In this last post before August, I sketch a conceptual framework for estimating the future of science. The framework relies on the close connections between explaining the past and estimating the future.
My dissertation and the work surrounding it focused on the question of how to understand the development of science. I have argued that
1. Historiography of science is best understood as the study of how and why the present science came into existence.
2. Historiographical explanations exhibit patterns of counterfactual dependencies.
3. An explanation tells why some X rather than an interesting alternative Y is the case.
How do we get a framework for estimating the future from these claims? (I will summarize the idea in a picture at the end of the post.)
The first claim (1.) says that the historiography of science should make understandable the causal chains that led to the present science. This has two implications for the estimating of the future:
(I) Historiography of science is not (mainly) a description of sciences in the past. Rather, it is an account of how different processes affected the way science developed to where it is now. The relevance of this “ontological” definition of the history of science is that (i) it avoids conceptual debates about what counts as a science (especially in the past when things were differently), and (ii) it allows us to see how science depends on non-scientific practices and events as well as on its “internal” developments. We should first know what happened and why, and only then – if there is any need – discuss the appropriate ways to categorize the past.
(II) Historiography of science can provide a picture of causal chains that led to the present science. Once we are able to see the present science as a part of wider causal chains (and “trends”) we are in a better position to estimate how those causal chains might develop in the future. A detailed investigation of individual links in the causal chain is important and might provide some analogies for the future, but the explicit focus on causal chains enables us to see processes of different scales and analyze the role of individual events and processes in different scales.
The second claim (2.) is more important. It says that a historiographical explanation never explains X simpliciter but answers the question “Why X rather than Y?” where Y is some alternative to X that did not happen. Moreover, an explanation tells us when Y would have been the case. An explanation has the form “X rather than Y because Z rather than W”, where X is an actual event, Y an counterfactual alternative, Z an event in the actual past, and W an counterfactual alternative to Z. For example, that Einstein explained the Brownian motion (Z) explains why scientists came to believe in (the real existence of) atoms (X): had no one explained the Brownian motion (W), scientists would not have believed in atoms (Y).
There are two ways in which explanations with this structure are central to the estimating of the future.
First, in order to estimate where the present situation is going to lead, we have to approach our present situation in a similar way as we approach the counterfactual alternatives to the actual past. This is due to the fact that we do not have any evidence of the (actual) future developments. Both the actual future and counterfactual past are such that we do not have any direct evidence of them. We have to use our knowledge of the actual past and present in our attempts to understand the counterfactual pasts and the possible futures. If we are able to explain the development of science (and I think we often are), then we are able to assess the plausibility of counterfactual developments; explanation consist of tracking down patterns of counterfactual dependencies. If we are able to assess the plausibility of a counterfactual development and if counterfactual pasts and possible futures are epistemologically similar, then we are able to estimate the possible futures. Surely, there are differences between explaining the past and estimating the future (for example, we might not be able to formulate some possible scenario and therefore we might be, obviously, unable to track what developments could lead to that scenario) but the epistemology of tracking down developments that have not happened is the same in both cases. I have discussed the many uses of counterfactuals in historiography of science in Virmajoki 2019.
Secondly, once we accept that our decisions affect the future, we can think of different decisions as different scenarios. In this case, we can estimate possible futures by tracking down what would happen in those scenarios. We can think of different decisions in the same way as we think counterfactual alternatives to the past. Unlike in the previous case, we do not ask where the present situation could lead. Instead, we ask where different futures (i.e. futures where different decisions have just been made) could lead. The difference between the cases is not fundamental. In both cases, the similarities between possible futures and counterfactual pasts exist. The only difference is that, in the second case, we are not literally focusing on the present situation but on many possible situations in the near future where different decision are made. This might add some complications but hardly anything serious. For example, we have to assume that there are no fundamental changes between the present moment and the moment when the decision is made in order to make the estimating of the decisions and their consequences meaningful. However, such an assumption seems necessary for any decision-making procedure and is therefore not a serious theoretical postulate.
Everything, then, depends on how well we are able to track down counterfactual scenarios. I do not think there is a single answer to this question, everything depends on the nature and scope of the scenario (as I have indicated here). There are many different types of considerations that could be relevant. However, we are able to provide a meta-methodological suggestion on the basis of above considerations: Every method of studying the future, as long as it is a method aiming to investigate the (“objective”) possible futures, must be based on some conception of how the method bridges the gap between actual evidence and non-actual scenarios and these conceptions must be critically examined. The connection between explanation and future estimation is not merely one of many methods to use existing knowledge in the estimating of the future. Rather, it is a general framework within which the use of more specific methods is embedded.
For example, while it is not generally true (see discussion here) that “We should consider as plausible or probable only those alternatives which we can show on the basis of contemporary evidence that contemporaries actually considered” we could still find it important to study what future alternatives are considered as relevant now. The importance of people’s conceptions of plausible futures does not stem from the (obviously false) idea that these directly reveal what futures really are plausible but it could stem from the fact that these conceptions shape their actions in the future. If this were the case, then we would understand the importance of examining people’s conceptions in the wider scheme of tracking down scenarios. It is an empirical question in which contexts people’s conceptions of future enable to track possible futures.
