Randomness, Explanation, Parfit and the Cosmos

It is widely accepted that explanation is a factive relation, that is, that it holds among facts (or true propositions, the distinction is not relevant for our purposes): if A explains B, A and B must be actual facts. Beyond that, it is a huge puzzle to determine what else is required for A to explain B. Some philosophers have defended the view that necessitation is another necessary condition for explanation, that is, that for A to explain B, A must necessitate or metaphysically entail B. The rationale for this condition is simple: for A to explain B, A must tell us why B happened instead of anything else that could have happened. If A does not entail B, then A is consistent with some possible outcome C different from B and this cannot explain why B happened instead of C.

I (in Barceló 2015) and others have argued against the necessity condition. An interesting counterexample to the hypothesis that necessitation is required for explanation are what I will call explanations by randomness. Both in physics and biology, it is not uncommon for random processes to be invoquen in the explanation of phenomena (see, for example Bak et al.1987, Sanders 1986, Santos et.al. 1989, etc.). By the very nature of randomness, these explanations do not satisfy necessitation, for if a process is truly random, it must be compatible with more than one possible outcome.

Derek Parfit gives a very nice example of this kind of explanation in (1998):

Suppose first that, of a thousand people facing death, only one can be rescued. If there is a lottery to pick this one survivor, and I win, I would be very lucky. But there might be nothing here that needed to be explained. Someone had to win, and why not me?

The basic idea is that if a process is truly random, as in a lottery, the question of why it had the outcome it had (instead of any of its other possible outcomes) is nonsensical. Appealing to the random nature of the process is enough of an explanation.

Interestingly enough, Parfit continues his paragraph with the following:

Consider next another lottery. Unless my gaoler picks the longest of a thousand straws, I shall be shot. If my gaoler picks that straw, there would be something to be explained. It would not be enough to say, ‘This result was as likely as any other.’ In the first lottery, nothing special happened: whatever the result, someone’s life would be saved. In this second lottery, the result was special, since, of the thousand possible results, only one would save a life. Why was this special result also what happened? Though this might be a coincidence, the chance of that is only one in a thousand. I could be almost certain that, like Dostoevsky’s mock execution, this lottery was rigged.

It is clear enough that Parfit takes  this new case as substantially different from his original lottery example and that he thinks that something extra needs to be said in this case in order to explain the lottery’s outcome. I think Parfit is right about this, but one needs to be very careful as to what this extra stuff is and what it tells is about the difference between these two cases in terms of whether and in what conditions the randomness of a process can explain its outcome. One might think that the difference between these two cases is that only in the first case it is enough to appeal to randomness to explain the outcome, while in the second something else besides the randomness of the process is required. However, the last sentence makes it clear that this is not so, because in that second case one can be almost certain that the process was not actually random. Thus, this second case is not a counter-example to the claim that whenever a process is actually random nothing extra is necessary for the outcome to be explained. All it tells us is that one must be careful about making sure that the process is actually random and not just apparently so.

Parfit is right in calling attention to this very important question. The issue of how to tell whether a state of affairs is the outcome of a random process, instead of an intentional design, is central to the philosophy of mind, but also to the philosophy of technology and the epistemological foundations of anthropology. After all, it is essential for the anthropologist to do her job to be able to tell whether an object or arrangement of objects is an artefact or not. Furthermore, Parfit is interested in it because it is necessary to answer a more specific question: was the universe designed? or, to be more precise, were the initial conditions of the cosmos truly random or the product of some kind of design? In this context, Parfit’s distinction does not challenge the cosmogenetic claim that if the initial conditions were random, there is nothing needs to be explained. It just claims that, given the low probability of any random process producing the fine tuned initial conditions required for life, much else needs to be said to prevent us from rejecting the hypothesis that those initial conditions were the result of a random process (leaving aside, for a moment, the question of whether it makes sense to talk about processes before the existence of the cosmos or not and/or whether explanation by randomness applies in cases where one cannot talk about process and result).

According to Parfit, if the actual outcome was just as likely as any other, this gives us good reason to believe that it was the result of some random process; but the less likely the actual outcome is from other possible outcomes, the less rational it is for us to believe that it was the result of an actual random process (unless further evidence of randomness is provided). This may be right (even tautologically so), but it is misleading to make this point by saying “something extra needs to be explained”, for it seems to imply that the issue at hand is whether randomness explains the outcome or not, when the question is whether randomness was the case or not. At the beginning of this text I mentioned that explanation is a factive relation, thus for something to explain something else, both things must be the case. This means that an outcome cannot be explained by appealing to the random nature of the process that produced it if the process that produced was not actually random. So telling us that, before appealing to randomness to explain a phenomenon one must be careful to make sure that the phenomenon was actually random does not tell us much about the nature of explanation. This means that if the initial conditions of the universe were actually random this would indeed answer the cosmological question why is the cosmos as it is?

Derek Parfit (1998) “Why anything? Why this?”, London Review of Books, Vol. 20 No. 2 · 22 January 1998, pages 24-27

Per Bak, Chao Tang, and Kurt Wiesenfeld (1987) “Self-organized criticality: An explanation of the 1/f noise”, Phys. Rev. Lett. 59, 381

Dale Sanders  (1986), "Generalized kinetic analysis of ion-driven cotransport systems: II. Random ligand binding as a simple explanation for non-Michaelian kinetics”, The Journal of Membrane Biology, February 1986, Volume 90, Issue 1, pp 67-87

Mauro Santos, Alfredo Ruiz and Antonio Fontdevila (1989), “The Evolutionary History of Drosophila buzzatii. XIII. Random Differentiation as a Partial Explanation of Chromosomal Variation in a Structured Natural Population”, The American Naturalist, Vol. 133, No. 2 (Feb., 1989), pp. 183-197