Values in Science: Five Challenges to Normative Accounts. Emma Harry Baitz. Master of Arts in History and Philosophy of Science (by Research)

Transcription

1 Values in Science: Five Challenges to Normative Accounts Emma Harry Baitz Submitted in total fulfilment of the requirements of the degree of Master of Arts in History and Philosophy of Science (by Research) February 2015 Department of History and Philosophy of Science Faculty of Arts University of Melbourne 1

2 Abstract Normative accounts of the proper role of values in science seek to regulate the influence societal values can have on scientific inquiries. Until recently, the most prominent normative account of the role of values in science was the value-free ideal, which seeks to restrict the influence of societal values during the justificatory phases of investigative practice. In the last few decades, the ideal of value-free science has been criticised by philosophers who doubt both the possibility and desirability of value-free science. Some have offered alternative normative accounts. The purpose of the thesis to follow is to expose limitations faced by alternative normative accounts of values in science. I focus primarily on Douglas (2009) prominent normative account. I will identify what I take to be five independent, though interrelated, issues that pose a challenge to normative accounts. Existing normative accounts are affected to varying degrees by the limitations I identify. The aim of the thesis is to point to issues that persist within prominent normative accounts. The first chapter focuses on the distinction between epistemic and non-epistemic values. Many normative accounts attempt to do away with this distinction. I examine the motivations for abandoning the distinction. I conclude that the distinction is problematic, though its abandonment reduces the scope of application of normative accounts. I focus on Douglas account, and argue that in abandoning the epistemic/non-epistemic distinction, Douglas constructs a normative account whose application is limited to only those scientific inquiries that have significant nonepistemic consequences, and which ignores important differences between the sciences. In this sense is fails to be generally applicable. I conclude with a defence of a modified version of the epistemic/non-epistemic distinction, based on Steel s (2010) criticisms. The second chapter focuses on the prevalent division between internal and external stages of science. I identify several problematic assumptions associated with the use of the internal/external distinction. I argue that if we continue to assume that the ostensibly external stages of inquiry are epistemically uninteresting, we risk ignorance of an important avenue of value-influence. The third argument I put forward is that many normative accounts do not adequately deal with value influence during heuristic appraisal. I argue that heuristic appraisal is under-recognised as an avenue of value-influence because it is assumed to be a function of the external stages of inquiries. I demonstrate that heuristic appraisal involves both epistemic and non-epistemic considerations, and argue that normative accounts ought to supply the guidelines for legitimate value influence here. In the fourth chapter I characterise the conceptual framework of inquiries as an important source of value-influence. The concepts underpinning inquiries are often deeply hidden, and do not 2

3 feature in the analysis of value-influence on the inquiry. I provide examples of normative components of conceptual frameworks that have influenced the methodoogy and conclusions drawn from inquiries. I argue that it is under-represented in current normative accounts, and requires further analysis. The fifth chapter argues that demarcating legitimate from illegitimate value influence during the analysis of case studies can be ambiguous and interpretive. Given their intepretive nature, I ask what work descriptive case studies can do for the values in science debate. I argue that the best information we can get from case studies is post-hoc, and context-specific. I conclude that while they may have heuristic value, case studies cannot be used to definitively test normative accounts. The neglect of these relevant and important avenues for value influence shows that extant normative accounts typically have a very narrow focus on epistemic justification. I believe that the debate needs to be broadened; that it must widen its focus to incorporate elements of scientific practice other than that pertaining strictly to the justification of theories. Acknowledgements My deepest gratitude to my supervisors, Dr Kristian Camilleri and Dr Gerhard Weisenfeldt for the invaluable support, feedback, advice and guidance they have given me throughout the duration of this project. Declaration This is to certify that: i. The thesis comprises only my original work towards the Masters except where indicated in the preface ii. Due acknowledgement has been made in the text to all other material used, iii. The thesis is less than 40,000 words in length, exclusive of tables, maps, bibliographies and appendices Signed 3

4 4

5 Table of Contents Abstract... 2 Acknowledgements... 3 Declaration Introduction The Problem of Wishful Thinking The Value Free Ideal Motivations for Rejecting the Externality Model of Value-Free Science Alternative Accounts Thesis Overview Conclusion The Epistemic/Non-Epistemic Distinction The Epistemic/Non-Epistemic Distinction in the Externality Model of Value-Free Science The Non-Neutrality of Epistemic Values Doing Away with the Epistemic/Non-Epistemic Distinction Steel s Rejuvenation of the Epistemic/Non-Epistemic Distinction Conclusion The Internal/External Distinction Defining the Distinction The Stages of Science Epistemic warrant Conclusion The Context of Pursuit The Narrow Epistemic Focus of Normative Models Heuristic Appraisal and Normative Accounts of Values in Science The Role of Heuristic Appraisal in Science Values in the Context of Pursuit Multiple Goals and the Context of Pursuit Conclusion Value-Ladeness of Concepts The Influence of Non-Epistemic Values on Concepts Conclusion Problem of Case Studies

