Losing My Religion. The Problem of Unconceived Alternatives

In the previous post, I wondered whether we can conceive a history (or future) of science that does not end up in the current state of science. We noted that the task is at least difficult. We have to use our current scientific knowledge in creating counterfactual historical scenarios of science and therefore our current scientific knowledge “pulls” the scenarios towards the actual state.

Prima facie, this use of scientific knowledge in creating the counterfactual scenarios is not a problem. We have empirical evidence for our best scientific theories and they are the best source of knowledge to be used in the scenarios-work. If scientific knowledge is used in a scenario, we should trust the scenarios.

There still remains a feeling of uneasiness. By using scientific knowledge in historical scenarios, we conclude that we could not have a science that differs vastly from the current science. But it seems possible that we could have an alternative science that is equally well supported by evidence. Given the inferential distance between theories and empirical evidence, it seems that we could have many (at least two) equally well-supported theories at any given time. The available evidence does not determine our theory choice. This is known as the problem of underdetermination of scientific theory.

While this problem is an old one and has produced valuable works on the logic of scientific confirmation, a standard argument against it is that well-supported alternatives are extremely rare and therefore the possibility of such alternatives is nothing but an age-old skeptical hypothesis. Maybe there are alternatives to our scientific theories; or maybe an evil demon deceives us! It is difficult to judge whether such skeptical possibilities should affect our attitude towards the current science.

However, P. Kyle Stanford has put some historical flesh on the logical bones of underdetermination. In his book, Exceeding Our Grasp Science. History, and the Problem of Unconceived Alternatives (2006), Stanford argues that the problem of underdetermination is not solved by noting that empirically well-supported alternatives have not been presented. The problem remains if we have reasons to believe that there exist well-supported alternatives to our best scientific theories that are presently unconceived by us. This is the problem of unconceived alternatives. Remember that mere possibility of such alternatives is not what we are looking for. We are looking for reasons to believe that such alternatives exist.

Stanford argues that there is a reason to believe that such alternatives exist. On the basis of the history of science, he performs what he calls the New Induction[1]:

“By contrast, I propose what I will call the new induction over the history of science: that we have, throughout the history of scientific inquiry and in virtually every scientific field, repeatedly occupied an epistemic position in which we could conceive of only one or a few theories that were well confirmed by the available evidence, while subsequent inquiry would routinely (if not invariably) reveal further, radically distinct alternatives as well confirmed by the previously available evidence as those we were inclined to accept on the strength of that evidence. For example, in the historical progression from Aristotelian to Cartesian to Newtonian to contemporary mechanical theories, the evidence available at the time each earlier theory was accepted offered equally strong support to each of the (then-unimagined) later alternatives” (2006, 19.)

Given that Stanford is correct, what does this mean for our ability to understand science and especially the futures of science?

(Methodological interlude. In the spirit of the project of attempting to formulate how the development of science from the history to the future can be understood, I do not directly criticize a philosophical position (without very good reasons). Rather, I attempt to explicate their structure, assumptions and consequences for our understanding of science. I ask “What if this view is correct?” and follow where the view takes us. In this way, we are able to understand how different views on the nature of science affect our thinking about the history and future of science. I think it is better to approach the estimating of the futures of science from a pluralistic point of view whenever meaningful (obviously rubbish theories do not deserve the status of “view of on science”) than to make substantial commitments early on. However, explication of the structure, assumptions and consequences of a philosophical view is not epistemically futile: The structure, assumptions or consequences might turn out to be contradictory or absurd and therefore provide a reason to reject the view. In other words, I do not conceive my task as judging what view of science is correct but to point towards the vast possibilities that philosophical reflections on science open to the estimating of the future of science.)

