Problem-Feeding as a Model for Interdisciplinary Research

ABSTRACT

Philosophers of science have in recent years become increasingly interested in the notion of interdisciplinarity. One important form interdisciplinarity can take is that of a dynamic exchange of problems and solutions between disciplines—what has recently been called problem-feeding. On this model problems arising within specific disciplines are sometimes solved more effectively by, or in collaboration with, other disciplines. In this paper we explore this model as a framework for thinking about, and actively structuring, interdisciplinary research. We point to the applicability of the problem-feeding model, and to some of the prerequisites of problem-feeding interdisciplinarity, highlighting in particular the harmonisation of goals and the establishment of mutual trust between disciplines.

1. Introduction

Sometimes problems are encountered in a discipline that cannot be adequately solved within that discipline. For example, when climate scientists run models of the climate system to project changes in the earth’s climate over the next fifty or a hundred years, they need to have an idea of what the future emissions of CO₂ and other greenhouse gases will be. Clearly, those emissions will depend both on how the various bio-geophysical systems respond to a warmer and more volatile climate and on what social, technological and economic developments we will experience in the coming century. Although climate scientists could make some more or less well-educated guesses, these issues really lie in other disciplines, such as (climate) economics. So, instead of trying to come up with some way of tackling the issues on their own, each problem is outsourced, or exported, to a different discipline—one that deploys a different set of tools, theories, concepts and models. In the particular case of climate change, the key tools are called integrated assessment models (IAMs). These couple representations of economic systems and sectors with various bio-geophysical models, often simple climate models. Climate economists then use the models to figure out what the emissions will be under various assumptions about, for example, the future carbon intensity of the economy, or how rapidly and widely carbon capture technologies will be deployed. The immediate products are scenarios such as the representative concentration pathways (RCPs) used in reports of Intergovernmental Panel on Climate Change (IPCC) (van Vuuren et al. 2011; O’Neill et al. 2014). These scenarios are then returned to the climate scientists, who run their models and make climate projections.

The process described above is a simplification. In reality, climate economics and climate science are deeply entangled disciplines—something illustrated by the fact that the IAMs themselves are interdisciplinary models of a sort. However, the simplification usefully conveys an exchange of problems and solutions across disciplinary boundaries of the sort that we, in what follows, will refer to as problem-feeding (PF) (Thorén and Persson 2013; Thorén 2015a, 2015b; Wahlberg and Persson 2017; Persson, Thorén, and Olsson 2018; Thorén, Persson, and Olsson 2021; Thorén, Soininen, and Kotamäki 2021). PF is of interest as a model of interdisciplinary research for several reasons. One is that, as we shall argue below, PF is a non-integrative form of interdisciplinarity and thus provides an example of interdisciplinary exchanges that do not conform to the standard typologies. Another is that these kinds of exchange appear both important and somewhat common, and connect to central themes in the literature on interdisciplinarity such as the idea that interdisciplinarity is problem-oriented. A third is that PF makes a lot of sense from the point of view of division of labour: emphasising the exchanges between disciplines can be helpful both generally—as we should strive, where possible, to direct problems to researchers with the right expertise—and as a way of approaching and structuring concrete interdisciplinarity.

This paper is primarily explorative. Our main aim is to examine and develop the PF model by filling in some of the gaps in the existing literature on it. We think of this model as having both descriptive and normative significance as regards the nature of interdisciplinary research. One central task of the paper is to outline some of the most important prerequisites of engagement in PF. Another is to discuss the PF model in relation to other conceptions of interdisciplinarity, focusing in particular on the notion of integration. We will argue that PF can be seen as a form of non-integrative interdisciplinarity.

The paper is structured as follows. In Section 2, in schematic form, we describe the structure of problem exchanges as these are understood in the PF framework and highlight some of the critical general issues. In Section 3 we discuss the distinction between problems and disciplines, and what this distinction means for our understanding of PF. Section 4 examines the prerequisites of PF in three dimensions: the ontological-conceptual, epistemological-methodological and axiological dimensions. Finally, in Section 5 we turn to the concepts of interdisciplinarity and integration, arguing that PF sits somewhere between what are conventionally considered multidisciplinarity and interdisciplinarity (more fruitful than the former and less demanding than the latter).

2. The Structure of Problem-Feeding

We begin by outlining the overall structure of PF. For convenience, let us call the discipline in which a problem arises the source discipline and the discipline to which the problem is transferred the receiving discipline. A simple outline of PF is as follows:

1. Problem P arises in discipline A

2. Problem P is solved by discipline B

A problem that ‘arises’ in the source discipline is recognised as somehow ‘belonging’ to another discipline and is subsequently transferred to that discipline, where it is duly solved.¹

The schema appears straightforward, but there are reasons to complicate it. One important reason is that it seems that often a problem must be fitted, or adapted, to the various cognitive and epistemic resources of the receiving discipline—its concepts, theories, models and so on. That is to say, the problem must be re-expressed so that it fits the cognitive and epistemic resources available in the receiving discipline. This frequently requires effort, at least initially. The point is that the transfer of a problem may be an important step in its own right. Consider a revised schema for PF:

1. Problem P is formulated in discipline A

2. Problem P is transferred to discipline B as P’

3. Discipline B provides a solution S to P’

4. Solution S is transferred back to discipline A

Following Thorén and Persson (2013), we can differentiate between unilateral PF and bilateral PF. The former omits step 4, perhaps because the relationship between P and P’ fails to be structured in the right way, and its interdisciplinary merits are weak at best. Bilateral PF, on the other hand, is ‘PF with solution-feeding’ (Thorén and Persson 2013, 346), a form of exchange in which all four steps in the schema are taken. The exported problems are solved in such a way that the solutions can be deployed (whatever that means) within the source discipline.²

In what follows we will use PF to denote the bilateral variety unless otherwise stated.³ Anticipating what is to come, we should say right away that whether bilateral PF can be achieved depends crucially on the relationship between problem P in the source discipline and problem P’ in the receiving discipline. Minimally, it must be the case that the relation between the two problems is such that the solution S is either considered a solution to both P and P’ or, otherwise, admits of appropriate transformation as it is transferred from the receiving discipline back to the source discipline.⁴