6 6.1 A Case Study in the Values in Science Debate Value Influence in the Case Study The Use of Case Studies in the Values in Science Debate Conclusion Conclusion Narrow Epistemic Focus Retrospective Analysis of Values in Science Implications, Limitations and Further Questions List of References

7 1 Introduction 1.1 The Problem of Wishful Thinking The term the problem of wishful thinking (used by Brown, 2014) describes a situation wherein one is permitted to draw scientific conclusions on the basis of societal or ideological values, rather than on the basis of evidence. Any account of the role of values in science is required to produce, first and foremost, a way of avoiding the problem of wishful thinking. On this, all philosophers agree. Those who support the notion of value-free science, and those who reject it, agree that values may not function as a justification for a scientific claim. Philosopher Heather Douglas (2000) characterises the heart of the problem of wishful thinking: the argument I want X to be true, therefore X is true, remains a bad argument, both within and without science (p. 578). If we allow societal values to act as justification for scientific claims, our science would be unacceptably relativist, subjective, and would lack the epistemic authority for which it is so highly prized. Most will agree that good science is not merely a product of wishful thinking in this manner. Yet, most will also agree that science is not, even ideally, totally free of influence from values. Some elements of research are laudably value-laden. For instance, we restrict the use of human experimentation within inquiry on moral grounds; this constraint on methodology is explicitly underpinned by values. Societal values of this nature may influence many aspects of inquiry, such as funding decisions, or the application of scientific research. Some values are inexorable; for instance, personal values will influence the choice of research question made by the individual scientist. Scientists hold the acquisition of truth and reliable knowledge as a value. Values influence of this nature is widely regarded as both necessary and harmless. To rid science of the influence of values completely is a crude impossibility. Allowing values to influence science without restriction would, however, result in the problem of wishful thinking. For this reason, it is a task of philosophy of science to create rules or legislation about the proper role of values; a normative account of the influence of values in science. The question about the role of values in science therefore becomes, in fact, many different questions. What kinds of values legitimately influence scientific inquiry? In what way do values legitimately influence decisions within inquiries? Which scientific processes are legitimately influenced by values? Normative accounts seek to answer these questions in order to construct legislation about acceptable and unacceptable value-influence within scientific inquiries, and are typically, primarily concerned with potential value influence during the processes by which we establish grounds for belief in a theory. The warrant for belief is taken by many in the debate to be based, ideally, on purely epistemic factors. Those involved in the debate are concerned to avoid basing 7

8 theory acceptance and scientific beliefs on non-neutral or non-objective grounds. The primary motivation for the value-free ideal, and likewise for those offering alternative normative models of values in science, is to avoid the problem of wishful thinking and secure a rational, objective basis for belief. The purpose of the thesis to follow is to expose limitations faced by normative accounts of values in science. I will argue throughout the thesis that there are some challenges that afflict value-free and alternative normative accounts equally. I will identify what I take to be five independent, though interrelated, issues that pose a challenge to normative accounts. At the end of this introduction I will adumbrate these challenges more specifically. This introduction will supply a conceptual and brief historical background to the main thesis. It functions mainly as a survey of the territory. I offer this survey to give the reader the philosophical context for the arguments to follow in my thesis. The thesis functions primarily as a constructive critique of the alternatives offered to a value-free normative account of scientific inquiry. Alternative normative accounts may borrow certain elements from the value-free ideal, and as such some of the criticisms I make will bear upon value-free normative accounts. The focus of the thesis is on alternative accounts. The alternative accounts are not equally afflicted by the issues I present. Certain accounts accommodate some of these issues and not others. For this reason, I discuss each of the issues separately, identifying and critically analysing their influence on alternative normative accounts. I focus primarily on the alternative normative model offered by Heather Douglas (2009). I turn now to a description of the normative model of value-free science. I will go on to outline some factors and arguments that have motivated its recent rejection. Following this I will introduce the Douglas (2009) alternative account that will constitute a focus for my thesis. In the chapters to follow I will identify common limitations shared by disparate positions within the debate, and attempt a conceptual clarification of some of the issues that I believe pervade the debate. 1.2 The Value Free Ideal Avoiding the problem of wishful thinking isthe primary motivation for the value-free ideal. The language, and many defences of the value-free ideal have been inherited from the work of logical empiricists of the Vienna Circle 1. The philosophers of the Vienna Circle were concerned largely with scientific rationality, and the sanctity of objective, neutral, scientific bases for belief. Most 1 Proctor (1991) notes in his discussion of the history of value-neutrality that the origin of the value-free ideal was not developed as a single notion, but borne of a collection of loosely associated ideals that emerged at different times to serve different functions. He identifies elements of the modern values in science debate throughout the history of philosophy, even as far back as Ancient Greece, Many scholars point to the the 17th century distinction between facts and values as a starting point of the value-free ideal. I choose to begin at the Vienna Circle because this is the most complete instantiation of modern valuefreedom and thus makes a good starting point for the contemporary discussion of the role of values in science. 8

