1. Philosophy, Disciplines and Interdisciplinarity

Philosophy is the oldest academic discipline, but where philosophy should be placed within the space of disciplines remains elusive. In the essay that inspired this special issue, Bernard Williams (Reference Williams and Moore2006b) is sharply critical of what he describes as scientistic approaches to philosophy, which assimilate philosophy to a model of science exemplified by physics or chemistry. In place of scientism, he offers a vision of philosophy as a humanistic discipline.

Williams says little about what it is for a discipline to be humanistic, but takes history to be a central exemplar. Humanistic inquiry is pursued, in Moore’s elucidation ‘by humans, from their unique position in the world, to make sense both of themselves and of that position’ (Moore Reference Moore, D’Oro and Overgaard2017: 45). Humanistic inquiry focuses on forms of sense-making, and on articulating in human terms the meaning of texts, institutions, norms, events, and artefacts through interpretation, criticism and argument. In Williams’s words, viewing philosophy as a humanistic discipline is to view it ‘as part of a wider humanistic enterprise of making sense of ourselves and our activities… in order to answer many of its questions it needs to attend to other parts of that enterprise, in particular to history’ (Williams Reference Williams and Moore2006b: 197).

Not all forms of inquiry are, or should aspire to be, humanistic. Williams was sympathetic to the motivations that lead some to hope that science will be able to attain to what he described as an absolute conception of reality. It is a virtue of physics that it aspires to (and has succeeded in) providing an understanding of the world, which abstracts to a very great degree from our ‘own perspective, our peculiar and local ways of apprehending things’ (Williams Reference Williams and Moore2006b: 185). However, he was sharply critical of the assumption that abstracting away from specifically human forms of sense-making will invariably improve our cognitive grip on ethical and political questions. As Williams argued, ‘in seeking to understand ourselves–we need concepts and explanations which are rooted in our more local practices, our culture, and our history, and these cannot be replaced by concepts which we might share with very different investigators of the world’ (Williams Reference Williams and Moore2006b: 185).

While humanistic scholarship remains crucial for self-understanding, the longue durée has seen an apparently inexorable decline in the prestige of the humanities in relation to the sciences, and a continued narrowing in the scope of the questions that are taken to be best answered with humanistic means. The result has been a profound shift away from the humanities and towards science, technology, engineering and mathematics (STEM) in the intellectual resources used to educate children and ready them for the workforce, and in prevailing assumptions about which disciplines bring the greatest social benefit. A sense of this long-term decline is already present in Williams’s text, and in the 25 years since then, there has been an ever-growing discourse on the decline of the humanities. The view that the humanities are in crisis has become ubiquitous even amongst their defenders.Footnote ¹

How should philosophy position itself in this changing intellectual landscape? One obvious move is for philosophers to reframe their discipline as a science – a move that Williams disparages as ‘scientism’. Scientism, on Williams’s account goes beyond a philosopher’s ‘interest or involvement in science’, and involves a ‘misunderstanding of the relations between philosophy and the natural sciences which tends to assimilate philosophy to the aims, or at least the manners, of the sciences’ (Williams Reference Williams and Moore2006b: 182). Williams is unusual in directing his concerns about scientism towards conceptions of philosophy, rather than, as is more often the case in philosophical discussions of scientism, the encroachment and/or overvaluation of science more broadly (Ridder, Peels & Woudenberg Reference de Ridder, Peels and van Woudenberg2018). Hence, unlike some who critique scientism, Williams does not disparage either the aims or methods of science. His goal is limited to providing a principled account of why questions relating to human self-understanding are not best interpreted as questions that can be answered using scientific means.

Williams presents two main objections to scientism in philosophy. First, philosophical writing needs to be reflective, and that reflection requires styles of writing in which the reader’s sense of what is important is engaged, rather than ignored or kept at a distance. Second, philosophy (or at least those parts of philosophy Williams is most interested in) is about sense-making. Sense-making is to a significant degree interpretive, historically situated and local. For both reasons, importing scientific methods, or assumptions that philosophers should adhere more closely to scientific methods, ends up obscuring or misinterpreting philosophy as an activity, rather than shedding light on it.

Williams’s wide-ranging essay takes for granted the intellectual landscape of the end of the twentieth century, and makes some deep and provocative claims about what philosophy is and how it should be done. My response starts from the somewhat changed intellectual landscape of the mid-2020s, is similarly wide-ranging, and also aims to say something novel and provocative about how philosophy should be done. I engage closely with what Williams says, but with the aim of thinking about what philosophy now should be, rather than providing an interpretation that takes up every strand of Williams’s argument or is definitive even in the interpretation of those elements I do discuss.

I am sympathetic to Williams’s reasons for reasserting the importance of the humanistic in philosophy, but will question some of the assumptions about disciplinarity that seem to shape his argument. In particular, I argue that it is a mistake to think that there is an overall answer to the question whether philosophy belongs with the humanities or the sciences. Rather, the philosophical methods that it is most fruitful to adopt depend on the research question to be addressed (a point that Williams’s own philosophical practice showcases, even if his metaphilosophy makes this less clear).

Williams’s objections to scientism are most persuasive in cases where the domain to be investigated is, at least to a significant degree, constituted by and through human self-understandings. Williams, of course, thought that value inquiry – whether in ethics or political philosophy – is such a domain. However, this is a point on which Williams and his moral realist critics sharply disagree, with both sides at times thinking the other’s position not only wrong, but hopelessly confused and obscure.