3. Problems and Disciplines

A problem, according to Stephen Toulmin (1972, 152) conforms to the following formula: $\mathrm{Scientific}\mathrm{problems}=\mathrm{explanatory}\mathrm{ideals}-\mathrm{current}\mathrm{capacities}$Along the same lines, we can characterise a problem as a discrepancy between, on the one hand, some ideal or expectation, and on the other, a (perceived) state of affairs (Thorén 2015b, 64). The ideal might be an explanatory one assigning to some theory, or discipline, a certain explanatory domain, or set of facts, roughly of the form ‘T should explain D’. Alternatively, it could be a practical ideal concerning, say, what the world should be like (e.g. ‘there should be a bridge here’), or a strictly theoretical ideal such as the (theoretical) unification of some domain or set of theories. The other component is some perceived state of affairs that runs counter to this ideal. Examples would include a set of observations that appear not to be explained by the relevant theory, a theoretical inconsistency within or between theories, or the absence of the aforementioned bridge. This characterisation of a problem covers empirical problems and conceptual problems (Laudan 1978) as well as what would count as social or societal problems.

To solve a problem is to relieve the tension in discrepancies of the kind just adumbrated. This can happen in many ways, including demonstrating that the tension itself is illusory or unimportant. Thomas Nickles (1981) notes that a problem, once appropriately well understood, contains within it a set of constraints on its solution—that is, it holds, not the solution itself, but what essentially amounts to a specification of what an admissible solution looks like. Understanding a problem well is to recognise what a solution should look like but does not amount to possession of that solution. Sometimes the constraints on solutions make problems impossible to solve (Thorén 2015b, Chapter 3).

Problems can also be stated with greater or less specificity. Thus they range from questions that can be researched immediately to extremely vague and unspecific enquiries (e.g. of the kind that arise when the problem is not well understood). An example of the former, call them research problems, would be: Is substance S at concentration C, over time t, statistically correlated with disease D in a patient, given a p-value of 0.05 in appropriately selected samples of a population? An example of the latter, which we might call general problems, is perhaps: What is the fundamental structure of the universe? Research problems can be answered directly by conducting research, and in that way solved, but this is less likely to be the case with general problems. Their function, rather, seems to be agenda-setting—to provide, within a discipline, a sense of overall direction and perhaps identity.

The kinds of problem that seem of be of interest in the present context lie somewhere between these two extremes. They are mid-level problems taking the form of a somewhat specific question, but the terms and concepts involved in them need to be further operationalised, or concretised, before they can be researched.

The typical situation mandating PF is one where the source discipline, in pursuing a solution to a problem, A, comes across another problem, B, such that it is the case: (i) either that, in order to solve A, B first must be solved, or that a solution to B is an essential component to a solution of A; and (ii) B is such that it cannot be solved by the source discipline alone. In other words, the problem that is subject to a PF exchange is part of a larger, decomposable problem.

Moving on to disciplines. What is a discipline? This is a simple question with a potentially very complicated answer, depending on what expectations we have. For present purposes it is not important to develop a concept of disciplines that can, for example, exhaustively and completely categorise all of science (or some such thing). We are less concerned with what might be referred to as the (inter)disciplinary structure of science as a whole, and more concerned with situations where two disciplines recognise overlapping interests, or some mutual dependencies, yet find it difficult to make the most of this recognition.

There is no consensus, among the comparatively few philosophers that have engaged directly with this issue, on how we should conceive of disciplines, and the distinction between disciplines and problems. Karl Popper (1963, 88), for example, dismisses the idea that disciplines are epistemologically significant entities out of hand. To him they are merely administrative entities and the consequences of historical accident, and simply not the right kind of thing to appropriately organise scientific inquiry around. Scientists qua scientists should follow problems wherever they may lead (ibid.). Problems are, to Popper, distinct from—indeed, almost entirely independent of—disciplines. He recognises only that some problems seem to belong to disciplines in the sense that they have sprung from that discipline’s own theorising (Thorén 2015a). Thomas Kuhn (1970, 182ff), on the other hand, imagines disciplines as central to science. In many ways they are what guides those working in them to what good science is. Interestingly, problems are perhaps the most central feature of disciplines for Kuhn. Exemplars are paradigmatic problems and problem solutions that shape the development of disciplines by illustrating, or modelling, the ways in which new problems should be tackled (i.e. by modelling them on old ones).

Neither of these attitudes offers a particularly fruitful point of entry to PF. Although Popper seems to endorse a kind of baseline interdisciplinarity in that scientists should follow problems wherever they may take them, his approach seems to guarantee that PF-like situations never occur. From Kuhn’s perspective, on the other hand, disciplines appear too insular. Within a discipline it can be useful to think of a particular problem as belonging to some other discipline, and in that way protect the paradigm from anomalies. Such a move, however, at best leads to unilateral PF.

Nevertheless, we will deploy a concept of disciplines that cleaves more closely to what Kuhn had in mind than it does to what Popper says. We will also borrow elements and vocabulary from William Bechtel (1986) and Darden and Maull (1977). Darden and Maull imagine fields—functional equivalents of disciplines, in their account—as having a number of elements. The most central of these elements are ‘a central problem, a domain consisting of items taken to be facts related to that problem, general explanatory factors and goals providing expectations of as to how the problem is to be solved, techniques and methods, and, sometimes, but not always, concepts, laws and theories, which are related to the problem, and which attempt to realize the explanatory goals’ (44). They also mention that a ‘special vocabulary’ is often associated with a field. Perhaps it is useful to differentiate between surface features and deep features of disciplines. The former might be those concepts or techniques that are deployed within the discipline but are, at the same time, both subject to constant change and modification and can comparatively easily themselves be subject to transfer. The latter, the deep features, are those commitments of disciplines that are more firmly held and subject to a lesser extent to change and exchange. That may mean things like epistemological commitments or ontologies (in the Kuhnian sense). It could probably be extended to include other things as well. In practice, we take it, these deep features of disciplines are embedded in more immediate practices and principles, such as Kuhnian exemplars or the general explanatory factors and goals that Darden and Maull mention.