First of all, because the argument is not based on mere skeptical fantasies but on the historical record, the evidence supporting the argument is (explicitly) fallible. This means that the argument itself and its consequences are subject to the problems associated with our historical thinking (discussed in a previous post). While it seems undeniable fact that there have been unconceived alternatives (for example, one has to write extremely distorted historiography if one is willing to argue that Newton could have conceived the general theory of relativity), the implications of the theory can be debated on historical grounds. For example, Stathis Psillos argues that historical record shows enough theoretical continuity to not challenge scientific realism (Psillos 2009, 4.2).

A related issue is whether the past science is a good base for the induction. As Psillos (2009, 4.3) points out, “one could argue that as science grows, theories acquire some stable characteristics (they become more precise; the evidence for them is richer and varied; they are more severely tested; they are incorporated into larger theoretical schemes and others) such that (a) they can no longer be grouped together with older theories that were much cruder or underdeveloped to form a uniform inductive basis for pessimism and (b) they constrain the space of alternative possibilities well enough to question the extent of the unconceived alternatives predicament.” The point here is that since (i) science is different today than it was in the past, (ii) most of the science has been produced in recent decades, and (iii) the science has been quite stable recently, it seems that the historical challenges do not apply to the current state. Again, we must remember that argument of unconceived alternatives gains its power from historical considerations. Even if (i)-(iii) do not exclude the logical possibility of unconceived alternatives, they could cut the link between the historical record and current science.

I think that even if the supposed link to the history breaks, the idea of that we have exhausted the space of possibilities is somewhat suspicious. In the field of technology – consider artificial intelligence for example – such argument would be brave. Now, given the intimate relationship between science and technology and given the amount of unconceived technologies, it seems that we should take the possibility of unconceived alternatives seriously. Surely, this is not the same argument as Stanford original induction but is based on essentially similar historically inspired humbleness.

Recently, Stanford has argued, in his paper “Unconceived alternatives and conservatism in science: the impact of professionalization, peer-review, and Big Science” (2019), that “we have compelling reasons to believe that they are actually less effective than those same predecessors in conceiving, exploring, or developing fundamentally novel theoretical conceptions of nature in the first place”. The funding structure, professional specialization and self-identity, and the social organization of Big Science have led to a situation where opening research paths that challenge the current theoretical “orthodoxy” is extremely risky and difficult.

This dynamic can be pushed even further. D. P. Rowbottom (2016) has analyzed different aspects of science that can have unconceived alternatives. The title “Extending the argument from unconceived alternatives: observations, models, predictions, explanations, methods, instruments, experiments, and values” is a rather good summary of those aspects. For example, Rowbottom points out that scientific theories often lack appropriate predictive force. Models need to be developed in order to achieve predictions. Sometimes the adequacy of a scientific theory to deal with phenomena remains open question until sophisticated models are developed. This was the case in classical mechanisms for over a hundred years. Sophisticated models can be initially unconceivable and, given that models are responsible for predictions, there can be unconceived predictions. Given the importance of predictions in theory-evaluation, the unconceived models can have impact on our theory choices.

While it is difficult to tell how much weight we should give to unconceived alternatives in different aspects of science, Rowbottom’s analysis serves as a valuable reminder that the high-level theories are not the only aspect where unconceived alternatives are an issue. The point that “What’s conceived is nonetheless limited, for a variety of reasons; limitations on time and material resources, contingencies about where attention is directed, and so forth” (Rowbottom 2016) is a good reminder of our historical situation. However, the fact that unconceived alternatives exist in many levels of science does not automatically have any novel consequences. No one denies that science develops and science changes. Moreover, I guess only few would say that all the future changes can be conceived now. The question is how fundamental these changes can be.

