Does everything reduce to physics? 

“If anything along the lines of the complete fundamental theory we have been trying to imagine here is true (after all) some crude, foggy, reflexive, largely unconscious but perfectly serviceable acquaintance with that distribution will have been hard-wired into us as far back as when we were fish, as far back (indeed) as when we were slime, by natural selection – and lies buried at the very heart of the deep instinctive primordial unarticulated feel of the world. If anything along the lines of the complete fundamental theory we have been trying to imagine here is true (after all) the penalty for expecting anything else, the penalty for expecting anything to the contrary, is extinction.” 

Many philosophers (and scientists) have denied the complete reducibility of special sciences to physics, citing physics’ inability to explain higher-level phenomena. Philip Kitcher, for example, writes that an explanation from physics will always result in making something that the special sciences (biology, chemistry, economics, history) would tell us was quite likely to happen seem like a funny, enormous coincidence. Kitcher uses an explanation from biology to outline why he thinks this is the case.  Physician John Arbuthnot discovered in the 17th century that in each year between 1623 and 1700, more male babies were born than female babies. If we explain this phenomenon by means of an explanation from biology, we’ll say something like “in a population in which sex rations depart from 1:1 at sexual maturity, there will be a selective advantage to a tendency to produce the underrepresented sex,” and “human boys were more likely than girls to die before reaching puberty between 1623 and 1700, and as such the birth sex ratio between 1623 and 1700 favored males.”

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

An explanation from physics, on the other hand, would say something like: “these were the physicochemical details of the 1st copulation-followed-by pregnancy between 1623 and 1700 and these details result in a child of male sex, and these were the physicochemical details of the ith copulation-followed-by pregnancy 1623 and 1700 and these details result in a child of the female sex, and you can see now that when we take the difference between the births of the first type resulting in male babies and the births of the second type resulting in female babies there were more male than female babies born between 1623 and 1700, and that is why the years between 1623 and 1700 had the birth ration that they did.” 

This explanation, Kitcher argues, fails to show us that Arbuthnot’s regularity was anything more than a “gigantic coincidence.” The explanation from biology, on the other hand, carries explanatory power – it tells us why, for a large population, all years are likely to have more male babies than female.

Albert responds to Kitcher by saying, actually, no, at the level of physics it was not a coincidence that more male babies were born than female in the time period in question, because the fundamental physical laws of the world include a probability-distribution over possible microscopic initial conditions. We know already this must be true because otherwise the things we are used to describing in terms of physics – the projectile of a ball, the movement of a level – would not be predictable and explainable in the way we know them to be. While we clearly do not regularly, explicitly consult such a distribution, this is not evidence against the hypothesis that such a distribution exists, nor is it evidence against the hypothesis that such a distribution is the “sole ultimate arbiter” of what is and is not coincidental.  And, again, our proven confidence in predicting the behavior of a trebuchet or the movement of gas around its container suggests that we already operate on the assumption that this distribution exists. This being the case, then, if more male babies than female babies were born between 1623 and 1700, then the physical details of male babies being born during that period would have been shown, in the probability-distribution (were we able to see it), to have a higher probability than the physical details of female babies being born during that period. There is no coincidence at the level of physics here.

 

As a quick aside, it does feel perhaps a little strange that Albert is saying yes yes, I know we cannot consult this distribution, but trust me this distribution exists and when things happen in the world that must mean they had a higher probability of occurring in the distribution because they ended up happening in the world! But, if I’ve understood his stance correctly, that is what he is saying, and, when we think about how we handle probability in physics generally, that is indeed what we do all the time. For those who practice physics, their first task is to study how fundamental physical particles move in the world, and then from those empirical observations about how things really are in the world, they begin to draw conclusions about what fundamental physical particles are likely to do and not do, which is to say what they have a high probability of doing. In other words, physicists see things happening in the world, and say those things must have had a high probability of happening in the future because they are happening now. This kind of reasoning rests on the assumption that the things happening now also had a high probability of happening. So, as Albert writes, “it is hard to see what Philip can possibly have in mind in supposing that something can amount to a ‘gigantic coincidence’ from the standpoint of the true and complete and universal fundamental physical theory of the world and yet (somehow or other) not be.”

Callender and Cohen respond to Albert that it is not obvious –that there is “not a shred of evidence” - that the probabilities of events occurring according to a probability space articulated in terms of statistical mechanics will be equal to the probabilities of events occurring according to a probability space articulated in terms of biology, or economics, or any other special science. Statistical mechanics describes the world at the level of microphysical phenomena. Something like biology describes the world at the level of biological phenomena. Callender and Cohen argue that because we describe a biological event with different vocabulary than we do a physical event, we can’t blindly assume that distributing chances over events described at the level of biology versus events described at the level of physics won’t be sensitive to those differences in vocabulary. And, from this it follows that we might not get the same probabilities at the level of a special science as we did at the level of statistical mechanics, precisely because we’re using a different vocabulary to carve the space of logical possibility. As they write:

 

A chance is relative to a particular measure over a particular state space. The statistical mechanical chance is based on Lebesgue measure over phase space (3n-positional dimensions and 3n-momenta dimensions, where n is the number of particles). Ecological systems are sometimes modeled via state spaces with measures on them—sometimes even Lebesgue measure. However, the physical and ecological chances are unrelated, to our knowledge, because the state spaces are different. The (classical) physical one is parametrized with respect to position and conjugate momentum, the ecological ones are parametrized with respect to ecological variables, such as ‘number of age 1 females’, ‘number of age 2 females’, and so on. Are the generalizations that are highly probable in the one space highly probable in the other? We have no idea, and neither does anyone else. The solution in question requires that all of this work out, but we don’t see any reason for such confidence.[2]

 

Albert’s argument assumes that the way of describing logical space that is relevant to Economics is the same as the way of describing logical space that’s relevant to statistical mechanics. Callender and Cohen reply that it is not. Who is right?