Williams’s views about the nature of ethics both support, and are supported by, his broader humanistic conception of philosophy, while other philosophers’ views of ethics as being akin to science (or to mathematics) both support and are supported by their distinctive conceptions of philosophy. This might lead to the worry that there is something question-begging about Williams’ argument. I shall suggest that appearances may be deceptive: to the extent that there are deep-seated and apparently intractable disagreements about the ontology and epistemology of ethical values, this is evidence in favour of Williams’s view rather than his moral realist opponents. I argue that it is helpful to view such disagreements through the lens of a transdisciplinary epistemology, in which incompatible methodologies and ontologies, and how to reconcile them, are central themes.

Looked at from this perspective, philosophy is an outlier: no other discipline suffers such a mismatch between the level of confidence of its competent practitioners that there are correct and highly general answers to its research questions, and the power of the methods adopted within the discipline to resolve disagreements between these same competent practitioners. Philosophers working in value inquiry would do better to be more honest about philosophy’s limits and more modest in what they expect it to do. Looked at in this way, it is Williams’s humanistic view of ethics that better comports with the totality of the evidence. I close by reconsidering the role of cases and thought experiments in ethics, and arguing that a suitably refined version of Williams’s position in ethics has much in common with Dewey’s pragmatism.

2. The Humanities and the Sciences

Williams (Reference Williams and Moore2006b) presupposes that his reader will agree that there is a clear distinction to be drawn between the methods and aims of the humanities and those of the sciences. Twenty-five years later, this seems less obvious. In the interim, work in many humanities departments has been transformed by the mass digitisation of documents and artefacts, and vastly increased computational power and storage. Completely new disciplines such as Digital Humanities have been created. Sense-making remains crucial for the humanities, but it’s now generally agreed that this sense-making may include as integral parts a range of methods which might previously have been thought of as distinctively scientific, such as data mining and model construction.

Williams was alive to differences in methods, approaches and ontologies among the sciences, which is one reason why, notwithstanding his gestures towards the absolute conception, his naturalism is non-reductionist (Williams Reference Williams2002b: 23). However, Williams (Reference Williams and Moore2006b) does not consider how the evident differences among the sciences complicate his critique of scientism in philosophy, both in theory and in practice. Considered theoretically, there is no consensus among either working scientists, or philosophers of science, that science requires us to commit to the absolute conception of the world, or that it is reasonable to expect scientific inquiry to converge on a single unified model (Dupré Reference Dupré1993; Chang Reference Chang2022). It’s now mainstream within scientific practice to acknowledge and to reflect on the fact that, even where science aims to improve understanding of objects that exist entirely separately from human beings, scientific inquiry is a thoroughly human endeavour (Douglas Reference Douglas2009; Kitcher Reference Kitcher2011). Social norms, institutional structures, and governance are recognised to be crucial in maintaining flourishing scientific communities of inquiry, in which levels of research misconduct are minimised, and structural factors that advantage some social groups over others addressed. So importing forms of scientific reasoning into philosophy need not presuppose implausible reductionism, inattentiveness to relevant differences, or even neglect of features such as social power.

Thinking more practically, it is evident that palaeontologists view the world very differently from physicists, just as ecologists differ in their methods and ontologies from electrical engineers. Would it amount to scientism on Williams’s account to import any methods or styles of writing from any science, or does the worry apply only to certain methods and certain modes of writing (e.g. those that aim to model themselves on those used in physics)?

My point here is not to resolve summarily these complex debates in the philosophy of science, but to explain why distinguishing the humanities from the sciences is more difficult than Williams’s paper might lead us to think, and to indicate how this affects the nature and cogency of his complaints about scientism. Williams famously ridiculed the distinction between analytical and continental philosophy as a cross-classification, ‘as though one divided cars into front-wheel drive and Japanese’ (Williams Reference Williams, Bunnin and Tsui-James2002a: 23). For all that has been said so far, someone could object that the distinction between the humanistic and the scientific may also involve such a cross-classification. Whether a discipline or topic area is a humanistic one is determined on Williams’s account by whether its subject matter is forms of human sense-making. However, whether an area of scholarly activity should be classed as a science is determined not by its subject matter (and certainly not by its subject matter being something other than human sense-making), but by features of the methods used. While detailed discussion of scientific methodology is beyond the scope of this article, uncontroversial features include that the methods allow competing interpretations or hypotheses to be tested with agreed ways of determining which are superior, and that the methods are apt to generate both reasoned convergence over time among competent inquirers, and an expanding body of knowledge (Hansson Reference Hansson and Edward2021). Whether a discipline or area of scholarly activity can be both humanistic in so far as it aims to shed light on forms of human sense-making, and scientific in that it is apt to generate reasoned convergence and an expanding body of knowledge, is a substantive question rather than one that can be determined by stipulation.