Initially, however, Stanford suggested that unconceivable alternatives are a problem in the fundamental domains of science, theoretical science. Stanley argues that eliminative inferences where conclusions are reached by ruling out alternative possibilities until only one remains work in cases where we are able to conceive the plausible alternative possibilities. However, the New Induction indicates that scientists have been unable to conceive the plausible theoretical possibilities and therefore the eliminative inferences have not worked in theoretical science (or they have been missing) (2006, 30-31.) The idea that theoretical science cannot be trusted as “really true” is important in its own right, but one may wonder whether we can really separate different levels or aspects of science form each other (see Stanford 2006, ch. 8, and Psillos 2009, ch. 4, for competing views). Prima facie, theoretical changes could lead us to rethink our eliminative inferences (or inferences in general) in other areas of science as well. On the other hand, we have seen that unconceived alternatives can be found below the highly theoretical level. It seems that if an unconceived theoretical change really is possible, then this possibility must be based on new, unexpected, findings in other aspects of science than the theoretical one. In order to rethink our theories, new predictions, methods, or values need to be found. So one could argue that if there are plausible unconceived alternatives in the theoretical level, there must be such alternatives in other levels of science as well. Theories and other aspects of science are so deeply intertwined that we perhaps should expect that the possible changes in each of the aspects are of similar magnitude.

Where does this take us with respect to the future of science and futures in general?

First obvious lesson is that an argument can be made that the future of science may be different in a now-unconceived way. At any given time, there have been changes that were not conceived earlier. The original argument was that these changes are fundamental, i.e. that is plausible that we have not conceived a theoretical science that is fundamentally different but equally well supported than our current theoretical science. The argument has been criticized but it is not flawed in any obvious way. In my view, the argument provides us with a history-anchored reason not to exclude different science as a mere skeptical fantasy.

Secondly and more importantly, the arguments for and against the problem of unconceived alternatives reveal important issues that need to be (and have been) addressed in order to understand the development of science. One of the issues is that of theoretical continuity. How should we define and confirm theoretical continuity through theory-change? Psillos (1999, 2009) has argued that there have not been as many fundamental changes in science as is sometimes suggested. The parts of the past successful theories that were responsible for the success of the theories can be shown to be retained through theory-change. This called the divide et impera strategy for scientific realism. Even if we accept this strategy, we still need to investigate whether the judgements of continuity can only be made ex post facto, i.e. whether the continuity can be underlined (in practice) only within the framework of the latter theory. If so, the future theoretical continuities may be incomprehensible from our perspective and therefore set serious limits to our ability to estimate the future of science even in the fortunate case that theoretical continuity exists (or in terms of taxonomies: whether the theoretical space of the future can be estimated even in those branches of our theoretical-structural taxonomies that allow theoretical continuity). Another issue is that if we adopt the divide et impera strategy, we can perhaps make the inductive inference that, since the theories of the past involved parts (or dimensions) that were responsible for their success as well as redundant parts, current theories probably involve redundant parts. And given the inability of past science to distinguish between the two areas of theory – successful and redundant – we are probably unable to do so until some serious theoretical turning point is achieved.

I think that one of the most interesting dimensions of the debate is Stanford’s argument that even though science has changed and therefore the inductive base can be questioned, we have independent evidence that the current structures in science may hinder the search for unconceived alternatives. In this type of reasoning, we identify a possible problem for science (or any other aspect of life, for that matter) by studying its history and then attempt to search for conditions in the present that could make the problem acute even if the present conditions differ from those of the past. So A was a problem in the past when B was the case. Now B is not the case anymore but C is, and C may lead to A. (For example, A = plausibility of unconceived alternatives; B = only few scientists, C = conservative incentives in science.) This type of reasoning could be fruitful in other projects as well.