9 members of the Vienna Circle were committed to a doctrine of ethical non-cognitivism, believing value statements to be always and only emotive, and therefore not proper candidates for truth claims 2. This position is perhaps best exemplified in the work of A.J. Ayer (1936). Because value statements can be neither true nor false, they were considered to be outlawed from scientific inquiry altogether. The logical empiricists were very concerned to scrutinise the basis of theory choice, and the warrant for belief in scientific inquiry. Philosopher Don Howard (2009) argues that in their support of the ideal of value-free science, the logical empiricists were driven partly by the worry that allowing values and motives a role in theory choice risked opening the door to dogma, prejudice, and other species of unreason (p. 203). This preoccupation with the epistemic sanctity of theory choice and belief continues into the current debate, and in the following chapters I will identify the philosophical consequences of this narrow epistemic focus. For now, I want only to point out that the motivation for the value-free ideal was, and continues to be, focused on protecting those stages that constitute grounds for belief from influence by subjective, societal values. Modern day advocates of value-free science, in whatever form, also argue for excluding values from scientific inquiry in the name of preserving rationality. It seems to be held among value-free advocates (e.g., Lacey 2013; Dorato 2004; Betz 2013) that to acknowledge and allow a role for values in the internal (i.e., justificatory) stages of scientific inquiry would undermine the attempt to acquire truthful, objective beliefs, leading to an unacceptably relativist science and losing all prospects of gaining significant knowledge (Lacey 1999, p. 216). The value-free ideal is underpinned by the belief that scientific practice ought to be independent of personal, social, and cultural values, and that the practice of science as scientists is independent of their subjective preferences regarding what ought to be (Longino 1983, p. 7). The notion of value-free science was borne of, and is maintained in, efforts to protect objectivity and still accommodate those areas of science where values can and do play a role. There are many different working definitions of objectivity (Douglas 2004), and many different ways to interpret the ideal value-freedom (Lacey 2010). The most prominent and longest-standing model of valuefree science is sometimes referred to as the externality model. The externality model of valuefree science forbids value-influence in the heart or core of science, which consists of the processes associated with the justification of scientific claims. This concept of a heart of inquiry persists in both value-free and alternative normative accounts (for some examples from recent literature from both sides of the debate see Douglas 2009; Kitcher 2011; Betz 2013; Lacey 2005; Anderson 2004). Importantly, the core of science is usually thought of as being constituted by 2 There were many disputes within the Vienna Circle on this point, and each individual had a slightly different view regarding the role and status of value claims. There was no unanimity on this (or likely any other) topic, and what I present here is a caricatured version. Their commitment to value-free science, in its varying forms, is however widely acknowledged and was put forward as their collective, public position. 9

10 the stages of data characterisation, interpretation and analysis, and subsequent hypothesis acceptance or rejection. Dorato (2004) describes the core of science as being constituted by socalled internal (that is, evidential knowledge-gathering) processes (p. 72). Rottschaefer (2003) offers a similar definition of the internal aspects of scientific inquiry such as its methods of justification, whose aim is to achieve such epistemic values as truth, prediction and explanation (p. 226). Douglas (2009) defines the heart of doing science as being comprised of the characterization of data, the interpretation of evidence, and the acceptance of theories (p. 102). These are the processes that proponents of value-free science seek to insulate from value influence. Many proponents of alternative accounts believe that there either is or ought to be value-influence during these processes. Both sides of the debate utilise this distinction, and philosophers on both sides agree that in external areas, such as deciding which scientific projects to undertake, Funding decisions, and which methodology to pursue (Douglas 2009, 98-99), value influence is pervasive and unproblematic. Because there is widespread agreement that the external processes are heavily and harmlessly value-laden, value-free idealists seek to protect only the internal processes of inquiry. So under this normative model of value-free science, only the internal processes are required to be kept free of values. Not all values are equally restricted, however. There are some values that are shared by the scientific community that can, under this model, legitimately exert influence and guide decisions during the internal stages of inquiry. These are epistemic values. These are the values that are taken to conduce to truth or rational belief (Longino 1996, p. 51), or that promote the attainment of truth (Steel 2010, p. 15) or that promote the acquisition of true beliefs (Goldman 1999). Kuhn (1977) introduced epistemic values with a list of five, and there have since been many more candidates for epistemic values. Kuhn s original list was comprised of empirical adequacy, predictive accuracy, simplicity, fruitfulness and explanatory breadth. To Kuhn (1977), these were the criteria of choice, the characteristics of a good scientific theory which provide the shared basis for theory choice (p. 76). Epistemic values are among the canons that make science scientific (Kuhn 1977, p. 76). They are those values that shape the rationality of science (Faye 2006, p. 125); that are the source of the rules determining what constitutes acceptable scientific practice of scientific method (Longino 1983, p. 8). According to the externality model of value-free science, these values can acceptably exert influence during the internal stages of inquiry. By contrast, non-epistemic values are, by definition, aimed at goals other than truth (Rottschaefer 2003, p. 230). These are societal, ideological or otherwise subjective values that value-free idealists fear could contaminate or otherwise taint scientific rationality. The normative model of value-free science utilises a distinction between internal and external stages, and epistemic and non-epistemic values. It therefore amounts to a claim that nonepistemic values are to be excluded from justificatory processes of inquiries. As Douglas (2007) 10