3. Disciplines and Research Problems

Williams recognises that his essay’s title could be interpreted as offering an answer to which grouping of disciplines we should take philosophy to belong to, but he claims that his real aim is to contribute to philosophy’s self-understanding as a discipline, by asking ‘what models or ideals or analogies should we look to in thinking about the ways in which philosophy should be done?’ (Williams Reference Williams and Moore2006b: 180). There is nonetheless a slight tension between what one might expect Williams’s answer to be to this question (given the title of his essay), and his acceptance that ‘some work in philosophy… quite properly conducts itself as an extension of the natural or mathematical sciences’ (Williams Reference Williams and Moore2006b: 182), and hence is not humanistic.

Williams’s language is oracular enough that, on the face of things he could be interpreted as making the strong claim that philosophy should draw its models, ideals and analogies from the humanities, except in the special case where its subject matter requires drawing on scientific models; or the much weaker claim that philosophy should draw on humanistic models, ideals and analogies to the extent that particular philosophical problems are, at base, humanistic questions about sense-making; or one of an indeterminate number of claims of a strength between these two extremes.

Williams (Reference Williams and Moore2006b) acknowledges that a good writer must have a keen awareness of the most likely potential misunderstandings, misinterpretations and objections to their position, and that such awareness will function as scaffolding during the writing process. Nonetheless, he does not think that making all this explicit in published work helps philosophy. Just as good building practices remove the scaffolding once the building is complete, good philosophical writing provides readers with an argument that can be appreciated without going through the explicit enumeration of a swathe of objections and potential misunderstandings which might be raised only by a reader who was ‘malicious… or clinically literal minded’ (Williams Reference Williams and Moore2006b: 183). Williams suggests a potential link between scientism and the mindset in analytic philosophy that he opposes, which ‘seeks precision by total mind control, through issuing continuous and rigid interpretative directions’ (Williams Reference Williams and Moore2006b: 183).

Given these features of Williams’s style, arguing with Williams – whether agreeing or disagreeing with him – frequently requires extracting from his texts claims that are more determinate than anything Williams himself explicitly commits to.Footnote ² My interest here is less in forcing Williams’s views on the nature of philosophy into a form more determinate than he chose to express them, than to argue for a conception of philosophical inquiry in which it is the research problem which should determine the method (rather than the discipline). If we adopt this more flexible approach, there is no reason to think that there is a single answer to where philosophy as a discipline ‘fits’.

Whole books have been written on how to define research problems, and what I say will be brief and programmatic. The world does not contain research problems ready to be discovered. Much of the creativity and skill in the early stages of a research project is devoted to the process of problematisation, whereby an inchoate situation is brought into sharper view as a problem. As Dewey put it,

Research problems can, but need not, be framed in ways that fall squarely within the domain of a single discipline, for example, being answerable only by the methods characteristic of that discipline. Research problems may also be multidisciplinary (requiring separate input from multiple disciplines that is later combined), or interdisciplinary (requiring the methods of at least two disciplines to be in dialogue in the problematisation, design and carrying out of the research), or transdisciplinary (framed in a problem-oriented way, and where solving the problem often requires unifying or reconciling existing disciplinary methods into bespoke forms).

There are a wide variety of interdisciplinary and transdisciplinary research questions that bring philosophy into dialogue and collaboration with a range of other disciplines and real-world stakeholders. As a result, we might ask whether it is sensible to assume that there is (or should be) a single answer about where philosophy ‘fits’ within the grouping of disciplines. Many areas of philosophy, including philosophy of law, philosophy of biology, philosophy of education, philosophy of art, and philosophy of medicine, reflect on the philosophical questions and challenges that arise for scholars and practitioners within another academic discipline. Scholars who contribute to the literature in these areas are often not based in philosophy departments and will frequently have had extensive training in the first-order domain as well as in philosophy. Most philosophers of law have law degrees in addition to their philosophical training and are also based in law schools. Many bioethicists and philosophers of medicine trained as healthcare professionals and are based in medical schools. As a result, the prevailing norms for what counts as good work, and appropriate methods, in these different areas of philosophy are influenced by features that are unique to the first-order discipline and are frequently interdisciplinary.

Philosophy can contribute to making progress on different interdisciplinary and transdisciplinary research problems in different ways: good work in the philosophy of physics may look nothing like good work in the philosophy of literature. We should not expect all the breadth of contemporary philosophy to fit comfortably within a single grouping of disciplines.

One possible response to such a line of thinking would be that these interdisciplinary forms of philosophy tend to be bad philosophy. However, this isn’t the line that Williams takes either in his pronouncements about philosophy, or his own practice. ‘Interdisciplinary’ and ‘transdisciplinary’ don’t seem to have been words that Williams used in his published writing, but his such descriptions would seem to be an appropriate way to describe his methods in Truth and Truthfulness. His later work often draws attention to the ways in which philosophical reasoning, to answer its research questions, needs to be informed by other disciplines such as history (Williams Reference Williams2002b) or law (Williams Reference Williams and David1997). While my approach advocates for a research problem led, rather than a disciplinary led approach to philosophy, it’s less an objection to how Williams does philosophy, than a friendly amendment to how he talks about doing philosophy.Footnote ³

4. Ethics, Sense-making and Vindicatory Explanations

Williams argues that scientists and philosophers view the history of their disciplines very differently, and that their attitudes reveal divergent assumptions about progress in their disciplines. The history of science is often not taught as part of science curricula, and working scientists generally assume that they can safely ignore much of the history of their discipline, because there is a vindicatory explanation of how earlier scientific theories were superseded by later ones:

Conversely, scarcely any moral philosophers think that the insights contained within texts such as Aristotle’s Nicomachean Ethics or Kant’s Groundwork of the Metaphysics of Morals have been assimilated or superseded in a way that makes it no longer useful to return to these texts. Williams notes that there is a tension between the thought that ethical theory has made solid and consistent progress over its history (and hence that there is a vindicatory explanation to be had of the ethical positions we now adopt), and the idea that the history of philosophy is an indispensable element of the discipline. He takes it that the continued relevance of such classic texts to contemporary debates in ethics implies that the story that should be given of how philosophical debates have evolved is ‘not notably vindicatory’ (Williams Reference Williams and Moore2006b: 190).