Disciplines have other institutional and social features that are also important, not least in the role they serve of stabilising and reifying commitments and practices through time. But, to simplify and focus the discussion, and following Mäki (2009) in a rough and ready way, we will focus on three dimensions of disciplinarity (see also Eigenbrode et al. 2007). The first is conceptual-ontological. This concerns the domain of items under investigation in a specific discipline, and how those items are conceptualised. The second is epistemological and methodological. This has to do with how knowledge claims are produced and evaluated as knowledge claims. The central concern here is with such matters as evidentiary standards and what kind of inferences are permitted or considered appropriate given a specific body of evidence, and so on. The third dimension is axiological, and it concerns the role of non-epistemic values and priorities within a discipline, which sometimes act as pointers to what is interesting or important or central. This does not exhaust the potentially relevant dimensions, and the problem could be analysed at higher granularity using finer categories. In order to keep the account reasonably brief, however, we limit ourselves to these three dimensions.

A final comment. What makes the prospect of transferring problems difficult is that disciplines provide some of the backdrop against which problems arise. Disciplines are at the same time constitutive of the problems themselves, and the means by which we explore problems to better understand and eventually solve them. There is always a tension between exploring a particular problem and formulating a new, and more interesting, one that better fits the commitments and resources of a particular discipline (Thorén 2015b). Moreover, as Kuhn (1970) famously noted, scientists are rarely interested in problems that are, for one reason or another, impossible to solve. Instead, they opt for challenges of a more technical nature, albeit preferably a difficult in nature as well, the tackling of which will demonstrate the individual scientist’s skills. The point is that the disciplines here can operate to break problem-solution coordination by pulling the problem-solver in what, from inside the discipline, seems to be a more convenient and interesting direction.

4. The Prerequisites of Problem-Feeding

We have already gestured in passing at a few prerequisites of PF. One is that the problem must be sufficiently specific. General problems can easily be transferred but, one suspects, leave too much unsettled for the exchange to bilateral. The main objects of transfer are what we have called research problems and mid-level problems. As research problems, if they are indeed adequately formulated, are immediately possible to transfer without transformation the interesting and challenging situations are likely to arise in connection with the mid-level problems.

4.1. The Conceptual-Ontological Dimension

Bilateral PF is essentially a way of organising a division of labour between two (or more) disciplines. For this to be motivated to begin with there has to be a recognition within the source discipline that the problem in question is important within that discipline but can only be solved with the help of some other identifiable discipline. This will not necessarily be obvious, as it seems to imply that members of the source discipline have recognised that their own discipline has limits, even with respect to problems that are important to it. As the literature on scientific imperialism indicates, some disciplines nurse an expansionist self-image, making such recognition unlikely (Dupré 1994; Mäki 2009). A first observation, then, is that PF appears to require, minimally, the recognition that one’s own discipline has limited scope, and that where one’s own discipline ends, another may begin.

A good place to start is with Darden and Maull’s (1977) idea of an interfield theory.⁵ This is a theory that makes ‘explicit and explain[s]’ the relationship between two fields in ontological terms (Darden and Maull 1977, 48), spelling out how items in the domains of the respective disciplines relate to one another. For example, an interfield theory may stipulate that the physical nature of an entity postulated in one field can be investigated by another field, or that one field studies the function of a phenomenon the structure of which is studied in another field, and so on. A very good example of this, detailed by Darden (1991, 80ff) is the interfield Boveri-Sutton chromosome theory. This theory, by stipulating that the location of the gene or factor (a Mendelian concept) is in or on the chromosome (a cytological concept), tied two fields together in a productive way. A remarkable feature of the theory is the way it allowed two disciplines to resolve curious anomalies—the fact that genes have locations in physical space means that their spatial proximity on the chromosome was capable of explaining small deviations, in the frequency distributions of independent traits, from what would have been expected under perfectly random assortment.

That disciplines exchanging problems need to have some interface vis-à-vis one another is beyond doubt. There simply has to be a reason for engaging with one another, and at some level of discourse this needs to imply a connection of some sort. It is equally clear that interfield theories can serve this function. However, they seem to be neither necessary (narrowly construed) nor sufficient (Thorén and Persson 2013, 347). As regards sufficiency, problem solutions also need to be admissible in the source discipline if bilateral PF is to happen. For example, climate scientists may well accept that there is an interfield theory that arranges climate science and sociology with respect to one another in a way that is adequate without for that reason finding sociological results convincing or useful. Moreover, an interfield theory stipulating that two disciplines have the same domain but providing nothing more also appears problematic from a narrow PF perspective. For it will not necessarily be conducive to problem/solution exchanges—at least, in the short term—even if it is clearly very fruitful inasmuch as it fosters, for example, debate and discussion across disciplinary boundaries. The historical relationship between philosophy of mind and neuroscience is perhaps a good example of this.

As far as necessity goes, the main worry is that it is not quite clear what kind of relationship the interfield theory demands. If the demand is for a strictly ‘ontological’ relationship, it seems that other problem/solution exchanges that have nothing to do directly with the objects of inquiry in the respective disciplines, or are more tentative and loose in nature than what Darden and Maull perhaps had in mind, will be possible (Thorén and Persson 2013, 347). For example, sometimes the interdisciplinary relationship can be organised with the aid of modelling templates. MacLeod and Nagatsu (2016, 2018) show how this happens in collaborative projects involving ecologists and economists. They identify several such templates. One is what they call substitutive model coupling, where members of different disciplines, in accordance with their expertise, contribute with ‘legacy’ models that are then coupled by harmonising input/output relationships. The ecologists in one of their examples contribute to the building of a coupled model, together with economists, by substituting simplistic population growth models designed by economists with their own more sophisticated version. This task is comparatively easy. Inputs and outputs have to be harmonised to allow for communication, and an eye has to be kept on overall model complexity—which can sometimes render such approaches non-viable (see e.g. Voinov and Shugart 2013).

Modelling templates are perhaps both more tentative and more general than interfield theories narrowly construed, suggesting that we do not need to rise to the level of a fully-fledged interfield theory to facilitate fruitful exchanges of problems and solutions.