11 puts it: the term value-free science really refers to the norm of epistemic values only in the internal stages of science (p. 120). 1.3 Motivations for Rejecting the Externality Model of Value-Free Science While any normative model for acceptable value-influence in scientific inquiry is required to protect against the problem of wishful thinking, in recent years it has been argued that the externality model in the form described above is not the best route for doing so. It lacks the capacity to protect the neutrality of internal, justificatory processes of inquiry. It fails to entirely secure the rationality of science, and allows a certain degree of relativism to infect the process of inquiry. Philosophical inquiry into the role of values in science over the last four decades or so has provided the field with excellent reasons to doubt both the efficacy and desirability of the value-free ideal. Much of this critical work comes from feminist science and feminist philosophy of science. Here I address some of the major reasons provided for the rejection of the value-free ideal, in order to provide crucial background for the concepts deployed in the thesis to follow. Feminist scientists and philosophers of science went a long way towards exposing the potential for non-epistemic values to infiltrate even good science. A good exposition comes from the 1978 issue of the feminist academic journal Signs, called Women, Science and Society. This issue considered topics of gender bias in science. Stimpson and Burstyn s (1978) editorial to this issue begins with a passionate lambast of the surliness of responses to the new ideas about the role of values in science borne of feminist critiques, and characterises the most pernicious form of resistance to feminist philosophy of science as perfunctory acknowledgement, followed by refusal to incorporate such ideas into teaching and research (p. 1). The editorial ends with revolutionary zeal, its manifesto culminating with an objective no less than reforming the way we think about science, to avoid manipulating interpretations of nature to justify the more oppressive and spiteful impulses of human nature (p. 4). Many of the issue s critiques of science focussed on the research into the anthropological origins or biological explanations of sex difference. Here, Adrienne Zihlman discussed the validity of the anthropological notion of manthe-hunter in theories about social organisation in early hominids; Donna Haraway wrote on conceptions of gender and social dominance in primatology; Helen Lambert analysed the presumptions of different scientific theories of sex differences and Ruth Bleier wrote about bias in the biological and human sciences. Bleier (1978) pointed out that then-current research in sociobiology served the function of justifying and validating oppressive norms through the assumption of evolutionary and biological determinism. The research, she writes is flawed by the assumption of male investigators, who view animal behaviour... in the light and language of their own experiences, values and beliefs (Bleier 1978, p. 159). 11

12 Feminist scientists and philosophers of science had exposed a situation of grave significance, and heralded the death of value-free science as a received notion. Feminist philosophy of science was set to reveal what the logical empiricists could not see that as much as we strive for valuefreedom in science, it is not and never was a wholly value-free enterprise. The light and language by which we interpret nature colours both the process of inquiry and its result, and is so deeply ingrained that it can be extremely difficult to identify in either the action or consequence of scientific inquiry. By the early 1980s, value-informed science was taken to be all but inevitable, and the view from nowhere deemed fancifully naive (see Knorr-Cetina, 1981 or Knorr-Cetina and Mulkay, 1983 for good examples). Kourany (2010) recounts the consequence that feminist scientists concluded that a new and more adequate understanding of scientific objectivity was needed (p. 57). The value-free ideal that was attacked by these scientists and philosophers had been the dominant conception of normative models of scientific inquiry for decades. There were many reasons and examples offered to substantiate doubt in its efficacy and desirability. These arguments can be roughly classified in to two separate, but related categories: arguments from underdetermination, and arguments from inductive risk. I will introduce each of these in turn. The underdetermination argument suggests that since theory choice is always underdetermined by evidence, background assumptions are necessary to attach theory to evidence. These background assumptions may be non-epistemic, thus endangering the neutrality of the internal processes of inquiry by inviting non-epistemic values to determine the evidence relevant to hypotheses. Another form of argument provided for the rejection of the externality model of value-free science is the argument from inductive risk. This argument shows that non-epistemic values have a role in decision making within the internal stages of inquiry. Such values assist in determining the evidential thresholds for scientific claims, affecting the conclusions of a theory that is to be put forward as a candidate for belief. Inductive risk arguments show the involvement of non-epistemic values in the internal processes of science to varying degrees. These processes provide the grounds for scientific belief and as such, according to the externality model, their neutrality with regard to non-epistemic values is essential for objective, good science. Underdetermination arguments show possible routes for non-epistemic values to influence these internal processes. Inductive risk arguments aim to show that there are some instances where value involvement is required for good scientific reasoning; that values have a contributory role to play in good science, and are required within the process for sound reasoning. Details of the specific deployment of these arguments will come later in the thesis; for now I wish only to introduce them, to provide a conceptual context for the discussion to follow. Underdetermination 12

