Put Your Money Where Your Mouth Is (PYMOWYM) principle: All research on historical phenomena both assumes and makes conclusions about the workings of the relevant phenomena. The commitments limit and shape the possible structures of the phenomena in the future.
In the previous post, I discussed Lakatos’s idea of framework-driven historiography. According to Lakatos, historical facts need to be ordered in terms of a framework. I mentioned that it is unclear how a mere sequence of actual history can be explanatory even if it appears to make sense because Lakatos did not inject any modal considerations into his account. In this post, I show that there is a straightforward way of connecting the so-called different-making account of explanation with Lakatos’s insights. But let us go back to our original problem.
How can we know what would have happened?
Currently, the interventionist account of explanation is a prominent version of the counterfactual account of explanation. In this account, X explains Y if and only if it is true that had X been different, Y would have been different, where X is changed by an intervention and certain other variables are held fixed (Woodward 2003). The interventionist account differs from Lewis’s in that it is not reductive. Intereventionism requires causal knowledge in its analysis of causal and explanatory claims. Which pattern of counterfactual dependencies between variables establishes the explanatory dependency between X and Y depends on the type of system they are embedded in, and thus we need causal models to track down these dependencies. For example, in normal circumstances kicking makes a ball roll. Had I not kicked, the ball would not have rolled. However, in other circumstances, there might be a backup person who kicks the ball if I do not. It is not true that had I not kicked, the ball would not have rolled. Yet, it is true that had I not kicked and had we fixed the backup so that she does not kick, the ball would not have rolled. There is a causal and explanatory relation between the kick and the rolling in both cases. but it is grounded in different patterns of counterfactual dependency.
It is important to note that the main function of the interventionist account is not to find out when some X explains Y – even though it can be used in this way – but to define how Y can be explained. This might seem like a small difference but it is not. To tell that some X is explanatorily relevant to Y amounts to a rather minimal explanation, as it does not specify anything else that is relevant to Y. On the other hand, if we focus on Y, we look for a pattern of counterfactual dependence between many factors and Y. An explanation answers what-if-things-had-been-different questions, and the more detailed the pattern of counterfactual dependence we find, the better our explanation is. This is why we should go straight to search for patterns of counterfactual dependency rather than ponder about a particular causal relationship between some X and Y.
Let’s go back to historiography. If we wish to explain why some historical event or process took place, we need to tell when that event or process would have been different (or absent). In order to do this, we need to know what types of factors were involved in the situation and how they affect each other. The question is: How can we know this much in historiography?
To be sure, the answer to this question depends on what we are trying to explain. For example, when we try to explain why Millikan achieved his result in the oil-drop experiment, we can cite the physical working of the instrument, the properties of electrons, and so on. This is a relatively straightforward case, and we can use scientific knowledge to construct the explanation. We could also ask why the gravitational deflection of light was observed in 1919 and cite the fact that Eddington was a Quaker which provided the motivational background. Things get a little more complicated here because it is difficult to justify claims about the relationship between particular details in the motivational background and the series of actions. The problem is not how to explain actions in general; that is a philosophical debate on a completely different level. The problem is simply that it is difficult to say whether Eddington would have had the motivation to unify the international scientific community even without explicit Quaker beliefs. However, our explanatory task is helped by the fact that we know a lot about the situation in 1919 around testing Einstein’s theory: We know that the theory was interesting but not many people were willing to go as far as Eddington to test it. We also know how Eddington’s attempt was criticized in advance due to possible problems in the observational procedure and how the criticism was treated. So we have a mixture of detailed causal background knowledge of the situation and open questions concerning the exact relationship between particular factors. In order to provide a comprehensive explanation, we need to fill the gaps. However, in order to explain the observation in 1919, we do not need to mention every possible explanatory factor. Some aspects of the situation can be left uncertain. We might know that the observation was made because Eddington was a Quaker rather than had lost all hope to humanity in war even if we cannot say whether the observation would have been made, had Eddington adopted a different religious identity during his life. I will come back to these uncertainties below. Here I just want to indicate that such uncertainties might reveal important blind-spots in our worldviews. And, as we know, our mind fills the blind-spots, and they are observable only in specific circumstances.
