1. ‘Death of evidence’ and the spectre of politicized science

Politicized science can be the subverting of sound science by outside influences as well as by political commitments of scientists. In recent decades, the US government under George W. Bush was more widely known for undermining science to further a pro-industry, anti-consumer protection, and anti-environment agenda (Mooney Reference Mooney2005). But also in Canada, Stephen Harper's conservative government currently selectively underfunds scientific areas that would provide results not conducive to its political-economic agenda, and prohibits scientists working for government research agencies to speak to the press without political approval (Harris Reference Harris2014; Rollmann Reference Rollmann2014; Turner Reference Turner2013). In 2008, the Canadian government decided to close the position of the non-partisan national science advisor, which had been created four years earlier (Hoag Reference Hoag2008); and it became known that Environment Canada ordered its scientists to refer all media inquiries to the government's communication officers, who would provide approved lines for the media (Ottawa Citizen 2008). Similar incidents followed, which the media described as the muzzling or gagging of government scientists (Reardon Reference Reardon2011; Saunders Reference Saunders2009). Examples are Environment Canada prohibiting its researcher David Tarasick to give interviews about his ozone layer research published in Nature, and the Privy Council Office barring Fisheries and Oceans Canada scientist Kristina Miller from talking about her findings on the decline of sockeye salmon that had appeared in Science (The Canadian Press 2012). In early 2012, the Canadian government announced the decision to close down the world-renowned Experimental Lakes Area freshwater research station and the unique Polar Environment Atmospheric Research Laboratory.Footnote ¹ This is to be seen against the background of the Harper government's reservations to climate change science findings and climate change mitigation efforts, having rescinded Canada's Kyoto protocol commitments and investing in a fossil-fuel-based economy, including the Alberta tar sands (which are worse than other fuel-producing technologies in terms of greenhouse gas emission and its environmental impact on the extraction site and during pipeline transport incidents; see also Goldenberg Reference Goldenberg2013b; Lukacs Reference Lukacs2015).

Combined with severe cuts to other government research agencies and the budgets of granting agencies that fund basic research conducted at Canadian universities, this led to the highly unusual event of a political rally held by scientists. On 9 July 2012, at a mock funeral on Ottawa's Parliament Hill, scientists mourned the ‘death of evidence’ (Davison Reference Davison2012; Pedwell Reference Pedwell2012). Canada's Government continues to be subject to media criticism (Galloway Reference Galloway2013b; Globe and Mail 2013; Goldenberg Reference Goldenberg2013a; Linnitt Reference Linnitt2013). This includes the charge of continuing to muzzle government scientists (Buranyi Reference Buranyi2015; CBC News 2012, 2015; Galloway Reference Galloway2013a; Gatehouse Reference Gatehouse2013; Globe and Mail 2014; Klinkenborg Reference Klinkenborg2013; for a detailed overview see Turner Reference Turner2013), which has prompted the Federal Information Commissioner to begin an investigation into the complaints about the government (Semeniuk Reference Semeniuk2015). Some of the critical commentaries also come from science journals (Nature 2008, 2012, 2013; O’Hara Reference O’Hara2010) and philosophers of science (Douglas Reference Douglas2013a; Gatehouse Reference Gatehouse2013).Footnote ²

The ‘death of evidence’ issue in Canada highlights the damaging effect of political interference with and undermining of science. At the same time, while upholding sound science and evidence as something that society should subscribe to regardless of one's individual beliefs and values, the scientists’ protest is also a political stance against the Harper government's political agenda, and the objection to this selective underfunding stems from their valuing of the environment and public health.Footnote ³ This likewise raises the spectre of politicized science, or at least the question of what role scientists’ values may play in science. Indeed, while clearly acknowledging some need for values, Psillos (Reference Psillos2015) wonders which social values have an improper influence or even infringe on evidence. This framing of evidence and values is rejected by Likwornik (Reference Likwornik2015) on the grounds that both are necessarily entwined, and Goldenberg (Reference Goldenberg2015) argues that also values are based on and in this sense embody empirical evidence. In what follows, I will contribute to general philosophical discussions on science and values. While many prior accounts make some room for social values while showing how this avoids politicized science, I shall argue for a stronger role for social values.Footnote ⁴