It is one thing to argue that the history of ethical theorising has not so far been notably vindicatory, another to claim that it could not become vindicatory. Someone could accept that ethical theory has made relatively little progress so far, but argue that there is every reason to think that it could do, given the sustained use of better methods. Thus, Parfit ends Reasons and Persons by noting how recent it is for ethics to be undertaken as an academic inquiry entirely outside of the limitations imposed by religious traditions, arguing that given its novelty, ‘we cannot yet predict whether, as in Mathematics, we will all reach agreement’ (Parfit Reference Parfit1984: 454). Williamson argues in a similar vein that lack of progress isn’t inherent to philosophy as a discipline; the problem has been that philosophy has not been done in a way that is precise enough to ensure progress (Williamson Reference Williamson, Greenough and Michael2006: Reference Williamson2007).

Williams argues that such expressions of hope – based on the assumption that method in philosophy could be assimilated either to mathematics or to the sciences – expresses a misunderstanding of the problems that value inquiry exists to examine. Value inquiry – whether in ethics or in political philosophy – is from Williams’s perspective a form of sense-making. There are important in principle limits on the extent to which we should expect inquiries that focus on sense-making to lead to convergence on a single truth, and limits too on the extent to which any convergence that does occur can be given a vindicatory explanation.

Williams makes this argument most fully in Truth and Truthfulness, in respect of history. The argument there is striking for the fact that it begins from premisses that are favourable to realists. Williams accepts that there are facts and plain truths about the past, and he spends much of the book providing a vindicatory genealogy of accuracy and sincerity as the twin virtues of truth. He allows that we not only can but should criticise failures of accuracy and sincerity such as misreading a document, deliberately downplaying relevant information, or being mistaken in what an argument is saying. Moreover, he argues that some events, such as the invention of historical time unambiguously embody intellectual progress (Williams Reference Williams2002b: Ch. 7).

Williams nonetheless argues that there are principled reasons for thinking that there will not and cannot be final convergence in historical scholarship, even when all those within an academic debate exhibit the virtues of truth. The combination of the facts and the intellectual virtues provide real constraints, but they systematically underdetermine what could count as a truthful and accurate way of making sense of a particular set of events. This is because what makes sense isn’t something that is fixed for all time, but reflects the needs of particular audiences at particular times:

As a result, there isn’t anything that even in theory could be uncovered or presented as the truth about the past (Williams Reference Williams2002b: 258).

Ethics and political philosophy are also fundamentally about sense-making on Williams’s account, and this sense-making, just as in historical scholarship, is systematically underdetermined by the constraints posed by factual accuracy and sound deliberation. This is the idea at the heart of Williams’s account of internal reasons, according to which an agent can have a reason to perform some action $\phi $ only if this reason could figure in a correct explanation of what they in fact do (Williams Reference Williams1981: 102). Individuals thus can only have reasons that they can make sense of (Williams Reference Williams2002b: 237). The reasons that make sense to them are those that are currently within their subjective motivational set, and those could be reached from it via a deliberative route. As Miranda Fricker argues, Williams’s account of sense-making, his account of reasons, and much else in his philosophy can be brought into focus by the realization that his most fundamental philosophical commitment is to ethical freedom, namely that:

Clearly, if it is true that we do have ethical freedom in this sense, this has implications for how we should think about methodology in ethics. Even if there are other areas of philosophy in which precise articulation of claims, and rigorous argumentation could plausibly lead not only to a consensus, but to a vindicatory explanation of this consensus, we misunderstand the nature of ethical inquiry if we expect this. While we can make progress by our own lights in ethical reflection, there is no final goal to which such attempts must tend. Deploying highly general ethical theories, which abstract from forms of local sense-making, may make our ethical judgement worse rather than better.

As a result of his commitment to ethical freedom, Williams had an abiding concern with the stability of our ethical views under reflection, and how to reconcile ourselves to the impossibility of the kind of vindicatory account of ethical principles and practices that philosophers have often thought it was the main purpose of ethics to provide. Williams came to think that the way forward is to acknowledge the contingency of our current values, and to revise the conception of philosophy that makes it seem that a vindicatory explanation of our values is either possible or required:

This line of reasoning is, however, barely intelligible for many moral realists (as Parfit Reference Parfit2011: 542–543 discusses), who simply reject the claim that that vindicatory explanations in ethics are impossible, and so have no use for sophisticated analyses of what we should do to overcome the sense that such explanations are needed. Viewing ethics as a sense-making activity through which we contend with the implications of our ethical freedom helps us to see how Williams’s characteristic commitments hang together in a unified way, but it does little to prevent the critic from contraposing the argument. Perhaps it is because Williams mistakenly assimilates philosophy to history, rather than drawing on models inspired by science or by mathematics, that he gives inadequate attention to moral facts, or to the role of the a priori in ethical thinking.