When we reach a more macrohistorical level where we abstract away the details from our explanation-seeking questions, providing explanations becomes even more difficult. For example, if we ask why the gravitational deflection was observed instead of asking why it was observed in 1919, many details of the causal structure of the latter case probably become irrelevant. For example, one might think that even if Eddington was not present, someone else would have made the observation sooner or later. Or one might wonder what would have happened, had the General Theory of Relativity not been formulated back then or if it had been dismissed due to the political tensions of the Great War. How is it possible to answer such questions?
Frameworks as backbones of scenarios
As we have seen, in the difference-making account X explains Y if there is a pattern of counterfactual dependence between X and Y. It is possible to understand Lakatosian frameworks as descriptions of the workings of the relevant domain. When an actual historical sequence “X, Z, then Y” is deemed as relevant to the outcome of Y, there are implicit patterns of counterfactual dependence assumed in the judgement. Notice that Lakatosian frameworks do not simply list the items in the sequence but also interpret them. For example, falsificationist framework does not provide a neutral description “theory T was formulated; experiment E was performed; scientists believe B” but a loaded one “T was formulated; test E for T was performed; scientists came to believe that T has serious merits”. The sequence makes sense because of the way it is described. I claim that this is due to the fact that the loaded description carries with it hints of dependence-relations. A framework tells us what factors were relevant and how they depended on each other. For example, falsificationism tells us that the outcome depended on the formulation of T and on the fact that it was tested and that the result of the test; had the theory not been formulated in enough detail to be tested, had there not been a test, or had the test produced a different result, scientist would not have believed that the theory is a good candidate. The falsificationist framework makes history understandable by suggesting patterns of counterfactual dependency.
Difficulties begin when we focus on historical details that do not match the expectations of a framework. In our example, these are cases where tests provided a result against a theory, but the theory was not rejected. The obvious reaction to these cases is that they are evidence against the account. They show that the world does not work in the way the framework describes. We simply cannot trust our judgements “had the test been different, the theory would not have been accepted” because in many cases a problematic result does not affect the acceptance. However, the issue is much more complex. As we noted above, what patterns of counterfactual dependence are relevant for an explanatory relation between X and Y depends on the causal structure of the situation. A framework that claims that attempts to refute a theory explain the acceptance/rejection of a theory does not have to commit to the truth of “had the result of the test been different, the acceptance would have been different” in every situation. For example, if a rumor is spread that the test was performed by cheaters, then it is natural that the acceptance did not depend on the test in this case. Lakatos famously argued that we should separate internal and external histories of science. Internal history accords to a framework. External history explains why certain cases do not accord with the framework: “when history differs from its rational reconstruction, [external history] provides an empirical explanation of why it differs”. This distinction is a rather natural one in the light of the considerations above. The breaking down of internal history might be due to the causal structure of the situation that can be mapped with further historiography.
A mirror-image of the case above (where the relevant factors exist but the outcome is wrong, according to a framework) is a case where the outcome is correct (according to a framework) but, in addition to factors that should explain the outcome (according to the framework), there are other factors that did affect the outcome. Again, there is no deadly problem. The problems these cases pose depend on the causal structure of the situation at hand. For example, X might be relevant to Y even if some Z pre-empted the influence of X. A more problematic example would be one where the alleged cause is present but the outcome would not happen without last-minute intervention from elsewhere. For example, imagine there is a result of a test that does not match the prediction of a theory. Despite this, scientists continue to accept the theory until a famous figure says on a podcast that the theory is idiotic. In this case, the acceptance does not depend on the test. However, even in this case one might search for external factors that made the test irrelevant in the eyes of the community (see the previous paragraph).