Callender and Cohen would like to argue that because the logical space that is relevant to a special science might be different from the logical space that’s relevant to physics we can’t be certain that the same probabilities are distributed across both. I don’t dispute that if we only take information pertaining to, say, ecological properties into account we will end up with different probabilities than those described for the same macro-properties under a statistical mechanical view. But I don’t think practicing scientists make it a habit to, or are even capable of, taking only the properties that pertain to their field of interest into account. A statistical mechanical view takes into consideration all the microstates that constitute all the macro-states in the universe, not just those that we have explicitly decided as relevant to Ecology. And the fact of the matter remains that, so long as physicalism is true, the ecological properties about the world that we pick out as relevant and name as ‘ecological’ are, under-the-hood, specific arrangements of microphysical particles situated in a certain position, moving at a certain rate, and possessing other fundamental properties along with every other microphysical particle in the universe. So even if we choose to focus on higher-level ecological properties, those properties are fixed by microphysical particles whose evolutions can be described and predicted by statistical mechanics. Callender and Cohen are certainly correct that we needn’t concern ourselves directly with particle physics when we would like to predict population dynamics. But that does not mean that particle physics stops baring on the outcomes of ecological phenomena that we witness and add to our mental bank of ecological properties. When I talk about bunnies and copulation I am not explicitly describing the micro-conditions of those two entities/phenomena and all the other micro-conditions of all the other phenomena that interact with them. But I am implicitly including them in my probability calculations.

Pretend for a moment that Albert’s probability distribution of the universe over initial conditions is directly accessible - that is, suppose we can read it, make predictions from it, copy things down from it, etc. Suppose also that an ecologist makes a probability distribution of her own, but instead of using predicates pertaining to statistical mechanics she uses ones that she has decided are relevant to the science of Ecology.  One point in her, say, three-point system represents “Bunny A being pregnant,” and another represents “Bunny A giving birth.” The third represents “Bunny A not giving birth.” From an ecological view, it is most likely that, in this system, “a bunny being pregnant” will evolve into “that bunny giving birth,” because, all else equal, the patterns we’ve found in ecology suggest that a state of being pregnant evolves into a state of giving birth.  What about from the statistical mechanical view? The system described in terms of statistical mechanics might assign a different probability to Bunny A’s giving birth, but only because the statistical mechanical probability distribution is armed with complete information on the micro-conditions of the entire universe, not in properties the ecologist has singled out as relevant. That is, the statistical mechanical probability distribution is better informed about what is likely to happen precisely because it takes quite literally everything into account when it assigns probabilities. Pregnant bunnies exist not in vacuums but as macro entities in a universe that is comprised, fundamentally, of microphysical particles. Ecologists and biologists and economists make predictions about the behavior of bunnies and cells and supply-demand responses, and they make these predictions based on how they have observed bunnies and cells and supply-demand responses to have acted in the past. These previous actions, are, of course, macro-events constituted by micro-physical particles moving and buzzing and existing in causal relations to all the other microphysical particles in the universe. A probability distribution at the level of statistical mechanics takes this fact explicitly into consideration in the same way that ecologists and biologists and economists implicitly do when they assume that bunnies are likely to act in a way that they have already. This is all to say that I don’t think it’s the case that an ecologist practicing as an ecologist usually does (by making predictions based on observations of a non-vacuum past) could come up with different probabilities of an event happening than those probabilities described by a probability distribution of the universe, precisely because the ecologist’s predictions are governed by observations – specific, ecological observations – of a world that obeys the laws of statistical mechanics. An ecologist could try and come up with vacuum-probabilities – probabilities that do their best to disregard the rest of the world and focus only on ‘ecological’ properties, thereby differing in their predictions from the statistical mechanical probability distribution - but I struggle a.) to see how an ecologist would go about doing this, because, after all, the things we decide to be laws at the level of ecology were decreed so after we observed macro-entities that reduce to microphysical particles taking up space in statistical mechanical phase space and b.) to imagine why we would want to come up with a probability distribution that is less informative than one that predicts over the entire universe.

 

I think I’ll end with Albert’s words, because as well as being a brilliant philosopher he was somewhat of a poet:

 

“If anything along the lines of the complete fundamental theory we have been trying to imagine here is true (after all) some crude, foggy, reflexive, largely unconscious but perfectly serviceable acquaintance with that distribution will have been hard-wired into us as far back as when we were fish, as far back (indeed) as when we were slime, by natural selection – and lies buried at the very heart of the deep instinctive primordial unarticulated feel of the world. If anything along the lines of the complete fundamental theory we have been trying to imagine here is true (after all) the penalty for expecting anything else, the penalty for expecting anything to the contrary, is extinction.”

907px-John_Arbuthnot,_physician_and_man_

John Arbuthnot's analysis of sex ratio patterns is credited as the first-ever use of significance tests, and, more impressively, the first example of formal reasoning about statistical significance and moral certainty.