5. The Meta-problem

In trying to adjudicate between rival research paradigms, which interpret ethics as engaged in very different kinds of inquiry, we face what I shall call the meta-problem. The meta-problem is at the heart of all interdisciplinary and transdisciplinary inquiry, and can be phrased as: how should we combine or reconcile different modes of disciplinary inquiry? Answering the meta-problem in any specific instance requires us to reflect on the phenomena that are constructed as research objects within each discipline or research programme, and the assumptions made within such constructions. This will include the rigour and the suitability of the methods used in the discipline or research programme for investigating the phenomena, and whether competing disciplines or research programmes are different but complementary, or are disagreeing, or simply talking past one another.

The meta-problem becomes pressing where inquiries undertaken by competent inquirers about the same domain of phenomena lead to different framings of research questions, different standards for success, and different results. Where (as in the case of Williams and his moral realist critics) there are incompatible research programmes within a discipline, with no obvious or agreed to criteria to determine which is most fruitful, the meta-problem must be faced within disciplines too.

Philosophy is one discipline among others. Some philosophers (particularly those who have little knowledge of other disciplines) are tempted by the idea that philosophical inquiry is more fundamental than inquiry in other disciplines. Such philosophers draw the comforting (and self-serving) conclusion that philosophers enjoy the asymmetrical freedom to critique forms of reasoning in other disciplines without needing to take seriously the ways in which philosophical research programmes fall short by other disciplines’ standards. However, once we bring the meta-problem squarely into focus, it is hard not to notice that philosophy is an outlier among disciplines for a different, and less flattering reason, namely that (to put the point in a slogan) philosophical research programmes tend to be ambitious without being progressive.

As I use the term here, the ambition of a research program is a vector of its generality, and the precision of the answers it provides to its research questions. A research programme is more general the greater the range of phenomena to which it is designed to apply: the laws of physics aim to be highly general, whereas theories in psychology may aim to explain more specific phenomena such as the effects of priming on research subjects’ judgements. The least general theories will be idiographic theories, aimed at (in Windelband’s words) ‘the complete and exhaustive portrayal of a particular, more or less protracted occurrence of a unique, temporally circumscribed reality’ (Windelband 1998 [Reference Windelband1894]: 12).

A research programme is precise to the extent that it enables researchers to make determinate and specific claims about the phenomena that fall within its scope. Precision should be distinguished from accuracy. A theory can be very precise yet wildly inaccurate (for example, Bishop Ussher’s chronology deduced that the world started on 23^rd October 4004 BC). A theory can be very imprecise yet accurate (for example, if it predicts that there will be between one million and one billion citizens of the USA next year).

A research programme is progressive to the extent that its history is vindicatory in Williams’s sense. Neither increasing generality nor increasing precision will automatically make a research programme more progressive. In the case of generality, what matters is whether the phenomena of interest in fact have the kind of regularity or structure needed for general theories to be informative and useful. Thus, if (as Windelband argued) historical events are in fact unique, it will make historical scholarship worse rather than better if historians attempt to invoke universal laws within historical explanations (Donagan Reference Donagan1964). In the case of precision, what matters is how accurately and reliably inquirers are able to test claims: making claims whose precision significantly outstrips our ability to test their accuracy is usually to be engaged in bluster rather than responsible scholarship.

The back-and-forth of philosophical discussion – even when conducted at the highest levels of sophistication and rigour over an extended timeframe – rarely leads to decisive reasons to endorse one theory over another. It is widely acknowledged among philosophers that in each area of philosophy, multiple incompatible research programmes will remain in play for the foreseeable future. As a result, David Lewis famously argued that philosophical progress consists in developing more refined versions of the theories we started with, and a better understanding of the price to be paid (in accepting implausible implications in some circumstances) for maintaining each particular position (Lewis Reference Lewis1983: x; Beebee Reference Beebee2018).

Despite a common agreement among philosophers that the power of the methods adopted within the discipline to resolve disagreements between philosophers is weak (and hence that philosophy is at best weakly progressive), philosophers’ level of confidence that the discipline should be organised around the construction of ambitious theories remains high. For example, some prominent philosophers working in ethics claim that the ethical principles they advance should guide actions not just here and now, but in all possible worlds at all times, while simultaneously acknowledging that other philosophers not only reject their particular position, but think they have decisive arguments against it, or even find it unintelligible.Footnote ⁴

No other discipline exhibits such a discrepancy between its level of progressiveness, and the ambition of the theories advanced. Thus, in anthropology, low progressiveness is matched by low generality. Anthropologists use ethnography and stress that even when competently practised, it will give partial and defeasible insights into some community at some time (Geertz Reference Geertz1983). Chemists make more ambitious claims (though notably only about elements and compounds in this world, and under a range of conditions), but those more ambitious claims are buttressed by an ability to form precise hypotheses and to test them much more definitively than is ever the case in philosophy, and to use such testing to build and refine theories. The mismatch between the ambition of philosophical theories, and the lack of power of the methods used by philosophers to determine which answers are correct, is one of the reasons why multidisciplinary funding panels often struggle to be persuaded of the value of philosophy proposals (Lamont Reference Lamont2009).

