4.3 The compatibility of causation and physics
4.3.2 How interventions do not solve the problem
The temporal asymmetry of the facts will be central for the argument how caus- ation and the nomic bidirectionality of physics can be reconciled with the help of interventionism. Before presenting this argument in the next sections, I wish to briefly consider previous, and to my mind unsuccessful, attempts to base a similar argument on interventions. That causal actions or interventions bring with them certain asymmetries has already be noted by a number of authors. Interventions seem to be temporally asymmetric in that we can only intervene on the future. And they seem to be causally asymmetric in that there is a difference between doing something and the result of that action. It is therefore no surprise that it has been tried to ground the asymmetry of causation on the asymmetry of intervention and thereby solve the two prominent problems that have been presented in the last sections. In this section I will discuss these solutions and give reasons for why I believe they are unsuccessful.
I have already criticised the agency theory of causation by Menzies & Price in section 3.3.4 as having no implications for a theory of physical causation and will not repeat this critique here. However, in a newer version of the agency theory, Price & Weslake (2009) claim to solve the problem of the temporal asymmetry of causation and it is worth to have a quick look on their proposal. According to them, the temporal asymmetry of causation can be reduced to the temporal asymmetry of deliberation, viz., the fact that to advance one’s goals it only makes sense to deliberate towards the future, but not the past. Omitting many details, their account of deliberation can be summarised as a counterfactual theory of deliberation, that is, when deliberating what to do, we have to evaluate the truth values of certain counterfactuals. This evaluation in turn, as Price & Weslake explain, crucially depends on the condition ‘holding the past fixed’. That is, to evaluate the consequences of one’s action, one must assume that nothing in the past of the action changes. A desideratum of this condition is the perspective of the deliberator:
If the relevant species of counterfactual reasoning develops from the kind of hypothetical reasoning needed in epistemic deliberation, the principle that one should hold the past fixed provides a simple codification of the asymmetry of the deliberator’s perspective [...]. (Price & Weslake 2009, p. 437)
Furthermore, since the aim of Price & Weslake was to explain the temporal asymmetry of causation by reducing it to the asymmetry of deliberation, it must be concluded that ‘holding the past fixed’ is what provides the asymmetry of causation as well.
However, already Reichenbach (1956) has put forward a criticism of the attempt to explain the asymmetry of causation by the asymmetry of intervention, which I believe
is also applicable for the asymmetry of deliberation. According to Reichenbach, to explicate the meaning of a claim like ‘We can only intervene on the future but not the past’, it is necessary to make counterfactual claims about what would have happened, if we had not intervened. And counterfactuals of the form ‘If intervention I had not happened, event A would not have happened’ presuppose that the past remains unchanged, or else there could be another event B that causes A, making the counterfactual false. Reichenbach (1956, p. 45) concludes:
No wonder that acts of intervention change only the future, and do not change the past; the term ‘intervention’ is defined by the condition that the past be unchanged. The statement that acts of intervention cannot change the past is a trivial tautology. This consideration leads to the conclusion that acts of intervention cannot define a direction of time. The term ‘intervention’ is defined only after a direction of time is given; acts of intervention are actions that leave the past unchanged.
I content that this criticism also applies to Price & Weslake. Their reduction of the asymmetry of causation to the asymmetry of deliberation merely shifts the problem and does nothing to explain or ground the asymmetry of causation; the same questions that can be asked about causation can be asked about deliberation, for example, whether the fact that we only deliberate towards the future can be grounded in physics. It might be that these questions are easier to answer in the case of deliberation than in the case of causation, but Price & Weslake do not give any reasons to think so. The lesson of this discussion is that the simple fact that we can only deliberate or intervene towards the future does not provide a sufficient explanation for the temporal asymmetry of causation.32
Similar to what has been discussed above, Frisch (2010) makes an argument for temporally asymmetric causal relation in physics based on the fact that we can only intervene on the future. However, Frisch analyses the asymmetry of intervention as a particular asymmetry of state preparation in physics, thereby supposedly giving it more substance than the mere tautology has, which Reichenbach was criticising. Frisch’s argument is the following: 1. There is a temporal asymmetry of state preparation in physics. 2. The best explanation for this asymmetry is that it is an asymmetry of causal relations. 3. Therefore, by inference to the best explanation, there are asymmetric causal relations in physics. But what does Frisch mean by the asymmetry of state preparation? Here is how Frisch (2010, p. 81) explains the asymmetry:
Consider a system S that is governed by both past and future deterministic laws. That is, let us assume that the final state Si(tf) of the system is
32Another noteworthy account is found in Kutach (2011), who is not building up on any asymmetry
of intervention per se, but only on an asymmetry of ‘useful intervention’. I believe it is obvious that such pragmatic asymmetry does not help in clarifying issues about whether physical processes are causal. Furthermore, Kutach equates intervention (he uses the maybe more general term ‘influence’) with the temporally symmetric notion of nomic dependence, and acknowledges that as a consequence there are causal processes in both temporal directions. In effect, suggesting a temporally symmetric variety of causation sets Kutach close to Ney (2009), who I have already criticised in section 4.2.3.
uniquely determined by the initial state Si(ti), where ti< tf, together with the
dynamical laws and the boundary conditions; and that the initial state Si(ti)
is similarly determined by the final state Sf(tf). Thus, if S is closed between
ti and tf, , then the initial and final states are both dependent on each other.