There also some fruitful differences between counterfactual pasts and possible futures. The tracking of counterfactual developments is limited by sources that happened to be produced and survived, while in the tracking the future we can create sources of information (by interviewing people. for example) that we think are relevant. We can use the limitations in our ability track counterfactual histories to suggest what kind of information to gather when estimating the futures.
I will return to this meta-methodology many times in the future. I wanted to take up the theme briefly to highlight how the similarities between explanation and future estimation enable the building of a general framework for estimating the futures of science and how the function of different (first-order) methods can be understood through that framework.
The third claim (3.) says that in an explanation-seeking question “Why X rather than Y?” the alternative Y must be “interesting”. That we choose an interesting contrast to X is important in order to keep irrelevant explanations at bay. We could, for example, ask why there is gender inequality in science rather than no inequality at all – equal scientific communities would be interestingly different from the actual with respect to epistemological practices (assuming that inequality has a negative impact on such practices). However, it would be pointless to ask in this context why there exists a gender inequality rather than inequality between people from different backgrounds, since there would still be an inequality and probably as problematic epistemological practices as in the other case. There is no relevant difference, with respect to epistemological practices, between two types of inequality; there is such a difference between equality and inequality.
In order to find interesting alternatives to X, we have to know why X is significant in the first place. In the example of gender inequality (X), X is significant because it affects the epistemic quality of science. Here a principle of significance guides our judgement: “X is significant if it affects the epistemic quality of science”. Take another example. We may say that the observation of gravitational deflection is a significant feature of science because it plays an important part in the acceptance of relativistic physics (and thus in the overall shape of our physics) and in the understanding and technology that it provides. The principle would be: X is significant if it gives us understanding and shapes technology. Then a sociologist can point out that, since the contribution to the technological state of our society is significant about science, the origin of the funding in science is also a significant feature of science because it shapes the distribution and use of technology in a way that matters (specified by the sociologists). In this way, we need to use principles of significance as our guide in judgements of significance.
Once we know what principle of significance guide the choice of X, the choice of Y must respect the principle. The principles of significance do not tell us only what is, in fact, significant but also what could be significant. As I have argued elsewhere in detail (Virmajoki 2019), the method of reflective equilibrium can be used to formulate the principles of significance in the case of science. The idea behind the method is that we begin with the features of science that seem prima facie to be significant. Then we try to formulate principles that capture these features as significant. If it happens that these principles do not allow us to judge as significant some features that seem to be prima facie significant, we need to replace the principles or to bite the bullet and say that the significance of the features is illusory. The more the we study the science, the more features of it we recognize. These newly recognized features can seem to be prima facie significant and if the accepted principles do not lead to this judgment, we need to reconsider which principles and prima facie judgments of significance they accept. In this way, a balance between principles and judgements is sought for.
The method of reflective equilibrium is useful in estimating the desirable futures of science. We can ask which futures of science are desirable. Answering this question is made possible by analyzing significant features of science by the method of reflective equilibrium. The method enables us (i) to use our own values and needs as the starting point of the evaluation of the future values and needs, and still (ii) to evaluate how the future changes in nature, culture, society and technology might change the values and needs. For example, we might currently consider as significant that science provides simple explanations and find out that this judgement of significance is (partly) based on our limited cognitive abilities and our willingness to be able to effectively cope with phenomena. However, we could estimate that the processing power of computers increases and that the main problems of the future, related to climate for example, are extremely complex, and conclude that, in the future, simple explanations are no longer preferable to complex models that enable us to make accurate predictions. Here we have projected the willingness to cope effectively with important phenomena onto the future and also re-evaluated the preference for simple explanations in the light of other estimates of future possibilities.
As a summary of the framework, here is a picture. I do not see any reason to believe that such framework could not be used to estimate the possible futures in general (and not just the futures of science). However, we have to be careful as there might be some differences between different targets of estimates. For example, the method of reflective equilibrium might be too thin to generate robust enough value-judgements that are needed to stop the use of fossil fuels because too many things in our current lifestyle might be judged significant. Still, I think the framework could provide a basis of estimating the futures that could be modified for different purposes. Only the future can tell.
Figure 1. P1 is the present science and P2 is an interesting alternative to the present science. H1 is the set of events that led to the present science and H2 is a set of events that would have led to P2, had H2 been the case. F1, F2 and F3 are possible and desirable futures and dotted arrows are causal paths that would lead to those futures. Fs are identified by a process similar to the one that identifies P2. The causal paths from P1 to Fs are studied in a similar manner as the the P1-H1-P2-H2 structure.
P1-H1-P2-H2 structure was studied in Virmajoki (2019). The current project analyses how to extend that structure into the future in order to estimate possible futures of science.
 For the sake of simplicity, I often write the future even though the framework is built to estimate possible futures.