Our analysis using the tools of transdisciplinary epistemology suggests that the debate between Williams and moral realists is not deadlocked after all. Sustained mismatch between the level of ambition and the level of progressiveness provides reasons to reduce ambition. Bringing the generality and precision of the claims made within value inquiry more closely into line with the ability to test them will allow for a more reliable way of attaining whatever level of progress turns out to possible. The sociologist of science Robert Merton recommended a similar move within social theory seventy years ago, arguing that sociologists do better when they shift away from totalising theories that attempt to explain everything in society, to what he described as theories of the middle-range, which are deliberately limited in their range and applicability (Merton Reference Merton1968: 39). While Williams does not use the language of middle-range theory, we can see a related set of concerns at play in his insistence on the importance of history for philosophy: ‘philosophy itself must involve more than abstract argument… it must engage itself in history. In this as in other respects, philosophy cannot be too pure if it really wants to do what it sets out to do’ (Williams Reference Williams2002b: 39).

The success of a middle-range theory depends on how useful it is for the kinds of research questions we want to answer, rather than on whether it has shortcomings in some conceivable circumstances (which it inevitably will). What matters is the theory’s adequacy for the class of cases that we actually do face or are reasonably likely to face. We might think of designing middle-range theories as like designing tennis rackets; we want to make sure that the theory’s sweet spot (with maximum insight) is as large as possible, and that the theory is feasible to wield. Just as we would not think a racket deficient if it proved useless in getting a cannonball or a grain of couscous over the net, so we may also be unworried if a middle-range theory would have implausible implications in circumstances widely different from those we in fact face.

Rigorous science aims to provide a reasoned account of the confidence we should have that a particular explanation is correct, given the limitations of the data. Calibration, understanding the limitations of measurement devices, and accounting for potential confounding factors are central. Straining for a greater degree of precision and generality than the evidence allows makes reasoning less, rather than more, rigorous. The research problem that a middle-range theory aims to solve is much more determinate than the research problem which a more abstract and general theory targets, and it is often easier and clearer to determine whether a particular middle-range theory is good enough to serve its purpose. As a result, middle-range theories are often more rigorous than more abstract and general ones. Good science in disciplines such as medicine or data science often involves doing middle-range theory.

Middle-range theory is often also more useful for the purposes of understanding and improving our current situation than more ambitious theory. It can be more useful, as it can be better tailored to the problem to be explored. For example, if we are interested in how to fairly distribute scarce health resources within a publicly funded health system, a middle-range theory of health justice is highly likely to be more useful than a completely general theory that attempts to provide principles for the allocation of any goods in any circumstances. As there are features of health as a good that are distinctive (as well as distinctive features of the institutions and social roles through which we protect and promote health), a highly general level theory of the distribution of goods will either be too abstract or give results that are inappropriate in context (Wilson Reference Wilson2023).

I acknowledge that the claim that middle-range theory is likely to be more rigorous may seem counterintuitive. Philosophers often assume that a philosophical theory cannot be correct if someone can produce a successful counterexample to it, and that it does not matter whether the counterexample is a real-world or a highly counterfactual one. Hence, philosophers tend to look down on middle-range theory: there are just so many obvious ways that the world could have been, which would render a middle-range theory that works for us now inadequate, were they to obtain. However, it is the combination of the commitment to a very high degree of generality in theories, with a view of what it is for a theory to be good enough that is so demanding that no philosophical theories in fact meet it, which creates the mismatch that we looked to middle-range theory to help solve.

The next section combines this account of middle-range theory with Williams’s account of analysis of sense-making, to analyse the role of thought experiments in ethical thinking. Section 7 concludes with a discussion of how Williams’s approach to ethical freedom opens a way in which the arts can inform ethical thinking, which Williams hinted at but was explored much more fully by Dewey.

6. Thought Experiments, Science and Fiction

Philosophers often describe thought experiments in ethics as if they are doing science. Thus, Kamm explains her method as follows: ‘Real-life cases often do not contain the relevant–or solely the relevant–characteristics to help in our search for principles. If our aim is to discover the relative weight of, say, two factors, we should consider cases that involve only these two factors, perhaps artificially, rather than distract ourselves with other factors and options’ (Kamm Reference Kamm1993: 7). An approach has developed that is often known as the method of cases, which consists in presenting thought experiments, gauging one’s own and other philosophers’ intuitions about them, and using this to drive progress in the testing and construction of ethical theories.

Williams also used thought experiments, and some have become classics. However, in advancing and discussing these thought experiments, Williams clearly took himself to be doing something other than science. In line with his humanistic conception of philosophy, Williams uses his thought experiments to draw attention to aspects of sense-making, and to problems of interpretation. For example, in his discussions of George the Chemist, and Jim and the Indians (Smart & Williams Reference Williams1973), he is at pains not to use the cases as a way of deriving a straightforward judgement or counterexample. Rather, he is interested in the kinds of ethical considerations that arise in cases such as these, and why it is that utilitarians find it difficult to account for them. In ‘The Self and the Future’, Williams (Reference Williams1970) presents two different thought experiments about body swaps and explores how features of the narration and framing make a difference to our reactions to the cases.