Upon reviewing some previous accounts of the role of values, in Section 3, I criticize the strategy of allocating different values to different steps of research, which has also been used by many sympathetic to social values in an attempt to provide a clear-cut distinction between the licit use of a value (in a certain research step) and illicit politicized science. The most convincing position for values in the context of theory acceptance has been the idea that social values can influence the evidential threshold. This argument from inductive risk maintains that we should demand more evidence (before hypothesis acceptance) if accepting an actually false hypothesis would have severe social and other practical consequences (e.g. Douglas Reference Douglas2009). However, such accounts have the consequence that when evidence increases (and thus any given evidential threshold is met), the impact of social values converges to zero. Using the case of past accounts of human evolution and primate social behaviour – which were both empirically flawed and sexist – in Section 4 I suggest that not only epistemic considerations, but also social and environmental values may determine a scientific theory's conditions of adequacy. Various considerations can be used by scientists to judge the adequacy of a theory, depending on the particular case, but in our context this includes what it means for the theory to be unbiased and complete. The role of conditions of adequacy is not diminished at all by evidence accumulating, so that my central aim is to argue for a stronger role for social values in the context of theory acceptance. This position is clarified and defended in Section 5, where I point out that my account cannot be captured by prior approaches that conceptualize theory acceptance solely in terms of making inferences from evidence, sketching a broader framework by creating connections to philosophical discussions of scientific explanation (which motivate and support my perspective, even though they did not engage with social values). The concluding section returns to the worry about politicized science.

2. Inductive risk and other prior accounts of the role of values

Philosophical discussions of the relation between science and values have gained prominence during the last two decades (Barker and Kitcher Reference Barker and Kitcher2013; Douglas Reference Douglas2009; Kincaid, Dupré, and Wylie Reference Kincaid, Dupré and Wylie2007; Kitcher Reference Kitcher2001; Lacey Reference Lacey1999; Longino Reference Longino1990; Machamer and Wolters Reference Machamer and Wolters2004). Furthermore, viewing science as including social values and being answerable to society has recently been complemented by a new vision of philosophy's aims, in terms of a socially relevant and socially engaged philosophy of science (Plaisance and Fehr Reference Plaisance and Fehr2010; see also Kourany Reference Kourany2010, ch. 5).Footnote ⁵ There are still some who uphold the ideal of ‘value-free’ science, but even a restrictive position does not deny that values can play a legitimate role in science. Rather, the strategy is first to distinguish between epistemic and non-epistemic values (Dorato Reference Dorato, Machamer and Wolters2004; Douglas Reference Douglas2000; McMullin Reference McMullin, Asquith and Nickles1983; Rooney Reference Rooney, Hull, Forbes and Okruhlik1992) – the terminology of cognitive as opposed to non-cognitive values is likewise used (Lacey Reference Lacey, Machamer and Wolters2004; Laudan Reference Laudan1984).Footnote ⁶ While ethical, social and environmental values are non-epistemic, epistemic values include a theory's internal consistency, its fit with evidence, its consistency with other theories, its predictive accuracy, its generality and unifying power, and its simplicity. It is uncontroversial that many, if not all of these epistemic values are rightly used by scientists. Second, different steps of scientific research are distinguished, for instance, research project choice, hypothesis acceptance and knowledge application (Douglas Reference Douglas2000). Everyone has to acknowledge that even non-epistemic values may be used in the choice of a research project and the technological application of knowledge, for example, scientists may study the effects of climate change on biodiversity because of an environmentalist agenda. Consequently, those who maintain that science ought to be ‘value-free’ distinguish different research steps so as to focus on the core step of science – accepting hypotheses by evidence – and maintain that in this context of theory acceptance, only epistemic values may be used.