To conclude this section, let us remark on shared concepts. Establishing conceptual connections of the appropriate sort is often emphasised as an important element in the fostering of exchanges of problems, and failing to acknowledge subtle differences between central concepts can lead to a failure of bilateral PF (see Thorén and Persson 2013). This may concern higher-level concepts such as cause (Wahlberg 2010) or uncertainty (Thorén, Soininen, and Kotamäki 2021). Here failing to recognise important conceptual differences may lead members of receiving and source disciplines to think there are opportunities for bilateral PF where there are none.

However, there seem to be situations where conceptual homogeneity across disciplinary boundaries is less important. Nancy Maull (1977), for instance, suggests that some terms are genuinely shared between fields in such a way that both fields have access to them and can attach to them various knowledge claims. Such shared terms enable what she calls ‘problem-shifts’ (149ff)—i.e. precisely the kind of problem exchanges in which we are interested here. Illustrating this, a concept such as mutation had certain associations in the late nineteenth-century thinking on evolution, where it was used to refer to successional subspecies. Then, with De Vries, it came to designate, within genetics now, a change in pangenes, and eventually, in biochemistry, changes in the genotype of an organism. The significance of this term-sharing, Maull suggests, lies in the way it allows for these fields to formulate and solve problems. Geneticists were unable to resolve issues relating to the physical nature of hereditary determinants they had stipulated, and the transformation of the term ‘mutation’ from genetics to biochemistry was necessary to address these issues.⁶

4.2. The Epistemological-Methodological Dimensions

The epistemological dimension concerns how disciplines interface with respect to their epistemic values and epistemological and methodological commitments. This has to do with, among other things, what passes for admissible evidence, what constitutes an acceptable inference from a given set of data, and what approach is thought to be suitable to deliver admissible or reliable outputs in a given situation. That such differences may occur, and be a source of problem-solution coordination between disciplines that would otherwise recognise one another as candidates for interdisciplinary research, is quite clear.

A helpful illustration of this can be found in Evelyn Brister’s (2016) account of attempts at interdisciplinary exchanges between conservation biologists and anthropologists. The case she examines involves the development of conservation solutions in central Africa. Broadly speaking, it is recognised that when one is developing conservation interventions one should keep an eye out for consequences and knock-on effects beyond biological and ecological parameters such as biodiversity. For instance, sometimes designating an area as protected will disrupt the livelihoods of people in or near that area who depend on its ecosystem services. Failing to take impacts like these into account is both morally hazardous and can substantially lower the chance that the intervention in question will be successful. Hence an ‘interdisciplinary’ approach that involves both conservation biology and appropriate social science disciplines is mandated. This will frequently assume the form of a PF relationship in which, for example, conservation biologists have particular identifiable problems that need to be resolved by social scientists.

Yet, as Brister points out, it is one thing to recognise that there is a need for interdisciplinarity and another to engage in actual interdisciplinary research. A problem that is typical of the kinds of case she examines is that the epistemic standards determining, say, what kind of evidence is acceptable, or what kind of approach will be reliable, disrupt the exchange of problems and solutions between the respective disciplines. The conservation biologists have questions, or problems—for instance, revolving around whether an intervention is likely to result in involuntary displacement due to loss of local people’s livelihoods. They also know, or have decided, where these problems should be exported. For example, that the issue is one for anthropology. But they are not willing to accept the solutions offered by the social scientists, as the approach through which these were devised is one they do not consider epistemically admissible.

One way to solve this problem is simply by establishing explicit new standards of inquiry that are embraced and accepted by all parties involved. Evidence-based medicine (e.g., Howick et al. 2011) and, to some extent, evidence-based conservation (see e.g. Bilotta, Milner, and Boyd 2014) have both struggled with this problem (Persson, Johansson, and Olsson 2018), attempting to develop useful and generally accepted evidential hierarchies for a range of common types of question within the relevant fields.

Evidence-based medicine de-emphasizes intuition, unsystematic clinical experience, and pathophysiologic rationale as sufficient grounds for clinical decision making and stresses the examination of evidence from clinical research. (EBMWG 1992, 2420)

Although there is no reason to think that new standards of inquiry would not aid collaborative efforts, it is often out of reach in practice, or at any rate highly labour-intensive to achieve, particularly in interdisciplinary settings. A better strategy, in situations where this is indeed the case, is to avoid conflict by minimising the points of contact. This allows for a degree of disciplinary independence and autonomy, and lowers the salience of disciplinary conflicts (see e.g. MacLeod and Nagatsu 2016, 2018). As members of the source discipline are unilaterally dependent, epistemically (Andersen and Wagenknecht 2013), on members of the receiving discipline, a fundamental component of this strategy is that the former have faith in the truthfulness and competence of the latter.

Harris and Lyon (2013) have pointed to the importance of building trust in transdisciplinary relationships and emphasise their interpersonal dimension, which is usually built through previous experience of working together. Norms of co-operation are important, as are ways of sanctioning transgressions of those norms. This suggests the scope of these relationships may be quite limited, and even, perhaps, confined to individual problem situations.⁷

At the same time, there are clearly interdisciplinary arrangements where trust is misplaced or even somewhat coercive. Some disciplines, as Mäki (2009) has pointed out, have considerable standing of a kind that can be put to good use under certain conditions. Social scientists are perhaps more likely to accept the work of climate scientists—regardless of how that work came about—under conditions in which the perception prevails that the results support, rather than conflict with, various social science assumptions and claims.

To conclude this section, intractable issues may arise when the values and norms associated with good science—what are sometimes called epistemic values (Kuhn 1977; McMullin 1982; Steel 2010)—in different disciplines clash with one another. This presumably mandates either an alignment of epistemic values ensuring that the disciplinary outputs can be accepted in all of the disciplines involved, or some other arrangement such as the establishment of trust (Andersen and Wagenknecht 2013; Thorén and Persson 2013). Notably, in the latter case, it could be pointed out that one reason for establishing these relationships of trust, even when they suppress conflicting epistemic values and priorities, is that interdisciplinarity itself is a value. A question that then arises is: Is interdisciplinarity an epistemic value? We will not attempt to answer this question in this paper, although we will consider an aspect of it in the next section.

