It is strangely underappreciated that when it comes to global catastrophic or existential risks the future cannot resemble the past.
Catastrophic risks illustrate the opposite of the more famous principle, first explicated about 1788 by the great founder of geology James Hutton, that “the present is key to the past.” The doctrine of uniformitarianism, promoted by Hutton, Lyell, and their followers in 19th and 20th century, actually lulled a large part of the scientific community into a false sense of security about global catastrophes.
Only in the last quarter century has the uniformitarian prejudice been shattered, and a richer, more nuanced “neocatastrophist” view emerged, due to people like Alvarezes, Gould, Palmer, Raup, Hsu, Huggett, Clube, Napier and others.
Why is the future is bound to be different? Because we happen to exist on Earth at this particular moment of cosmic time, which creates our “anthropic bias.” In order to predict the future we need to analyze traces from the past. This has particularly serious consequences when naturally occurring catastrophes, such as asteroid impacts, supervolcanoes and nearby supernovae are investigated. Regardless of the specific methodology pertaining to each individual scientific discipline, there is a common denominator: from the a posteriori (“empirical”) distribution function of such events we need to derive the a priori (“real”) distribution function.
For instance, terrestrial impact craters are the empirical traces of past catastrophes. We commonly use statistics about impact craters, along with assumptions about the relationship of the impactor size and the size of the crater, to derive an a posteriori distribution function for Earth-crossing impactors, that is, how likely we are to be hit by a massive asteroid.
However, in order to calculate the probability of a collision between Earth and an object in a given size range in the next thousand years, we need an a priori distribution, i.e. what “real Nature” offers us. Trying to reason past the anthropic bias of our unique moment in time tells us that we cannot empirically sample the part of the a priori distribution function which is incompatible with our existence at present, i.e. catastrophic events that would have made all life on Earth, or at least human life, impossible. But those probabilities are as real as the rest.
No matter how hard we search we can never hope to find traces of a 20-km asteroid impacting Earth in the last million years. We wouldn’t have been here if such a catastrophe occurred.
This bias skews the basic symmetry between impact events in the past and those in the future. Even though such an impact couldn’t have occurred in our past, there is still a non-zero probability of one occurring in our future. Ditto for similar random disasters from supervolcanoes to naturally-triggered vacuum phase transition.
This means that our existence causes a Bayesian probability shift when stochastic global catastrophes are concerned. This observation selection effect leads to the systematic decrease in the amplitude of the a priori distribution function for the same a posteriori distribution.
In other words, future catastrophes are generally systematically underestimated, with far-reaching and possibly tragic consequences.
Astrophysicist and philosopher Milan M. Ćirković is Assistant Professor of Physics at the University of Novi Sad in Serbia, and Senior Research Associate at the Astronomical Observatory of Belgrade, Yugoslavia.
IEET Blog