13 The Duhem-Quine underdetermination thesis relates the idea that no scientific hypothesis has, by itself, any observational consequences (Duhem 1954; Quine 1951) 3. Hypotheses have empirical consequences only when conjoined with auxiliary hypotheses or background assumptions about the world. Background assumptions, according to this argument, are needed to form evidential relations between empirical observation and scientific hypotheses. Strong underdetermination arguments of this sort might be referred to as global underdetermination, and is deployed in opposition to the value-free ideal most notably by Kitcher (2001). I narrow my focus here to transient underdetermination, rather than the global underdetermination of the Duhem-Quine thesis. The reason I restrict my focus to transient underdetermination arguments is that this weaker version is sufficient to undermine the value-free ideal (see Howard 2006; Nelson 1990; Potter 1996 for arguments in support of this). 4 Transient underdetermination states that some theories, hypotheses or models are underdetermined by the evidence that is currently available. Transient underdetermination theses provide a potential role for non-epistemic values within scientific inquiry in constituting the background assumptions and auxiliary hypotheses required for constructing the crucial link between theory and observation in the event that a claim is underdetermined by the currently available evidence. Helen Longino (2002) embraces this role for non-epistemic values, characterising their function as a rich pool of varied resources, constraints, and incentives to help close the gap left by logic (p. 128). Lynn Hankinson Nelson (1990) is another proponent of this sort of role for non-epistemic values, arguing that there is no way to rule out the possibility of non-epistemic values acting as auxiliary hypotheses. Elizabeth Anderson (2004) is also a proponent of this sort of underdetermination argument, explaining that various background assumptions could be legitimately selected for any reason (p.2), so there are no logical or methodological proscriptions against choosing moral or political values to constitute background assumptions. Laudan and Leplin (1991) also utilise this argument from underdetermination in their efforts to legitimise the role of non-epistemic values. Underdetermination arguments are thus frequently utilised in order to provide a legitimate role for non-epistemic values in the internal aspects of scientific inquiry, and consequently deny the functionality of the value-free ideal. 3 It is widely accepted now that Duhem and Quine were espousing different arguments of underdetermination. Duhem s claim was restricted to physics, and argued that no hypothesis can be sufficiently isolated from its background assumptions in order to test it (1954). Quine argued a stronger version of underdetermination that encompassed all human knowledge, arguing that there will always be multiple hypotheses that are consistent with all the evidence we have at any one time (1951). The Duhem- Quine thesis that I utilise in this section refers to the idea that no single hypothesis, taken by itself, has any observational consequences. 4 See Biddle (2013) for a survey of the literature on transient underdetermination and its application to the values in science debate. 13

14 Many alternative accounts of values in science rely on the underdetermination thesis. They relegate value influence to auxiliary or background assumptions that underpin scientific inquiries and argue that their influence here contributes to good science. Longino s (1990) contextual empiricist alternative normative model of values in science makes such a claim. Longino s is a values-of-the-gaps philosophy insofar as she is talking not bout scientific claims or or their justification, but about the factors that inform scientific choices beyond considerations of evidence. Of contextual empiricism, Longino s alternative to value-free science, she writes: evidential relevance of data is secured... by background assumptions, with the consequence that the same data can in different contexts serve as evidence for different hypotheses (Longino 1996, p. 39). This is a classic invocation of the argument from underdetermination, which places values in the gaps, somehow extraneous to, yet bearing influence over, the heart of scientific inquiry. Longino is concerned with the influence of subjective preferences on background assumptions and hence theory choice (Longino 1996, p. 40), rather than with the justification of scientific claims and theory choice in terms of empirical evidence alone. Janet Kourany s Socially Responsible Science model is also, to an extent, a values-of-the-gaps argument. Kourany s project encourages scientists e.g., through funding initiatives to choose research programs that promise support for egalitarian goals. And second, the feminist project encourages those same scientists e.g., through the provision of epistemic values supportive of egalitarian goals as well as a choice procedure that favors egalitarian options in cases of underdetermination to pursue their research as far as empirically possible in ways that maximize that promise. (Kourany 2003, p. 10) These two concrete requirements of the socially responsible science project paint value-influence as broadly external to the justification of scientific claims, though exercising deep influence over scientific claims via auxiliary assumptions and choices about pursuit. Kourany seeks to encourage the influence of societal values in the pursuit, from among the available possibilities, of only those systems that support, or most support, egalitarian goals (2003, p. 10). Decisions about pursuit are usually considered as external to scientific inquiry in some important sense, despite its great influence on the product of scientific inquiries. Kourany s account therefore invokes a values-of-the-gaps type underdetermination argument in her normative account of the proper role of values in science. The externality model of value-free science denies that there are any necessary or legitimate roles for non-epistemic values in the internal processes of inquiry, in forming the warrant for belief or acceptance of theories. Although arguments from underdetermination highlight the possible role for non-epistemic values via auxiliary assumptions, they do not show that such a role is necessary (Intemann 2004). Underdetermination arguments show only that auxiliary assumptions may determine the relationship between theory and evidence, but dictate no necessary involvement of non-epistemic values in determining what these auxiliary assumptions are. Underdetermination 14