4.3. The Axiological Dimension

This dimension, as we think of it here, has to do with the role of non-epistemic values in disciplines involved in bilateral PF. Non-epistemic values include such things as the social and ethical values that influence and shape a range of activities and choices within disciplines (McMullin 1982). They can enter PF in several ways. First, they play a central role in determining what problems, or hypotheses, to pursue, presenting those undertaking bilateral PF with coordination challenges. Just because a discipline is suited to solve a problem in the sense that it has access to certain problem-solving resources, it does not follow that this problem is also interesting to that discipline (Thorén 2015a). In the receiving discipline a problem exported from another discipline may be perceived as trifling and uninteresting. Those in the receiving discipline may believe that the problem does not enable them to make progress towards their central aims and priorities. As this highlights, the benefits of interdisciplinarity can be unevenly distributed among the participants.

Second, if we think of the problems that are being exchanged as often being mid-level problems standing in need of further specification, there is also always the possibility that the transformation of the problem, a process that may also be shaped by non-epistemic values, will weaken the connection to the problem originally encountered in the source discipline—perhaps to breaking point. If this connection is not adequately maintained, solutions will cease to seem relevant and bilateral PF will fail. Often the actual maintenance work will, perhaps, be the responsibility of the receiving discipline, but this is not the only conceivable way of doing things. How the needs that arise in the transference of problems are dealt with can be a matter of methodological negotiations between both involved disciplines. It is also possible for the source discipline to be compelled to adapt a problem’s formulation in ways entailing a substantive departure from what was originally encountered, in which case it is the source discipline that is adjusting. In such cases the relationship between the original problems and other problems in the source discipline are likely to impose constraints. Nonetheless, adjustment in the source discipline is certainly a possibility under certain conditions.

What this means, in practice, is that interdisciplinarity itself has to be valued and prioritised. That is to say, at least one of the involved disciplines needs to prioritise the interdisciplinary effort over their pursuit of central problems in their own discipline.⁸ The reason for this is that disciplines are generally not well aligned in this respect out-of-the-box. Any interaction between two disciplines involves adapting and adjusting to the needs and priorities of other researchers with different disciplinary backgrounds. This connects, albeit somewhat loosely, to a point philosophers have often made about interdisciplinarity—namely, that it (interdisciplinarity) is instrumentally rather than intrinsically valuable. That is, whether interdisciplinarity should be pursued should depend on the problem situation: sometimes disciplinary approaches are better, sometimes interdisciplinarity is to be preferred (Hansson 1999; Grüne-Yanoff 2016; Mennes 2020). Again the demands of the problem situation should determine which approach is suitable, and a premature commitment to interdisciplinarity is no less dogmatic than a premature commitment to disciplinarity (see Popper 1963, 88). It seems that it is commonly assumed that the values guiding the formulation of, and prioritisation among, problems are fixed. This is not unreasonable, but interdisciplinarity will often, we suspect, involve adjustments in precisely these values. Perhaps a better way of characterising the situation is roughly as follows. At the outset, and well before they have anything like a complete understanding of the problem-situation, members of the respective source and receiving disciplines must appraise the situation, assessing how fruitful and promising it is likely to be (c.f. Nickles 2006). Once engaged in the exchange, however, interdisciplinarity itself becomes the dominant value. Now it is what motivates value adjustments that feedback on to the structure of the original problem.

A further challenge, or set of challenges, emerges from the fact that non-epistemic values are tangled up with epistemic judgements in various ways. Recent discussions in the philosophy of science focusing on inductive risk arguments have highlighted the ways in which non-epistemic values are involved in setting epistemic standards and methodological choices (Rudner 1953; Douglas 2009; Steel 2010, 2013; Elliott 2011; Steele 2012; Parker and Winsberg 2018; Winsberg, Oreskes, and Lloyd 2020). Non-epistemic values are also built into many scientific concepts (Putnam 2002; Dupré 2007) and may be involved in closing the gap between theory and observation (or data) (Anderson 2013). In practice, it can often be difficult to distinguish between these types of value influence. The epistemic and non-epistemic values are entangled and affect one another.

What this means is that even when we are examining what appear to be exclusively epistemic standards, or purely epistemic methodological choices, we will often have to consider, also, the disciplinary differences that turn on non-epistemic values, both to understand why coordination is difficult to achieve, and what to do about any disciplinary discrepancies. The tenuous relationship between climate science and climate economics is a case in which we can see both epistemic and non-epistemic values becoming actively involved in an interdisciplinary dispute. For example, one problem with many climate-economic models is that they fail to represent central feedback mechanisms in the climate system adequately (Pindyck 2017; Pezzey 2018; Keen et al. 2021). This issue has to do with epistemic values, such as how faithfully the models represent their target systems and what that means as regards what we can learn from the models. Equally, however, it has to do with non-epistemic values and questions about how to represent and evaluate risks and costs.

Whether differences in non-epistemic values impede collaboration will depend on the several specifics of the case. First, presumably not all values are actively in play all the time. As disciplines are amorphous and changing entities that themselves contain diversity, the key need is probably to examine value conflicts that are made salient by a particular problem situation. Second, assuming that value entanglement is a fact, it seems that where such things as epistemic standards and methodological choices are concerned more than one route may well permit arrival at the same destination. This means that there are degrees of freedom, for participants in an exchange, to find ways of mitigating their value differences. Third, thoroughgoing harmonisation is presumably not necessary, although it is difficult to say, in principle, how close one needs to get to this ideal.

4.4. Problem-feeding when it Fails

In concluding this section, we want to say something about the implications of PF failure. What exactly does it mean for PF to fail? Failures can happen when there is a need for PF, perhaps of a certain kind, but this need is not met. This is one way of thinking of Brister’s example. There, there was an arguably justified basis for engaging in bilateral PF, but the attempt failed at the solution-feeding stage owing to a combination of factors including differences in epistemic (and perhaps non-epistemic) values and lack of trust between actors. Our suspicion here is that unwillingness to adapt to the needs of one’s partners is a common issue. In such cases the idea is that while PF is possible, it fails to work because the members of the disciplines involved have, in essence, a skewed perception of the costs and benefits of engaging with one another.