Nevertheless there is an asymmetry of state preparation in the following sense. We can prepare the system in its initial state Si(ti)without making use of any
knowledge we might have of the system’s dynamical evolution between ti and
tf; and we can subsequently calculate the system’s future evolution for times
t > ti from the initial state, the dynamical laws, and the boundary conditions.
But we could not similarly first prepare the system’s final state at tf without
using our knowledge of the dynamics and then take the final state together with the laws to calculate the system’s past evolution for t < tf. (Of course we
cannot first prepare the system in Sf and then let it evolve into Si. That is not
what the asymmetry consists in. Rather the asymmetry consists in the fact that we cannot first prepare the system in Sf without making use of facts about
the dynamical evolution and then calculate what the system’s past evolution from Si to Sf must have been, given the dynamical laws and the boundary
conditions.)
What Frisch has in mind seems to be the following difference between the initial and the final state. Under the assumption that the system is closed between ti and tf
we can prepare the initial state by intervening on it directly at ti, but we can only
prepare the final state by first preparing the initial state and then letting it evolve into the final state. If we were to prepare the final state directly by intervening on it at tf then the system would not be closed anymore, which contradicts the assumption.
This difference in how we can prepare the initial and final states provides us with a temporal asymmetry and therefore with a difference between past and future, according to Frisch.
Frisch then continues to argue that state preparation is a kind of intervention that falls under the interventionist account of causation by Woodward (2003):
In particular, if Sf is an effect of Si, then according to an interventionist
account of causation there are two ways by which one can intervene on the system to set Sf to a particular value: first, we can intervene on Si, which in
turn will affect the value of Sf; or, second, we can intervene directly on Sf [...].
(Frisch 2010, p. 82)
To identify state preparations with causal interventions then enables Frisch to explain the asymmetry of state preparation with the asymmetry of causation. Finally, to complete his argument, by inference to the best explanation there must be causal relations in physics.
I do not agree with Frisch’s argument and have a couple of reasons why. First, one might wonder whether causation really is the best explanation for the asymmetry of state preparation. According to Frisch (2010, p. 83) “there is no other fully worked out and equally as successful non-causal alternative explanation of the asymmetry.” But what about a theory of time according to which time flows from past to future and only the present exists? This would do the same job by giving a reason why we cannot intervene on the final state directly without violating the closed system
constraint. Granted, there might be other reasons for not believing into a flow of time, but there might just be the same reasons why many refuse to believe in time asymmetric causation. In any case, one would wish to have more reasons from Frisch to accept the explanatory relevance of causation. Second, Frisch does not make appropriate use of Woodward’s interventionist theory. As can be seen from the quote in the previous paragraph, according to Frisch, if there is a causal relation between Si and Sf, then an intervention on Sf can either go via Si or be on Sf directly.
However, according to definition IV.I3 in Woodward (2003, p. 98) a well defined intervention cannot act on Sf on a path that does not go through Si (cf. section
3.3.4 above). This might seem like a nitpicking point, but it is crucial to Frisch’s account that there is an asymmetry between certain ways of intervention that can be identified as causal interventions such that then the asymmetry can be identified as a causal asymmetry.
Most damaging to Frisch’s account however is that upon closer examination the asymmetry of state preparation is nothing else than the temporal asymmetry that we can only intervene on the future but not the past. Which in turn makes Frisch vulnerable to Reichenbach’s point that this statement is a mere tautology and therefore does not prove anything. Again, the asymmetry of state preparation according to Frisch consists in that we can intervene on Si directly to prepare it,
but not on Sf, given that the system is closed for times between Si and Sf. The
problem with that is that it seems to presuppose that time is directed from ti to
tf. Because if this were not the case, one could turn Frisch’s example around and
look at it from the other temporal direction, in which case we could intervene on Sf directly, but not on Si. Frisch does not give any reason why only one temporal
direction should be allowed. Being charitable, at this point one could make use of the temporal asymmetry of the facts, for which I have argued in the previous section, in which case only one temporal direction is the case. Nevertheless, even then Frisch’s account does not solve the problem of nomic bidirectionality. Even if there is an asymmetry in preparation in that we can prepare Si directly, but Sf
only by first preparing Si and letting it evolve into Sf, the problem remains that
Si and Sf nomically depend on each other. Hence, Frisch’ reasoning might solve
the problem of temporal symmetry, but not the problem of bi-directional nomic dependence. I will therefore not follow Frisch, but propose a different account for the asymmetry of causation, based on a localised understanding of determinism and interventions, in the next section.