Thus, even at the more specific level of the role of cases, profound disagreement between Williams and other thinkers remains. Are thought experiments in ethics best understood as a way of doing science, or do they form part of humanistic inquiry? As I argued, such a blunt categorical question is unhelpful. There is no reason why it can’t be fruitful in some circumstances to treat cases as scientific experiments, and in others to think of them as literary narratives. So better questions would be informed by transdisciplinary epistemology: what are the kinds of research questions in ethics that can fruitfully be answered by thought experiments, and what are thought experiments’ limitations even where they are useful? Last, can narratives or artworks provide kinds of ethical insight that philosophical argument cannot, and if so, how can these insights be integrated into philosophical thinking? I’ll consider each question in turn.

Thought experiments are unparalleled in their ability to raise points of principle, in at least a somewhat concrete form. In so doing, they allow fine distinctions to be drawn between different kinds of cases, and allow sustained attention to be brought to bear on whether and if so how, these differences are ethically relevant. To that extent, the analogy with scientific experimentation is apt. However, good science requires rigorous attention to experimental design, and to the limits of the power of any particular experiment to rule out competing hypotheses. It also requires the humility to recognise that even where an effect can reliably be shown in a controlled environment, this may tell us little about how to respond to a similar but more complex real-world problem.

Scientific experiments in general face two different challenges, which are conventionally known as internal and external validity. Internal validity relates to whether the experiment tests what it is supposed to be testing in a clean and unbiased way; for example whether adequate weight has been given to mitigating biases and confounding factors, or whether the sample size is large enough to detect the expected effect size, and whether the experiment and its results are reproducible. External validity relates to the question of how transportable a result is: even if an experiment does reliably establish an effect in the experimental context, and hence is internally valid, it is often far from clear what its implications are for the real world (Cartwright & Hardie Reference Cartwright and Hardie2012).

Doing experimental science with creatures as complex as human beings is hard. Even in cases such as drug development, where teams of hundreds or thousands work to design an intervention and then test it in a randomised clinical trial, it’s rare rather than common that rigorously conducted scientific experiments lead to a drug that works in the experimental environment without unacceptable side effects. Even where an experiment is internally valid and results can be reproduced in an experimental environment, there are deep and persistent problems of external validity. Experimental science that takes human beings as its subject matter will typically lead to middle-range, rather than to more ambitious theory.

As a result, conceiving of thought experiments in ethics as a way of doing science may support rather than undermine reasons for shifting to middle-range theories. Thought experiments have unfortunately not received anything like the same systematic approach to their design, methodology and real-world extrapolation as is found in other disciplines, and literature on how to achieve external validity in thought experiments is rare (Wilson Reference Wilson2016; Wilson Reference Wilson2021: Ch. 3). The analogy with science may make thought experiments seem less, rather than more, reliable when we look across to the seriousness with which methodology is taken in other sciences.

Thought experiments themselves are short pieces of fiction that the reader needs to make sense of. As Williams’s account of narration in history made clear, sense-making is local and to a significant degree relative to the interests of the interpreter. This is even more strongly the case for fictions. The interpreter’s response to a work of fiction (and artworks more broadly) is always a creative one, in which the interpreter completes the work. Following Gombrich (Reference Gombrich2004) and Seth (Reference Seth2019), I’ll call the latitude that artworks offer the interpreter the Beholder’s Share. All fictions have a Beholder’s Share, but some have more than others. The more that a fiction requires the beholder to do to complete the work in interaction with their imaginative explorations, the greater the Beholder’s Share. There are, of course, better and worse ways of making use of the Beholder’s Share, but there isn’t one single correct way: it doesn’t function like a Sudoku puzzle, where what’s left out can be deduced in its entirety from what’s been presented.

The Beholder’s Share not only poses a methodological challenge for achieving internal validity in thought experiments,Footnote ⁵ but also presents an opportunity for rethinking what we want from a good case in ethics. Rather than aiming to construct only thought experiments that function like precisely calibrated machines, each testing one and only one point of principle, we could also think of cases as works of fiction that the reader responds to creatively in the light of their own interpretation. If, like Williams, we are persuaded that we have ethical freedom (and that hence that the point of ethics is not convergence on a set of unchanging moral facts), we may come to think that it is fecundity of response, rather than homogeneity, that is the sign of a good case.

7. Ethical Inquiry, Ethical Freedom, and the Arts

Williams’s love of the arts is well known, and his papers are peppered with examples from opera and literature, thereby drawing into the ambit of philosophical reasoning cases that have the resonance, depth and ambiguity that is characteristic of good art.Footnote ⁶ Nonetheless, given the mostly critical tone of Williams’s engagement in ethics – destroying certainties rather than building anything new – his use of such examples serves mostly to undercut the pretensions of ethical theory. Williams, of course, accepts that the thick concepts and exemplars by which we navigate ethical life are local and human created, and would agree that the arts play a crucial role in articulating and developing novel modes of valuing.

