with his “drinking buddy,” a male friend suffering from cancer. After placing 2 shells in a .38 special, the victim died with the first pull of the trigger. Four of the victims had pulled the trigger at least 3 times before their fatality. A 19-year-old white man significantly increased the likelihood of a Russian roulette fatality. He had a history of depression with a previous commitment at a mental hospital and several previous suicide attempts. Playing a variation of traditional Russian roulette with his brother and 2 friends, the victim placed 5 live rounds in the cylinder, leaving one empty chamber, of a .357 Traus revolver. He spun the cylinder, put the gun to his right temple, and pulled the trigger. <..> The decedent had played Russian roulette on 2 occasions in the previous several weeks, each time placing only one live round in the cylinder." 


(**) A formal statement of Bayes' theorem is:



In the problem of the Russian Roulette, X= you pull the trigger and A= you die


P(A|X) is what we want to estimate: what is the probability that you die when pulling the trigger? P(X|A) is the probability that when you die it is because you pulled the trigger. We may take it as equal to one, neglecting the occasional heart attacks that might strike the player when they pull the trigger on an empty chamber. For the P(A)/P(X) term, we need statistical data but it is obvious that, in a limited number of attempts, as long as the player is alive, the more times they pull the trigger, the larger P(X) becomes and hence P(A|X) becomes smaller. So, you wold be led to conclude that the more times you pull the trigger, the safer you are. Then, of course, for a large number of players (and a large number of deaths), the Bayesian analysis will converge to the reductionist results: n*3.5 attempts are needed in order to kill n people, assuming that the drum contains six chambers and one bullet. The problem is that the Russian Roulette doesn't allow a large number of attempts when played by a single person (or by a single planet in the form of a climate catastrophe).

See also this article on freakonometrics

 

Who

Ugo Bardi is a member of the Club of Rome, faculty member of the University of Florence, and the author of "Extracted" (Chelsea Green 2014), "The Seneca Effect" (Springer 2017), and Before the Collapse (Springer 2019)