Rather differently, PF can fail when problem-solution coordination is not achieved and the parties involved do not recognise this. For example, subtle differences in the understanding of concepts may make it seem that discipline B has a solution to discipline A’s problem, although in actual fact this is not the case. This kind of situation is sometimes labelled a type III error—i.e. the provision of the right answer to the wrong question (Mitroff and Featheringham 1974; Wahlberg and Persson 2017; Thorén, Persson, and Olsson 2021). Conceptual differences, or unacknowledged contrasts in the epistemological and axiological backdrop within disciplines, may result in problems and solutions coming undone or failing to connect. If these differences are subtle, the decoupling or original lack of connection may not be seen and acknowledged.

5. Problem-feeding, Interdisciplinarity and Integration

The PF model provides a way of understanding certain forms of interdisciplinary interactions. One promise that this perspective offers is that it opens up connections with certain common themes figuring in the debate on interdisciplinarity, such as problem-orientedness.⁹ To be clear, we do not mean to suggest that interdisciplinarity is always of the PF sort. Jan Schmidt (2022) distinguishes problem-oriented interdisciplinarity from object-oriented, theory-oriented and methodologically driven interdisciplinarity. Nor are we suggesting that all instances of PF are interdisciplinary, even in a broad sense. The PF model gets a purchase on its subject when epistemically differentiated actors pool their resources to tackle common, but decomposable problems. Conceivably, this can also happen within disciplines.

In practical terms, the model can serve a normative role in certain settings by highlighting complementary rather than conflicting aspects of disciplines that engage in interdisciplinary problem-solving and promoting a sensitivity to problem-solution coordination which can be useful when diagnosing problems in specific attempts at interdisciplinary problem-solving (c.f., O’Rourke and Crowley 2013). The point we wish to convey here is that problem-solution coordination is not always easily maintained in interdisciplinary research, and drawing attention to the way problems and solutions are exchanged in constructive ways is of practical value to interdisciplinarians.

In the remainder of this section we will offer some observations on the place of the PF model in the wider context of current thinking on interdisciplinarity in the literature. We have thus far been using the term ‘interdisciplinarity’ loosely to designate ‘the’ way (i.e. any and every way) in which two or more disciplines might engage with, or relate to, one another (see Mäki 2016). However, interdisciplinarity comes in many different forms (Huutoniemi et al. 2010; Klein 2017). Here we will focus on some of the more prominent of these, and on themes around these.

By some distance, the most familiar typology is that separating multidisciplinary, interdisciplinarity and transdisciplinarity (see e.g. Apostel et al. 1972; Klein 1990; Alvargonzález 2011). Within this tripartite typology, multidisciplinarity is understood to be ‘additive’ and non-integrative, involving only the ‘juxtaposition disciplines’ (Klein 1990, 56)—a process that leaves the respective disciplines ‘neither changed nor enriched’ (Klein 1990, 56). Interdisciplinarity is the next rung on the ladder. It is often explicitly said, or strongly implied, to be a superior kind of crossdisciplinarity, differing from multidisciplinarity in going beyond the ‘merely’ additive and involving some degree of integration. Transdisciplinarity, finally, is distinguished from the other two types by its involvement of non-academic actors in some substantial capacity—although this term, too, is in fact associated with various meanings (Bernstein 2015; Koskinen and Mäki 2016; Thorén and Breian 2016). In our view the PF model can be applied to at least some transdisciplinary situations, but we will confine ourselves in what follows to inter-academic situations.

Where does PF fit in this framework? Is it multidisciplinary or interdisciplinary? As we have noted, there are two aspects to this question. One is about whether a given exchange between disciplines drives change, or development, or progress, or some such thing, where the disciplines involved are ‘enriched’. The other has to do with integration. As regards the former, it seems to follow by definition that bilateral PF is enriching in some significant sense for at least one of the disciplines, and sometimes for both (such as in Darden and Maull’s case). Problems important for the source discipline are solved, and by solving problems progress is made.

Integration is a stickier issue. While it is central to the way many think of interdisciplinarity ‘that to question whether ID involves integration is almost heretical’ (Holbrook 2013, 13), very few publications in the interdisciplinarity literature have attempted to specify its meaning (O’Rourke, Crowley, and Gonnerman 2016). A common idea is that integration involves the reduction of conflicts or inconsistencies between disciplines, and the establishment of a common methodological, theoretical and perhaps ontological framework. This unificationist understanding of integration is often regarded as too demanding, especially in the context of real-world interdisciplinary projects, since these mostly operate with significant time and resource constraints.

Yet it is a mistake to think that integration is one thing. Many of the numerous philosophical accounts of integration are in tension with unificationist ideas. An example is Sandra Mitchell’s integrative pluralism. In this, explanations can be integrated at the level of concrete phenomena, but not at the theoretical level, where explanations remain distinct (Mitchell 2009). Beyond this kind of explanatory integration, O’Malley and Soyer (2012) highlight two further kinds: methodological integration and data integration. Grantham (2004) draws a different distinction, separating theoretical integration, which involves relations between explanations, ontological relations and other conceptual relations, from practical integration. The latter, Grantham suggests, include methodological integration, confirmation dependence and heuristic dependence. Going further, Grantham (following Kincaid (2008)) argues that integration is a matter of degree. Integration revolves around the various kinds of connections that can be established between disciplines. The more interconnections there are, the more integrated the disciplines can be said to be.

Michael O’Rourke and colleagues (O’Rourke, Crowley, and Gonnerman 2016) have developed an ‘integrative’ account of integration—essentially a framework into which one may fit different understandings of integration. Their idea is that integration is an ‘input-output process’ in which elements of different sorts are fitted together according to different more or less precise principles. When we consider an integrative episode, we therefore need to be specific about such things as the kinds of element involved (data, theories, practices, models, and so on), how many there are, how they are to be characterised (are they social, epistemic, cognitive, or what?), and various aspects or expectations of the process itself, as well as the kinds of output that the integrative event is expected to yield. Different kinds of interdisciplinarity may emphasise different steps in this process. Sometimes the emphasis is on integration as a property of the output, as seems to be the case in Schmidt’s theory-oriented interdisciplinarity. At other times, the process is what is central.