But even excluding social values from theory acceptance would not undermine the philosophical need to study such values, given that other aspects of scientific research raise epistemological questions and have social impacts. The ‘death of evidence’ issue in Canada vividly highlights the societal effect of choosing certain lines of research (or not being able to do so due to selective underfunding), and of communicating and applying scientific knowledge (or being prevented from communicating results). Against the idea of a global aim of science, Kitcher (Reference Kitcher2001, ch. 6) has argued that scientific aims are particular aims arising locally in a discipline, where their scientific significance often combines intellectual and practical, application-related considerations. Going beyond this, Kourany (Reference Kourany2010, ch. 5) stresses the importance of critically – based on social values – evaluating the research problems scientists work on and the application of their results, and argues that philosophers of science should devote more of their work to these research steps. How even the choice of research projects can be philosophically scrutinized involving a combination epistemic and ethical-social considerations can be illustrated by de Melo-Martín and Intemann's (Reference de Melo-Martín and Intemann2011) investigation of the development of HPV vaccines. Current virus-like-particle vaccines are expensive to produce, have to be stored using refrigeration, and are effective only when three doses are administered over a six-month period. As result, the available vaccines will not benefit people in developing countries, who are most in need of HPV prevention. Yet de Melo-Martín and Intemann argue that at least in the past researchers did have the option of exploring other feasible types of HPV vaccines (naked DNA and live bacterial vectors), which are much more likely to be of benefit in developing countries. This yields guidelines for how more socially responsible science would look like, which are concrete given the epistemic–empirical considerations used.

In any case, there have been two basic arguments that social and other non-epistemic values matter even in the context of theory acceptance (Brown Reference Brown2013; Rottschaefer Reference Rottschaefer2003). One is the argument from underdetermination. Sometimes, it is phrased in terms of auxiliary hypotheses (Anderson Reference Anderson2004; Campbell Reference Campbell1994, Reference Campbell1998; Intemann Reference Intemann2001). Starting with the assertion that a theory cannot be tested in isolation and has testable consequences only once combined with an auxiliary hypothesis, the observation is that there are actual cases where the auxiliary hypothesis used embodies social values or that scientists’ preference for using this auxiliary hypothesis was based in part on social values. Moreover, different auxiliary hypotheses could be used (resulting in a different outcome for the theory to be tested), and since empirical considerations alone cannot settle which auxiliary hypothesis is to be chosen, there is nothing to rationally prevent one from choosing a theory because it aligns with one's social values. The latter idea is often directly phrased in terms of underdetermination (Intemann Reference Intemann2005; Longino Reference Longino1990, Reference Longino2002; Psillos Reference Psillos2015). The argument is that a theory is always underdetermined by evidence alone, so that additional considerations are needed to fill the gap between evidence and theory. Values, including social values, are not only an unavoidable, but also fruitful way to achieve theory choice. In the remainder of my discussion, I focus on the other type of argument for social values in the context of theory acceptance, given that it is often seen to be the more convincing one.

The argument from inductive risk stems from Rudner's (Reference Rudner1953) seminal account, according to which ethical and other non-epistemic values may influence the amount of evidence that is deemed to be sufficient for accepting a hypothesis. Using Hempel's (Reference Hempel1965) terminology, this approach is nowadays typically discussed under the label of inductive risk (Douglas Reference Douglas2000; Elliott Reference Elliott2011b; Steel Reference Steel2010, Reference Steel2013), given that evidential support for an empirical theory is always inductive, and given that accepting even a very well-supported theory is a risk, as it could still turn out to be false. But accepting and acting on a hypothesis that is actually false may have bad social and other practical consequences, for instance, approving a new drug on the erroneous assumption that it has no serious side effects. Thus, the more severe the consequences of endorsing a false hypothesis would be – by the light of social and other ethical values – the more evidence is needed before this hypothesis can be accepted. In line with this, Kitcher (Reference Kitcher1985) emphasized that the standards for sociobiological explanations claiming human social behaviours, e.g. racism or social hierarchies, to be evolutionary adaptations ought to be at least as high as the evidential standards used in evolutionary accounts of animal behaviour (and then offered an empirical argument that often the stark opposite was the case).