We can see that a framework is not easily refuted by seemingly anomalous cases. There are always ways of finding the error from elsewhere. However, this does not mean that we can accept whatever theory of the face of whatever cases. Rather, this only means that there are no crucial tests for frameworks – something we know is true of all theory-testing. Constant failures and constant inability to provide correct historical verdicts surely speak against a framework even though even this is highly dependent on the context and background knowledge. As Lakatos himself pointed out “the theory of scientific rationality [i.e. a framework] progresses if it constitutes a ‘progressive’ historiographical research programme. [Framework-driven historiographies] remain for ever submerged in an ocean of anomalies. These anomalies will eventually have to be explained either by some better rational reconstruction or by some ‘external’ empirical theory.” (1979, 118.)
Now, one might think that the crucial test for a modally loaded framework is that it provides correct judgements in the form “had X been the case, Y would have been the case”, i.e. that a framework is acceptable only if and only if it implies that Y depends on X and Y really does depend on X. However, this way of thinking is not possible. Remember that we introduced frameworks as something that guides explanations. We noticed that we seem to be able (or at least willing to) identify explanatory sequences and relations from the vast sea of history. This requires background assumptions of how the world works, and the frameworks were introduced as something that describe the workings of the world. We cannot test frameworks against counterfactual scenarios. Rather, we build the counterfactual scenarios in terms of frameworks.
We arrive at the following picture. In order to judge that X explains Y, we have to (a) know what factors were involved in the situation (b) know the dependencies between these other factors and X and Y, and (c) adopt a framework that describes the relationship between X and Y. It is of utmost importance to notice that frameworks are needed to approach relatively tricky counterfactual scenarios that are common in history rather than more mundane ones. In such cases, a framework does not provide straightforward answers in itself but only in relation to background knowledge concerning the causal structure of the situation. The framework provides a backbone for the construction of explanatory counterfactual scenarios.
Let’s take an example. Why was the gravitational deflection observed in 1919? Was it due to Eddington’s effort? We know many things about the situation: The war had damaged the international scientific community. Einstein’s theory was introduced and it was a distinctively interesting one. Eddington had a specific background that made him willing to reunite the community and able to test Einstein’s theory. Against this background, we can judge that Eddington’s effort explains the observations in 1919: There seems to have been no one else that was willing to execute an expedition to observe the gravitational deflection. Moreover, while it is usually true that distinctively interesting theories are tested, this factor was cancelled by the damaged relations between British and German scientists. This first example is not all that tricky. Surely, it assumes background knowledge of how things like politics shape the course of human actions but the scenario without Eddington is a trackable one on the basis of rather mundane knowledge of how the world works. Of course, one can question such knowledge and claim that, had Eddington been removed, aliens would have sent a copy of him to the Earth, but this is beside the point. We are attempting to understand historiographical explanation, not fighting skepticism. All explanations require knowledge of how the world works, and in this particular case the knowledge required by historical explanation seems rather mundane.
Let’s then ask why the deflection was observed (no matter when). In other words, we need to ask what should have happened so that the deflection was not observed at all. In this case, it is more difficult to find explanatory factors. For example, the acute influence of the war would probably have diminished but the interest in Einstein’s theory probably would not. Moreover, this seems to imply that Eddington was dispensable. Someone else could have been eager to make the observations. But how can we arrive at judgements like these? It is here that the frameworks become a necessary tool. We need to make assumptions of the types of factors that affect the development of science in general. For example, falsiciationism says that science proceeds in terms of novel theories with rich contents and attempts to refute those theories. Had Eddington not existed, there would still have been a novel and rich theory. If we remove Eddington from a scenario and leave the theory there, the scenario leads to the testing of the theory as it could have refuted the theory. Whether something is observed depends heavily on the situation on theoretical parts of science, according to falsificationism.