Is PF integrative? A first observation is that PF, like any kind of interdisciplinarity, requires a degree of integration to be already in place in the sense that some connections between the disciplines needs to exist in order to facilitate the exchange. Hence the process of making PF work will sometimes involve establishing, maintaining or bolstering the connections between disciplines which, plausibly, involves integrating those very disciplines. The outcome, however, is not necessarily, or even typically, ‘integrated theories’ (see Schmidt 2022) or ‘a common language, joint categories, methods, and a common research design across disciplines’ (Grüne-Yanoff 2016, 347). Successful PF solves problems.

As Thorén and Persson (2013) note, Grantham’s practical integration comes close to capturing some key aspects of PF. Confirmational dependence, for example, covers situations where ‘[t]he methods and/or data of one field may be used to confirm hypotheses generated in a neighboring field’ (Grantham 2004, 143). Such situations may be reconstructed as instances of unilateral PF, albeit perhaps of a shallower sort.

One strand of the literature on interdisciplinarity contains the idea that there are non-integrative forms of interdisciplinarity. Thus, drawing on examples illustrating exchanges between economics and disciplines like biology and psychology, Till Grüne-Yanoff (2016) argues that interdisciplinary success can be achieved even without constructing integrated frameworks. In other words, there is space between what is conventionally labelled multidisciplinarity and interdisciplinarity in the narrow sense for fruitful exchanges that do not result in integration.

Britt Holbrook (2013) has also considered alternative accounts of interdisciplinarity that sit alongside the integrationist approach he labels the Habermas-Klein thesis. Holbrook calls one of these alternatives the Batallie-Lyotard thesis. This says that communication between disciplines is often straightforward and proceeds as if everyone understands one another and abides by the same rules. Under these conditions, the parties operate within the same genre of discourse, allowing us to ‘litigate’ between them to settle disputes. The crosswise communication becomes truly challenging when parties, or disciplines, are operating, not with the same genre of discourse, but different and incommensurable ones. Normal (weak) communication then breaks down, and an alternative must be pursued. An entirely new genre of discourse must be invented: ‘one that is not merely an integration of the previously existing genres, but a novel co-creation of those who have risked and relinquished their previous disciplinary identities’ (Holbrook 2013, 1876).

The other alternative is called the Kuhn-McIntyre thesis. Here, too, the emphasis is on the incommensurability of disciplines and, in a certain sense, the possibility of the kind of consensus-seeking that is central for proponents of the Habermas-Klein thesis is denied. Interdisciplinarity here is not motivated by the aim of integration (although the rational resolution of conflicts is under certain conditions possible). Rather—in a formulation that comes close to invoking PF—it is motivated by situations ‘when members of one discipline realize that they do not possess the ability to address a problem in a satisfactory way using only the resources native to that discipline’ (Holbrook 2013, 1872). Interdisciplinarity in the Kuhn-McIntyre paradigm is demanding, however. Often it requires the interdisciplinarians effectively to learn a ‘second first language’ in order to avoid efforts to translate between disciplinary languages (which is impossible) and adjudicate positions on their own terms, from the ‘inside’. Here PF may demand less. It is a partial adjustment between disciplines along certain dimensions to coordinate efforts.

To bring this line of reasoning to a close, it seems to us that the PF model emerges as an alternative to integrationist accounts of interdisciplinarity that is less demanding and builds on more tempered, and perhaps more realistic, expectations. In its descriptive and normative modes, the PF framework could serve to assist interdisciplinarians in understanding and shaping collaborative enterprises by increasing their sensitivity to the capabilities and limitations of their own and other disciplines. This promotes a certain kind of interdisciplinary interaction that emphasises complementarity over conflict.

We want to conclude this section with a note on the collaborative aspects of PF and interdisciplinarity more generally. In practice the main challenges of interdisciplinarity that animate practitioners revolve around interdisciplinarity as a concretely collaborative activity rather than the abstract bridging of theories, or bodies of knowledge, that has traditionally been focused upon by philosophers of science (Schmidt 2008). Collaboration involves intentional interpersonal connections of some sort between researchers, or scientists, from different disciplines with the purpose of formulating and/or pursuing common problems. This, too, can come in different forms. Margret Boden (1999) distinguishes between three such. Shared interdisciplinarity involves a situation in which ‘different aspects of a complex problem are tackled by different groups with complementary skills’ (17). Boden considers this to be a ‘minimal’ kind of collaborative interdisciplinarity—it is like children sharing cake. There is some communication, and some monitoring of progress, but little else. In co-operative interdisciplinarity the exchanges are deeper, with ‘several groups with complementary skills work towards a common goal, actively co-operating on the way’ (Boden 1999, 18). This is altogether more demanding. Boden writes:

Co-operative interdisciplinarity requires four things from the individual researchers: respect for other disciplines; willingness to communicate not only results, but also progress and problems; willingness to give, and to receive, suggestions about how a method used by another group might help address the problems faced by one’s own; and, willingness, when appropriate, to co – operate in developing new research methods and/or research areas. (Boden 1999, 18)

The final form, which Boden calls integrative interdisciplinarity, is ‘an enterprise in which some of the concepts and insights of one discipline contribute to the problems and theories of another—preferably in both directions’ (Boden 1999, 20). This, according to Boden, is the ‘strongest’ form of interdisciplinarity. Really, it is the only form she thinks of as being truly interdisciplinary, as it is the riskiest. It involves genuinely exposing two disciplines to one another in ways that can be very fruitful but also challenging, and even damaging, at least in certain ways, to the disciplines involved.

This last form of interdisciplinarity approaches disciplinarity in ways we have already indicated. Although some have taken this to be the most collaborative of Boden’s categories (Kaiser, Kronfeldner, and Meunier 2014), it seems to us that while it is perhaps likely to involve collaboration, this is not necessary. Integrative interdisciplinarity arises where two disciplines perceive each other as so highly relevant that there is often, or always, reason from one to engage with the content of the other. That may, of course, be as a kind of collaboration, but it need not be.