Heather Douglas (Reference Douglas2009) has prominently elaborated this approach by distinguishing between a direct and an indirect role for values. A hypothesis must not be accepted because it fits one's social agenda – which would be a direct role for social values – it is to be accepted because it is supported by evidence. But social and other non-epistemic values may be invoked to require more evidence, or to tolerate less uncertainty, when accepting a theory.Footnote ⁷ In this sense, social and other non-epistemic values play an indirect role, by modulating the evidential threshold, but not functioning as evidence.Footnote ⁸ Thus, Douglas's direct–indirect distinction is to lay out how non-epistemic values can play a legitimate role in theory acceptance, while articulating a safeguard against a politicized science, which is not strictly committed to scientific objectivity but would endorse theories merely because they align with a political agenda:

Moreover, the direct–indirect distinction goes beyond the context of theory of acceptance – on which most other accounts of inductive risk focus – given that according to Douglas, social and other non-epistemic values may indeed play a direct role in various other aspects of scientific research, e.g., research problem choice.

While this basic framework makes sense, let me question some of the details. Unlike many others, Douglas does not use ‘epistemic value’ and ‘cognitive value’ as synonymous. Inspired by Laudan (Reference Laudan, Machamer and Wolters2004), she views only internal consistency and predictive competency (i.e. fit with evidence) as epistemic, whereas she classifies other empirical values (such as broad scope, fruitfulness and simplicity) as cognitive values, on the grounds that such features of a theory are not indicative of its truth – the hallmark of the epistemic. In fact, Douglas does not take the epistemic criteria of internal consistency and predictive competency to be genuine values, Footnote ⁹ and sharply separates them from cognitive, ethical and social values. Epistemic criteria can always be used in a direct role, whereas values – be they cognitive or social – may only have an indirect role in the context of theory acceptance. Thus, not only is the distinction between direct and indirect roles a novel idea of Douglas's, for her it is the only important difference, unlike the traditional distinction between cognitive and social values:

A minor issue is that Douglas strictly separates internal consistency and external consistency (see also Douglas Reference Douglas2013b), where unlike a theory's internal consistency, the ‘consistency of a theory with other areas of science is [merely] a cognitive value’ (Douglas Reference Douglas2009, 93). Apart from the fact that others may argue that using the criterion of ‘being (in)consist with other theories’ is epistemically legitimate (when the other theories are likely to be true), Douglas's account is incoherent, given that ‘T being consistent with theory X’ is simply ‘T&X being internally consistent’. Of course there are pragmatic reasons for why a body of assumptions T is individuated so as to distinguish it from X, but this cannot yield the categorical distinction that Douglas postulates between the epistemic criterion of ‘being internally consistent’ and the (cognitive) value of ‘being consistent with other theories’, where only the former may have a direct role in theory acceptance.

More problematic is that Douglas (Reference Douglas2009) takes the distinction between a direct and an indirect role to hold for all values and to apply across contexts, Footnote ¹⁰ without actually explicating the distinction for all these situations. First, consider an aspect of science apart from theory acceptance, say choosing a research project. In this context it is clear what it is to use a social or environmental value in a direct role (e.g. studying climate change because of one's environmentalist values). But what could it possibly mean to use the same value in a different, indirect role when choosing a research project, or in other situations outside of theory acceptance? Second, the context of theory acceptance is where Douglas's distinction promises to have most traction, given that epistemic criteria may play a direct role, while cognitive, ethical and social values may play an indirect role only. For social and environmental values, there is indeed a distinction: endorsing a hypothesis because acting on it would have consequences that are positive in light of the value (direct role), as opposed to increasing the evidential threshold to the extent to which acting on the actually false hypothesis would have consequences that are negative in light of the value (indirect role). But Douglas includes cognitive values, such as a theory's simplicity or its breadth of scope. She views accepting a hypothesis because it is broader in scope than its competitors (and rejecting a hypothesis because it less broad in scope) as illicit, and classifies this as using the value in a direct role (102). Yet, she also seems to assert that a legitimate use of this value in an indirect role would be to ‘choose the broader scoped theory’ (108), without explaining how this indirect role differs from the direct one in the case of this and other cognitive values.Footnote ¹¹