15 arguments merely show the potential involvement of non-epistemic values. Rottschaefer (2003) takes issue with the notion that auxiliary hypotheses are necessarily influenced by the processes of discovery. He notes that there are no arguments to support the claim that non-epistemic values influence the justification of auxiliary hypotheses legitimately. The consequence of these objections is that in order to utilise these varieties of the underdetermination thesis to undermine the value-free ideal, one has to do more than prove the possibility that non-epistemic values might operate in science through the background assumptions used to connect theory and evidence. The value-free ideal makes the claim that non-epistemic values have no legitimate role to play in justification of a theory. Critics of the ideal must therefore show not only that non-epistemic values might possibly have a role to play in scientific justification, but that their possible role is legitimate, and contributory to good science (Intemann 2004). We find an argument for the legitimate influence of non-epistemic values as auxiliary assumptions in the next argument discussed, Douglas (2000; 2009) argument from inductive risk. The strongest interpretation of arguments from underdetermination is represented by the normative claim that non-epistemic values provide us with good reasons with which to judge the epistemic weight of a theory, and that in the absence of this function for non-epistemic values, the reasons one may adduce for theory acceptance are incomplete (Douglas 2009). The interpretation of data, the choosing of a theoretical framework, the preference or privilege granted to a particular epistemic value over another, and the adoption of a standard of evidence, may all constructively rely on the influence of non-epistemic values. Douglas (2009) re-interpretation of the role of non-epistemic values in scientific inquiry seeks to salvage this necessary influence of nonepistemic values without exacerbating the problem of wishful thinking. The argument from underdetermination shows that the externality model of value-free science is untenable- nonepistemic values may function as auxiliary or background assumptions in forming crucial relationships between theories and relevant evidence. The argument from inductive risk, to which I now turn, shows that the externality model of value-free science is undesirable as an ideal- nonepistemic values are needed to execute rational decisions in the internal stages of inquiry. Inductive Risk Heather Douglas is the main contemporary proponent of the argument from inductive risk. Douglas (2000) argues that non-epistemic values are needed to weigh the consequences of the possible errors one makes in accepting or rejecting a hypothesis (p.562). This is the argument from inductive risk, which refers to the setting of the epistemic standard for belief or acceptance of a hypothesis, based on the significance ascribed to the non-epistemic consequences of error. The argument is based on the notion that all scientific claims or proposals carry the possibility of error. Scientists are therefore required to assess the potential consequences of this error, in an 15

16 effort to achieve the optimal balance between false positives and false negatives. For instance, in assessing the toxicity of a new drug, the optimal balance between false positives and false negatives is likely to hinge on the non-epistemic value of public health. A false positive in this instance would result in the belief that a drug is safe for consumption when it is, in fact, dangerous. A false negative would see the drug declared unsafe, when it is in fact safe. The nonepistemic consequences of distributing a drug that is actually unsafe for consumption are dire for those who consume the drug. You might sympathise with the scientist who sets the optimal balance between false positives and false negatives at a lean toward false negatives. Better to be needlessly cautious in this instance than reckless, given what is at stake. This intuition is based on the value of public safety, and illustrates the potential of this value to influence the internal process of characterising data. If it is shown that non-epistemic values can or should be involved in the internal stages of inquiry, this constitutes a denial of the externality model. Determining the ideal balance is a function of an internal stage of inquiry, and requires the influence of nonepistemic values. Between 1948 and 1954, C. West Churchman (1948), Philip Frank (1953) and Richard Rudner (1953) utilised the argument for inductive risk to criticise the ideal of science as value-free. Douglas (2000) draws on this work and uses the argument from inductive risk to attack the normative standard of value-free science. In setting standards for statistical significance, the optimal level of inductive risk can be partially determined by the non-epistemic consequences of error, which in turn is calculated with consideration of non-epistemic values. The setting of thresholds and standards of significance, Douglas argues, are functions of the internal processes of inquiry. Douglas argues, therefore, that the infiltration of non-epistemic values via considerations of non-epistemic consequences results in the violation of the internal, justificatory processes by non-epistemic values. Non-epistemic values are required by the scientist in this case in order to make responsible and rational decisions regarding the setting of thresholds. Douglas makes both a normative and descriptive claim that scientists do and ought to consider nonepistemic values where non-epistemic consequences hinge on their decision. She concludes that the argument from inductive risk overthrows the externality model. The argument from inductive risk legitimises the role of non-epistemic values in the internal aspects of science, and thereby undermines value-free science as a laudable or tenable ideal. Consideration of non-epistemic consequences highlights the need to incorporate non-epistemic considerations in the internal aspects of science. 1.4 Alternative Accounts Underdetermination arguments and arguments from inductive risk expose the externality model of value-free science as being largely decrepit, and in need of either replacement of rehabilitation. It fails as a descriptive claim because non-epistemic values do play a role in the internal, 16