However, we can add complexity to the example. The complexity reveals why frameworks need to be thought of as the backbone of a counterfactual scenario. It is well known that Eddington (and Dyson) were selective in the data he selected for his analysis of the deflection. It has been claimed that the data can be reconstructed to yield different results concerning the deflection. If this is the case, then falsificationism leads to a different development of the scenario: If we remove Eddington, the observation would not have been made because his idiosyncratic data-selection would not have been in play. The case seems to go against the falsificationist framework but, as we have seen, there are ways to reinterpret the causal structure of the situation in such a way that framework is saved. Falsificationism would say that (i) given the interest in the theory, there would have been an attempt to refute it, (ii) given that the data, the observation of the deflection could not have been established, but (iii) given Eddington’s idiosyncratic judgement, the observation was still announced. We have a pattern of counterfactual dependency that explains the observation. The pattern would not be tractable without a framework and therefore the framework provides a necessary backbone for the explanatory scenario.
What falsificationism says about the scenario might sound trivial because it corresponds to a straightforward picture about science. The point of using frameworks is not to justify some intuitive or naïve explanations of science. Notice that different frameworks often lead to different judgments. For example, Lakatos argued that science does not abandon easily the theories that form its “hard core” in the face of evidence. Rather, the “protective belt”, a set of auxiliary assumptions, is modified. What matters is the overall progress of the scientific research program over time. This is the “research programme framework”. Given this framework, Eddington’s idiosyncratic interpretation of data might not have affected the outcome of the process. Even if someone else would not have concluded the observation as made on the basis of similar data, there would have been further attempts to observe it. Scientific research programs are resilient in the face of contradictory evidence. But it is also true, in the research programme framework, mere formulation of Einstein’s theory would not have been consequential to the importance of the observation of gravitational deflection. Had people not been aware of problems that Newtonian physics had difficulties solving, an alternative research program would not have been as attractive as it was. In the research programme framework, the central explanatory patterns of counterfactual dependencies hold between the overall state of a research field, theoretical novelties, tests, and protective modifications. These factors answer what-if-things-had-been-different questions and thus explain. But again, they do not provide answers without background knowledge of the causal structure. Rather, they form the backbone of the explanatory scenario where the structure is taken into account.
Finally, one could adopt a strongly sociological framework according to which theoretical developments and tests are epiphenomena of social dynamics of science. What scientist do and accept does not depend on theoretical or empirical considerations but on social incentives. When one tracks what would have happened without Einsteinian theory or Eddington, one constructs the counterfactual scenario in terms of social factors and generalizations that govern them. This example shows how different the frameworks can be from each other. Framework-driven explanations might therefore differ greatly from each other. There is no guarantee that frameworks provide convergence in historiographical judgements. On the contrary, they are a central source of disagreement, not least because it is difficult to refute a framework (see above).
To sum up, in order to explain a historical outcome, one needs to answer what-if-things-had-been-questions and find out when the outcome would have been different. In order to answer these questions, one has to construct counterfactual scenarios. The construction of the scenarios depends on assumptions about how the world works. In many important but tricky cases, it is not enough that we introduce common-sensical assumptions or low-level empirical regularities. We have to make much wider assumptions about the workings of the world. An interconnected set of such assumptions is a framework. This means that many important historical issues can be explained only by relying on frameworks. A sequence of historical events is explanatory if and only if there are counterfactual dependencies between the events. A framework provides the conceptual tools to identify a possibly explanatory sequence (see 1-dimensional PYMOWYM ) and also explanatory tools to provide the actual explanation in terms of counterfactual dependencies.
We arrive at 2-dimensional (modality, no axiology) PYMOWYM principle: In order to explain important historical cases, we need to rely on assumptions of how the world works. These assumptions are provided by frameworks. These frameworks provide the backbone of counterfactual scenarios that enable us to answer what-if-things-had-been-different questions and thus explain. Thus, explanations in historiography are committed to frameworks. A framework generates a set of possible futures, as it tells how the world works. Thus, every explanation in historiography generates a set of possible futures, and a commitment to an explanation is a commitment to a set possible futures.