It appears that Douglas has not offered a convincing articulation of her separation between epistemic criteria and values (which include cognitive values), and her distinction between a direct and an indirect role. This is a drawback insofar as the distinction promised a clear boundary between properly value-using science (committed to objectivity and evidence) and illicit politicized science. Be it as it may, I now move on to more interesting limitations of her framework, given that they also hold for many other accounts of the relation between science and values.

3. The limited relevance of segregating different research steps

Arguments from inductive risk as such pertain to theory acceptance; and in the following sections I will focus on the context of theory acceptance as well. But I start out with broader considerations that have implications for different accounts of the role of values in science (and which will turn out to be relevant for my particular account of theory acceptance, as Section 5 will explain). Not only those who still maintain that non-epistemic values should not impact theory acceptance in any way (while acknowledging that values may be used in other aspects of research) separate theory acceptance from other steps of scientific research. Even some proponents of inductive risk arguments, to the extent to which in the context of theory acceptance they restrict the role of social values (to the setting of the evidential threshold), have to distinguish different steps of research. Douglas's (Reference Douglas2009) position is that non-epistemic values may play a direct role at some steps of research (e.g. research project selection and knowledge application), whereas the situation is different ‘in the heart of doing science – during the characterization of data, the interpretation of evidence, and the acceptance of theories’ (102), given that ‘a direct role for values at this point in the scientific process is unacceptable, but an indirect role is legitimate’ (96, emphasis added).

While there is nothing wrong with distinguishing different aspects or steps of research, I argue that there are limits to the philosophical strategy of segregating different research steps with the aim of being able to adjudicate whether a particular value was licitly (or illicitly) used by looking at this one step. One basic reason is that the actual outcome of research, e.g. a journal article endorsing a hypothesis, requires a sequence of several steps, so that a critical evaluation of this research – be it on epistemic or on social grounds – sometimes cannot assign blame to an individual research step regardless of the outcome it only yields in combination with other steps. A good example is the well-known fact that drug trials sponsored by the companies producing the drug are more likely to find it effective and without serious side-effects than independent studies (Lundh et al. Reference Lundh, Sismondo, Lexchin, Busuioc and Bero2012). Something wrong is going on here, but it is usually not the case that the authors would falsify data or use shoddy statistical methods – the papers pass rigorous peer review precisely because they follow standard epistemic procedures. The explanation for the bias is that industry-sponsored trials know to ‘ask the right questions’ (Smith Reference Smith2005), e.g. having the trial drug run against too low a dose of a competitor drug (making the trial drug seem more effective) or against too high a dose of a competitor drug (making the trial drug's side-effects appear relatively minor).

This asking of particular questions to get a more favourable outcome does indeed belong to the step of research question selection or the step of experimental methodology selection (which of these steps, or both?). But there is nothing intrinsically wrong about asking such a question, e.g. comparing the drug against a particular competitor drug with a certain dosage. What is first and foremost problematic is the impact that the published results would have, where the results follow from the question pursued in combination with an experimental methodology and statistical means of data evaluation – the latter being the ‘hypothesis acceptance based on evidence’ step. Likewise, using a statistical inference method is not licit in every case, but whether it is legitimate depends on the research context, where the relevant context may contain other ‘steps’ of research. Thus, even if one can assign (epistemic or social) blame to a particular aspect of research, e.g. the asking of a certain question, this verdict can often be made only because of how this step connects up with other research steps. Conversely, in general one cannot give a free pass to a value's use as long as this use pertains to a specific step of research.