17 justificatory processes of inquiry, in determining the standards for thresholds and statistical significance, and forming auxiliary assumptions by which we determine evidence to be relevant to a hypothesis. It fails, also, as a normative claim. We need to utilise non-epistemic values in setting thresholds and standards of statistical significance, in order to rationally accept or reject scientific claims. Alternative accounts or amendments have been offered to accommodate these objections. Douglas (2009) proposes a normative account to rival the externality model, in an effort to accommodate the objections mentioned in the preceding sections. In this account, she proposes to do away with the distinction between kinds of values (i.e., epistemic/non-epistemic), which she takes to be a leaky and unreliable distinction, and instead distinguish between the roles values may play at different stages of inquiry. The motivation for abandoning a distinction between kinds of values will be made clear in the first chapter of this thesis. For now, I wish only to give a brief overview of Douglas normative account that restricts all values equally. Douglas (2009) utilises a distinction between direct and indirect roles for values in science: In the first direct role, the values act much the same way as evidence normally does, providing warrant or reasons to accept a claim. In the second, indirect role, the values do not compete with or supplant evidence, but rather determine the importance of the inductive gaps left by science (p. 96) To avoid the problem of wishful thinking the influence of values is restricted within the internal, justificatory processes of inquiry. In the external processes, values may act directly, as reasons in themselves to accept a claim, providing direct motivation for the adoption of a theory (Douglas 2009, p. 96). In the direct role, values determine our decisions in and of themselves, acting as stand-alone reasons to accept a claim (Douglas 2009, p. 96). For instance, values can act directly in the external stage of study design. Moral values can directly determine, by themselves, the limitations on human experimentation that are placed on methodological designs. A scientist may legitimately cite any manner of non-epistemic values as having motivated their choice of research question; another example of values acting legitimately in an external and direct role. During the internal processes, values may only act indirectly, in assisting scientists in their evaluation of the potential consequences of erroneous choices. In the internal processes, values: serve a crucial role of helping us to determine whether the available evidence is sufficient for the choice and what the importance of uncertainty is, weighing the potential consequences of a wrong choice and helping to mitigate against this possibility by requiring more evidence when such consequences are dire. (Douglas 2009, p. 97) 17

18 This legitimises their role during considerations of inductive risk. During the internal processes, a value serves as a reason not to accept or reject the current level of uncertainty not as a reason to accept or reject the options per se (Douglas 2009, p. 97). They cannot act directly in these stages, serving independently as motivation for a claim or as the sole reason for a choice. Douglas alternative normative account is centred on the distinction between direct and indirect roles. Values acting in an indirect role may saturate science without threat to its objectivity or rationality. There are other normative accounts that serve as alternatives to the value-free ideal, but none so complete as Douglas. As Douglas shares motivations with these alternative normative accounts, much of my critique will be directed at Douglas account. Many of the challenges I offer in this thesis will be applicable to varying extents to those normative accounts that I do not outline in detail here. Throughout the thesis I will note those normative accounts that do better at accommodating the challenges I will raise. In particular, I will make mention of Kourany s (2003) socially responsible science, Longino s (1990) contextual empiricism and Kitcher s (2011) wellordered science, though these challenges bear, to varying degrees, upon other normative accounts and their critiques. The thesis to follow will focus on theoretical tools and elements I believe to pervade the values in science debate, and is thus applicable, to varying degrees, to many normative accounts. 1.5 Thesis Overview At this point, one may begin to recognise some uncomfortable assumptions and limitations that persist across both the accounts I have described. The aim of my thesis is not to answer questions as to whether or not science is, or can be, value-free. Nor does it offer a normative account of the role of values in science, or defend any extant accounts. My thesis looks instead at the problems that afflict both value-free and alternative normative accounts of the role of values in science. I conclude that the debate is narrowly focused on a caricatured and severely limited conception of scientific inquiry, failing to accommodate many facets of actual scientific practice. I argue that the debate needs to broaden its focus to incorporate more of the philosophical resources, issues, arguments and opinions raised in general philosophy of science, that it may apply more broadly to the practice, rather than the constrictive theory, of scientific inquiry. The first chapter examines the distinction between epistemic and non-epistemic values. As I have mentioned, the externality model of value-free science places this distinction centrally, whereas Douglas alternative account dismisses it entirely. Douglas account privileges empirical adequacy and predictive accuracy as basic standards or criteria of good science. With this move, Douglas ensures that the values empirical adequacy and predictive accuracy may directly determine theory acceptance or rejection. All values other than these two, whether they are 18

