Limits of a statistical mechanical account of the arrow of time

Abstract

The temporal aspect of our experience exhibits a distinctive asymmetry; time seems to have a direction associated with it. This is manifested in different ways: our present actions seem to have consequences for the future but not the past; causes always precede effects; our knowledge of the past is different from our knowledge of the future; we perceive processes like the dispersion of ink in water and the growth of a tree from a seed that never happen in reverse. How are all these temporal asymmetries related, and what explains this feature of time? In addition, there are physical laws--such as the 2nd law of thermodynamics (that the entropy of an isolated system must never decrease)--that exhibit such a temporal asymmetry, while more fundamental laws--like those governing classical mechanics--do not. There has been a long tradition of appealing to statistical mechanics to explain the temporal directedness of thermodynamic phenomena, and recently David Albert, Barry Loewer, and others have attempted to explain all temporally asymmetric phenomena using a statistical mechanical account. This thesis argues that while such an account is more fruitful than previous attempts, it doesn’t satisfactorily explain the temporal asymmetry described by the 2nd law of thermodynamics and similarly doesn’t explain the various temporal asymmetries listed above. However, I argue that the above asymmetries are indeed intertwined with each other and with statistical mechanics, but in a different way than that proposed by Albert and Loewer. Specifically, the asymmetry of causation (that causes precede effects) is better understood as being a matter of perspective, where the perspective is determined by the kinds of beings who make use of causal relations in deliberation. There is no agent-independent fact about which temporal direction the cause-effect distinction goes. I argue that taking causal asymmetry to be a matter of perspective does a better job at relating all the different temporal asymmetries and better accounts for a number of thought experiments regarding our causal intuitions. The upside is that the insights provided by the statistical mechanical account of Albert and Loewer are still necessary. However, I argue that they aren’t enough on their own; a perspectival account is needed to make full use of their explanatory potential.