Kourany's (Reference Kourany2010) call for the ideal of a socially responsible science assumes that considerations involving social values hold for every aspect of research. She does not offer an argument for this, given that she does not explicitly address the rival position that social values may play a direct role at some research steps, but not in the acceptance of theories. But one of her examples illustrating socially responsible science is highly instructive. The studies by psychologist Carolyn West (Reference West, Collins, Dunlap and Chrisler2002) on domestic violence aim at improving the condition of black woman in the USA. Despite uncovering some similarities among black and white communities, this research faces the challenge of exhibiting domestic problems encountered specifically by black women without perpetuating the stereotype that black men are inherently more violent. The pursuit of this dual social aim

In particular, it requires broadening the definition of ‘partner violence’ to include emotional and sexual in addition to physical abuse and to change the measuring of sexual violence by taking its context into account.

This suggests the second reason for why one cannot base one's account of the role of values on a separate treatment of steps of research. Scientists’ use of particular definitions, categories, and concepts impacts all aspects of research. And sometimes a category (e.g. ‘partner violence’) embodies value-judgments, or more precisely, its scientific use leads to an effect that is more problematic than an alternative definition would have.Footnote ¹² An example that goes beyond social science is climate change research (Schienke et al. Reference Schienke, Baum, Tuana, Davis and Keller2011). Climate management models may use the concept of ‘utility’ as a measure of global wealth, which is to be maximized by an appropriate climate change mitigation strategy. But if utility is defined in terms of average global wealth, the model ignores any global inequities, even though climate change will have particularly severe impacts on some human populations and mitigation efforts may enhance existing inequities. Conversions of floods, droughts and other climate change-induced events into units of monetary value likewise embody values. In summary, the formation of a proper scientific category, and a value-based criticism of a category used (regardless of whether more epistemic or more social considerations are involved), are to be done in the light of several research steps and their joint operation.

Some philosophical accounts segregate different steps of research and restrict the role of social values to certain steps, in an attempt to make room for values without compromising the objectivity of science. My considerations have questioned the success of this strategy. For instance, allowing values to be used in the choice of research questions may not prevent such values having a (problematic) impact on the acceptance of theories, given the wide-reaching ramifications of scientific categories (which may be value-laden), or given the way in which asking a certain research question may bias what the evidence-based outcome will be (as in the case of industry-sponsored drug trials). Douglas actually recognizes the latter issue:

However, this admission qualifies her basic idea that values may be used in a direct role in the context of research problem choice – without offering an account of the conditions in which the direct role is illicit. Given that one would have to look at a context larger than research problem choice, such an account would go beyond Douglas's philosophical distinction between direct and indirect roles.Footnote ¹³ That said, I do not recommend a solution in terms of restricting the general influence of values even further. In contrast to any philosophical account segregating steps of research, there may be no general articulation of the role of social values – at least one which considers research steps in isolation – ensuring a clear boundary that prevents the use of values to lead to an improperly commercialized or politicized science. But this situation is unlikely to be specific to social values, in that sometimes the problematic use of an epistemic value or method can only be criticized based on a consideration of the overall process of research.

4. Toward a role for social values beyond setting the evidential threshold

Arguments from inductive risk clearly demonstrate that social values matter even in the context of theory acceptance. However, unless one adds a further role for social values, one implication of inductive risk approaches is that the impact of social values converges to zero upon evidence accumulating. For such accounts the role of social values is to influence the evidential threshold, and increasing evidence will meet any evidential threshold. As Heather Douglas puts it:

This shows that Douglas fully endorses this diminishment of the role of social values, as it excludes a politicized science where values would function as evidence (Douglas Reference Douglas2009, 113, 122). However, my main aim in this paper is to argue for a stronger role for social and other non-epistemic values in the context of theory acceptance, a role that does not decrease with the accumulation of evidence.Footnote ¹⁴ I do so by calling attention to a 1995 paper by Elizabeth Anderson, in which she proposed a ‘cooperative model of the interaction of normative and evidential considerations in theory choice’ (52). An insight of hers deserves to be quoted:

Anderson illustrated this by discussing the Nation of Islam's book The Secret Relationship between Blacks and Jews, which, even if all alleged facts were true, would still offer a biased (anti-Semitic) account of the role of Jews in 17^th-century Atlantic slave trade. Here I point to the natural sciences to make a case for a strong role for social values, namely, recent accounts of human evolution and primate social behaviour.