Nonetheless, Williams gives little more than scattered hints about how the arts might play such a positive role. What he does say is perhaps expressed most clearly in his writing on opera, where he stresses ‘the power of music to express feeling that cannot be fully captured in words, a power which underlies the special opportunities that opera possesses as a form of drama’ (Williams Reference Williams2006a: 9). More specifically, Williams suggests that Janáček’s operas provide an exemplar of how to reconcile reflexivity with wholeheartedness (which for Williams is the central problem of modernity):

Other philosophers of ethical freedom have proposed accounts that explore the ethical potential of the arts much more fully and provide a good complement and partial corrective to Williams. Dewey’s Art as Experience provides an extended argument that the kind of insight we can get from the arts into how to live is importantly different from what science or philosophical argument can offer, and complementary to it:

Dewey thus argues that taking art seriously has the potential not just to act as an adjunct or extension to traditional philosophical methods in ethics, but to fundamentally challenge them by allowing us to ‘concentrate and enlarge an immediate experience’ (Dewey 2006 [Reference Dewey and Boydston1934]: 277–278). Dewey thought that philosophers too easily slip into the assumption that it is philosophy that gets to hold art, and immediate experience, up to philosophy’s standards. They forget that philosophy may look equally as questionable from the perspective of art or of immediate experience. As we have seen, adopting a transdisciplinary epistemology, there’s no reason to think that philosophy should have an asymmetric ability to criticise other forms of insight and methods of inquiry, while itself being immune from critique from these perspectives.

Williams referred to Dewey in only one of his published writings, his review of Rorty (Reference Rorty1982), and even there did so only in passing (Williams Reference Williams1983). Elsewhere, Williams took Rorty’s position to define pragmatism for the purposes of his critique (Williams Reference Williams2002b: 59–60, 128–30). It seems that Williams took himself to be opposed to pragmatism because he profoundly disagreed with what he took to be Rorty’s flippancy and subjectivism, rather than as a result of engaging with pragmatism as a broader intellectual tradition beyond Rorty’s idiosyncratic take on it. As a result, Williams may not have been fully aware that the pragmatist tradition encompasses both subjectivists such as James and Rorty and objectivists such as Peirce, C. I. Lewis, Ramsey, and Dewey. Williams’s anti-pragmatist arguments have some force against subjective pragmatists, but none against objectivist pragmatists, whose philosophical aims and projects often have much in common with Williams’s (Misak Reference Misak2021). Moreover, as Queloz explains, the overarching argument Williams makes in Truth and Truthfulness is best understood as pragmatist:

Thus, notwithstanding his protests to the contrary, there is a good case for taking Williams’s late work to be a contribution to the pragmatist tradition. Of course some elements of Williams’s overall perspective remain in tension even with a Deweyan pragmatist outlook – namely the absolute conception of reality, and his insistence that there is a fundamental difference between scientific and ethical inquiry (Williams Reference Williams2008 [1985]: Ch. 8). However, even here, there are reasons to think that Williams could have increased the coherence of his position if he abandoned these elements.

Williams (Reference Williams1978) introduced the terminology of the absolute conception of reality, and he continued to use this terminology until the end of his life, though he seems to have shifted both his attitude towards the absolute conception, and perhaps also his understanding of what it entails.Footnote ⁷ In Williams (Reference Williams1978: 197), the absolute conception of reality appeared as a challenge: knowledge requires us to presuppose the absolute conception, but we need to rethink it free from Descartes’s conception of certainty. Williams (2008 [Reference Williams1985]) defends the legitimacy of perspectival knowledge alongside the kind of non-perspectival knowledge associated with the absolute conception, and argues that striving for non-perspectival knowledge in ethics is mistaken. In his late work, the role of the absolute conception was diminished still further. Despite Truth and Truthfulness’s aim of providing a vindicatory explanation of sincerity and accuracy as the twin virtues of truth, the absolute conception is mentioned only once, in a footnote which says that it ‘remains controversial whether the idea of an absolute conception is coherent’ (Williams Reference Williams2002b: 295).

For Dewey, the extent to which inquiry within a particular domain can (or ought to) aspire to eliminate our ‘own perspective, our peculiar and local ways of apprehending things’ (Williams Reference Williams and Moore2006b: 185) is something to be clarified through the iterative process of inquiry, rather than something to be imposed on particular inquiries (whether into ethical or scientific questions) as a requirement from outside. For Dewey, inquiry is a form of activity, rather than a passive receptivity to how things are (Dewey 2008 [Reference Dewey, Boydston, Poulos and Nagel1938]). As I alluded to in Section 3, Dewey argues that inquiry has the same kind of structure across all domains, integrating experience (including aesthetic experience), reasoning and action. As a result, for Dewey, inquiry in ethics, science and art have much more in common than Williams allows: in each kind of inquiry, we act on the world on the basis of our hypotheses, and then feed the results of this action back into the process of inquiry through further rounds of observation, reasoning and action (Brown Reference Brown2012).

Inquiry on the Deweyan picture is iterative, in that each new inquiry is able to draw on the problematisation, observation, reasoning and judgements of previous inquiries, and generates resources that can be used in future inquiries. It is through iterated inquiry that we move from a position of fragmentary and unreliable understanding to one that is more coherent and reliable, a process that Hasok Chang describes as epistemic iteration (Reference Chang2022: 244–45). In some domains, epistemic iteration may push inquirers in the direction of eliminating what is peculiar to our local ways of apprehending things, while in others it may lead us to insist on maintaining the local and the contextual. Adopting this Deweyan account of inquiry might have allowed Williams both to bring more sharply into focus what it would take to vindicate the absolute conception, and also made more perspicuous to him that doing good science is not dependent on such a vindication.