19 epistemic or non-epistemic, are restricted from exerting direct influence within the justificatory stages of inquiry. I argue in this chapter that distinguishing between kinds of values is useful. Without the distinction, a normative model cannot accommodate the differences between the sciences. In different modes of inquiry and across different sciences, epistemic values are weighed and traded against one another. What counts as an epistemic value in one context may not be considered as important or virtuous in another. There could be an instance in which a value counts as an epistemic value in one context does not count as an epistemic value in another at all. For this reason, Douglas normative model, with its criteria of good science cannot apply uniformly across all the sciences. Doing away with the distinction results in a restriction of the account s scope of applicability. Yet, embracing the distinction (with important alterations that I will explicate further within the chapter) comes with its own issues. While its scope of application is unaffected, it raises questions as to the justification for a value s status as epistemic. I argue that this can be done only retrospectively, and that a uniformly applicable normative account is restricted in its application to retrospective analysis. The treatment of the epistemic/non-epistemic distinction therefore poses a general problem for normative accounts of values in science. This issue intimates a question more fundamental than the treatment of epistemic values itself: Is a temporally general and universally applicable normative account of values in science possible? The second chapter focuses on the ubiquitous and pervasive distinction between internal and external stages of inquiry. This distinction, as noted earlier, persists throughout both value-free and alternative normative accounts in the debate. The internal aspects of scientific inquiry are taken to be those processes that provide the grounds for belief in scientific claims. They are characterised with relative uniformity across the debate, as the characterisation and interpretation of data, and theory acceptance or rejection. In slightly older parlance, they might be identified as the constituent stages of the context of justification. In this chapter, I identify issues with the assumption that we can carve science up into neatly discretised stages, and that the grounds for belief are provided exclusively by what is taken to be internal stages. Moreover, I argue that the dismissal of external stages as epistemically uninteresting is both harmful and misleading. I demonstrate in this chapter that those stages deemed external do contribute in important ways to the grounds for belief. From these arguments, I propose that we ought to disentangle the notion of grounds for belief from that of internal stages. If one cannot neatly discretise the stages of science from one another, this weakens the normative accounts of values in science whose legislation depends on such stages. Moreover, the arguments in this chapter for the epistemic significance and import of the external stages imply that important value influence has gone unnoticed and unanalysed in the debate. This demonstrates an aspect of the narrow focus of extant normative accounts on the internal, or justificatory 19

20 processes of inquiry, and its subsequent lack of application to many important areas of valueintrusion. The third chapter argues that both value-free and alternative normative accounts of the role of values in science maintain a narrow focus on the epistemic appraisal of theories. In this chapter I examine the role of values under heuristic appraisal- the appraisal of theories (or problematics, techniques, or models) with a priority not of empirical adequacy, but of expected utility or promise. I argue that values function differently under heuristic appraisal than they do under epistemic appraisal. From this argument, I urge that normative accounts ought to widen their focus to those areas and modes of inquiry that do not have to do strictly with epistemic appraisal. This chapter draws heavily on the preceding chapter, in which I demonstrate that the debate is blind to important areas of inquiry, deeming them epistemically uninteresting. The debate retains a narrowly epistemic focus that is damaging to its scope of application. Heuristic appraisal would be deemed external under the internal/external schema that pervades the debate, and hence dismissed as epistemically uninteresting. In this chapter I go about demonstrating the epistemic significance of heuristic appraisal and the complex interaction of epistemic, social and personal values during this mode of appraisal. I identify that heuristic appraisal can have epistemic consequences and does utilise epistemic factors, and as such I urge the incorporation of considerations of heuristic appraisal into normative accounts of values in science. The fourth chapter of the thesis continues along this theme, focusing on another area neglected by current normative accounts; namely, the concepts that facilitate and enable scientific inquiry. I argue that these concepts, though themselves so deeply fundamental to inquiries as to be themselves beyond epistemic justification, are an important avenue for value-intrusion. Values do not act here to justify a scientific claim, but deeply influence, and sometimes even directly determine, the kind of claims being made. Because of their lack epistemic justification, current normative accounts do not analyse the potential role of values here. The narrowly epistemic focus of the debate again allows important value-influence to go unanalysed. The deeper implication of the value-ladeness of concepts is that we may not be able to legislate about this role of values at all. Normative accounts focused on epistemic matters do not extend to this fundamental level, and indeed because of the importance of the conceptual framework of inquiries, it is very difficult to legislate about or against value-involvement at this level. The fifth and last chapter of the thesis focuses on the use of case studies within the debate. I argue that radically different conclusions may be drawn from the same case study regarding the legitimacy of value-influence during that historical episode. I argue that the use of case studies appeals to information that is, at best, not wholly reliable and, at least, partially interpretive. I identify the potential to beg the question by interpreting case studies in a way friendly to a 20