Thirdly, Rowbottom’s analysis of unconceived alternatives in different aspects of science highlights the necessity of studying the development of different aspects of science together. Even though it might seem that some aspects of science – for example, observations or theoretical values – do not have threatening unconceived alternatives, this is probably is not the case. As I argued above, if there exists a problem of unconceived alternative theories, there probably also exist all sorts of interesting unconceived alternatives in different aspects of science. Many aspects of science may appear not to contain important alternatives. Sometimes we think that an aspect of science, like observations, is not controversial enough to make the problem of unconceived alternatives fruitful. In other cases, we think that some aspects, like values, are too controversial or unclear (consider the many dimensions of simplicity) to make the problem of unconceived alternatives important. In the controversial cases, we do not have a neat picture even of the actual alternatives at play and so there is no point of adding unconceived alternatives into the mix. However, Rowbottom’s approach indicates that such attitudes might be misleading, and we should attempt to understand the possibilities of unconceived alternatives in different aspects of science. Whereas the original problem formulated by Stanford focuses on the theoretical level of science and draws a conclusion concerning that aspect of science in general (a tendency that characterized the realism debate for decades), similar general analyses of the issue of unconceived alternatives in other aspects/levels of science is not in sight. Rather, the implications of the problem of unconceived alternatives in other levels of science require local analysis and depend on the particular contexts of research. To give an obvious example, unconceived alternatives in historiography and science probably differ greatly (although there has been some criticism towards the basic tenets of historiography, see Rosenberg 2018 on recent attack).

Finally, the philosophical debate on unconceived alternatives can have implications for futures research. However, the implications are not straightforward. The problem of unconceived alternatives in science threatens our trust towards current scientific theories. The problem is therefore about our epistemological underpinnings. It is a problem of “how do we know”. In contrast, if there exists a problem of unconceived alternative in some other field, it is difficult to say what that problem concerns in addition to our ability to know the future or possible alternatives to the current world. The world is what it is and unconceived alternatives do not challenge it. Even if Yesterday could have been performed differently, this does not have the implication that perhaps the actual Yesterday is erroneous (or a wrong version). Such questions would be nonsensical.

However, the problems that unconceived alternatives pose to futures research are not trivial. Surely, all the problems of estimating the future go back to our inability to conceive events and processes that are novelties and affect the course of the future. Despite this – or even for this very reason – it would be important to analyze the logic and causes of different types of unconceived alternatives. Probably not all inconceivability has similar causes and consequences. (I separate logic and causes because we can say, in the case of unconceived alternatives in science, that the logic of the problem is that there have not been sufficient eliminative inferences in theoretical science, and this logical problem is caused by things like the cognitive limitations and incentive structure of science.) For example, we could ask whether there are unconceived alternatives for the society and how this affects the futures research. We could wonder whether political tensions cause a fragmentation of the space of future possibilities. Is it possible that we miss some possible ways of organizing the society because only politically clear visions of the future (i.e. futures that fit some currently existing political agenda) get their voice heard in the atmosphere of increasing political tensions? We could add that the logical problem in the politically fragmented futures is that if we cannot conceive a future outside the ends of political (n-dimensional) spectrum, it becomes difficult to answer normative questions on how people should live in a society. If there were unconceived alternatives that would have been better for many people, can we expect them to conform to the rules of society? In this way, we can extend the thinking through unconceivable alternatives beyond science.


Psillos, Stathis (1999). Scientific Realism: How Science Tracks Truth. Routledge.

Psillos, Stathis (2009). Knowing the Structure of Nature: Essays on Realism and Explanation. Palgrave Macmillan.

Rosenberg, Alexander (2018). How History Gets Things Wrong: The Neuroscience of Our Addiction to Stories.

Rowbottom, Darrell (2016). “Extending the Argument from Unconceived Alternatives: Observations, Models, Predictions, Explanations, Methods, Instruments, Experiments, and Values”. Synthese (10).

Stanford, P.K. (2006). Exceeding Our Grasp: Science, History, and the Problem of Unconceived Alternatives. New York: Oxford University Press.

Stanford, P. (2019). Unconceived alternatives and conservatism in science: the impact of professionalization, peer-review, and Big Science. Synthese 196 (10):3915-3932.

[1] Old one being the pessimistic metainduction: There have been false successful theories. Therefore our successful theories are probably false.

Leave a Reply

Your email address will not be published. Required fields are marked *