The other two of Boden’s three forms of interdisciplinarity are, however, clearly collaborative. If we take shared and co-operative interdisciplinarity for granted, PF, as we conceive of it, is a kind of co-operative interdisciplinarity. It typically requires active adjustment and engagement, as solutions need to be substantively incorporated into the source discipline. Moreover, we think the PF model and the various aspects of it we have highlighted here can support collaborative efforts in practice in at least two ways. Drawing attention to the way different disciplines are complementary and interdependent assists in fostering a spirit of both interdisciplinary exchange and epistemic humility that is conducive to collaboration. It is also positive because it makes clearer, and to an extent creates, the conditions conducive to exchange.

6. Concluding Remarks

In this paper we have discussed in some detail the nature and prerequisites of PF as a model for interdisciplinary research. First, drawing on previous work on the concept, we laid out a schema showing the PF process. Second, discussing the concepts of a problem and a discipline, we brought out some of the basic reasons why this kind of interdisciplinary exchange can be challenging. Third, in the main part of the paper we outlined the prerequisites of PF. We focused on three dimensions of disciplines. The first we examined was ontological-conceptual. It has been suggested that so-called interfield theories are needed here, but we argued that less seems to be required in order to facilitate the exchange of problems. Modelling templates seem to do. The ontological-conceptual dimension is probably the most difficult to capture, since underlying intuitions and ideas about the nature of science come into play here. In seeking to explain why an episode of PF appears to work, a realist might propose that if the interfield theory, concept or modelling template obtains sufficient purchase on the underlying phenomenon, a baseline condition of PF is in place. The instrumentalist, on the other hand, may emphasise other things, and bracket any consideration of the ontological basis. Minimally, however, there must be a recognition of mutual relevance that is then reflected somehow in the very exchange of problems and solutions. Along with the epistemological-methodological dimension, we highlighted two central points, for here the options seem fewer and more straightforward. Either an explicit, shared epistemological framework is developed—if it does not already exist—in which the admissibility of solutions is independently assessed in both disciplines or the relationship rests on some degree of (mutual) trust.¹⁰ Finally, what we labelled the axiological dimension revolves around the influence of various non-epistemic values operating within disciplines. We concentrated on the role of non-epistemic values in setting the priorities and aims of disciplines, and the way such values come to play a role in methodological choices and the acceptance of hypotheses. Two important points were made here: (a) that the exchange itself, i.e. the interdisciplinary work, often needs to be valued to provide an impetus to adjust priorities and interests, and (b) that as a result of value entanglement there may be different ways of reaching the same goals. This last point seems to generalise. These different dimensions also interact in ways that make it hard to provide additive criteria of when the conditions for PF are in place. Finally, we argued that PF seems to provide an example of non-integrative interdisciplinarity and therefore sits somewhere between multidisciplinarity and interdisciplinarity in the standard taxonomy.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by Formas [grant number 2020-00202].

Notes

1 What it means for a problem to ‘belong’ to a discipline is itself not straightforward: see, e.g. Popper (1963, 88) and Thorén (2015a) for discussions.

2 As a reviewer of this paper pointed out, the fourth step also will involve some transformation and translation if the solution is to be expressed in terms that make sense in discipline A. If this is not possible, bilateral problem-feeding cannot be achieved.

3 Occasionally, and perhaps often, the solutions are more or less present in the source discipline, and the entire challenge lies in the problem transformations. We thank an anonymous reviewer for pointing this out.

4 See Thorén (2015b, Chapter 3) for a longer discussion of these kinds of relationship. See also Wahlberg and Persson (2017) and Persson, Johansson, and Olsson (2018).

5 Some, such as Bechtel (1986), have drawn a distinction between fields and disciplines. Here, however, we have followed Darden and Maull and not applied this distinction.

6 Another idea that comes to mind here: a boundary object (Star and Griesemer 1989) is such that it is flexible enough to be adaptable to different contexts, yet rigid enough to ‘maintain identity across sites’ (393). Its function is to mediate communication between otherwise disparate ‘social worlds’. Boundary objects can be both concrete things like maps and abstract things like concepts. The main challenge with this idea is that it seems entirely descriptive. In other words, it is merely part of the definition of a boundary object that it aids the exchange of information in this particular way, and thus we fail to obtain an idea of how that happens beyond the stipulation regarding flexible and rigid parts.

7 Issues of these kinds have become increasingly central not only in the philosophy of science but elsewhere. Of note here is the science of team science. This has developed since the turn of the century and is devoted to empirical and theoretical study of collaborative and integrative scientific approaches from a wide range of perspectives (see e.g. Hall et al. 2018, 2019).

8 There is no particular reason to think that all problem-feeding interactions are, or indeed should be, symmetrical in this respect.

9 It is common to differentiate interdisciplinarity from conventional disciplinary research in that the former is problem-solving or problem-oriented or problem-driven while the latter is (supposedly) not (Schmidt 2022). On its face, this way of drawing the distinction has some drawbacks. As Schmidt (2011) points out, many philosophers including Karl Popper and Larry Laudan have considered science in general to be primarily a problem-solving activity. Nevertheless, PF clearly is about problem-solving and should therefore be of relevance to the idea that interdisciplinarity is problem-solving. For example, a case can be made for the view that what is meant by ‘problem orientation’ and similar constructions is that science as such, as well as its individual disciplines, should be more sensitive to what is considered to be problematic outside the disciplinary or scientific context. Specialisation, which in the interdisciplinarity literature is frequently portrayed as a main challenge (Brewer 1999), is then the process of generating one’s own problems. The PF model may help here by offering some guidance on the constraints and challenges with transferring and sharing problems between contexts where different values and practices dominate.

10 As a reviewer of this paper pointed out, trust and shared epistemic standards are rather different things, at least in certain ways. One way to think about this is that the trust situation actually implies a kind of shared framework, albeit a pluralist one acknowledging the contextual nature of epistemic standards. Another reaction, as the reviewer suggested, is that trust engenders the kinds of social process that are needed to develop a shared framework.