Past theoretical accounts of human evolution and archaeological field studies had an androcentric bias toward men's innovations (Schiebinger Reference Schiebinger1999; Wylie and Nelson Reference Wylie, Nelson, Kincaid, Dupré and Wylie2007; Zihlman Reference Zihlman and Hager1997). This was aggravated by some approaches trying to pinpoint a particular cognitive, behavioural or technological innovation that was especially instrumental in making our ancestors modern humans. Dominant until the 1960s, the man-the-hunter account assumed that hunting furthered the development of bipedalism, an enlarged brain, and tool use. It featured hunting, especially big-game hunting, as an activity of men, which as a socially coordinated activity required new cognitive abilities. Later on this overall vision was qualified by the addition of the woman-the-gatherer account, which saw women not just as passive in human evolution, but contributing to subsistence and the development of tools required for gathering. It was shown that for the majority of our ancestors most of the caloric intake did not come from big-game hunting, where women are likely to have provided more overall food by weight than men.

While initially the archaeological record about even stereotypically female activities was largely ignored, sometimes rationalized by the assumption that women's activities are archaeologically largely inaccessible due to their more perishable artefacts (unlike stone tools), nowadays the role of gender and the specific contributions of women has become an important dimension of investigation. Even the woman-the-gatherer account has come to be seen as stereotyping gender roles as well as being empirically inadequate. This has led to the reinterpretation of previous data. For instance, pestles and other tools had originally been interpreted as indicating female household activity in many cases of being found with a woman, yet as evidence of men manufacturing such tools when the same kind of tool was found with a man. Novel kinds of data have also been made possible by the new perspective, such as the search for traces of more perishable artefacts, or analyses of bone breakage patterns and stone tool wear patterns, which provide evidence of secondary processing (after hunting and butchering), and thus sophisticated tool use involving women.

A similar situation was to be found in accounts of non-human primate social behaviour (Hrdy Reference Hrdy1981, Reference Hrdy and Bleier1986; Schiebinger Reference Schiebinger1999; Strum and Fedigan Reference Strum, Fedigan, Strum, Lindburg and Hamburg1999, Reference Strum and Fedigan2000). Early studies happened to devote more observational and theoretical attention to the activities of male primates. Individuals were categorizes as dominant males, peripheral males and females/young, so to the extent that the social role of female primates was investigated, it often boiled down to reproduction and rearing offspring. Indeed, even if the scope of research is not overall social behaviour but more specifically sexual behaviour, female sexuality cannot be restricted to reproductive sex, as philosopher Elizabeth Lloyd (Reference Lloyd1993, Reference Lloyd2005) has argued. She points to studies who run afoul of this by observing male–female sex only or more explicitly considering female behaviour as sexual only if it is reproductive, Footnote ¹⁵ even though it is known that in many primate species females engage in sexual activities outside the oestrus (and thus independently of reproduction) and in bonobos among others there are widespread female–female sexual interactions.

The observational focus of early primatology led to accounts that were similarly androcentric as past theories of human evolution, in that they entailed more active behaviours and a more influential social role for male primates, while females were initially seen as being non-competitive and submissive, trading sex for food and protection. Observations and theoretical interpretations were also guided by the theory of sexual selection. Introduced by Darwin, sexual selection maintains that in animals it is the males that compete among each other (for access to females), while a female chooses a suitable male, which has led to the stereotype of the promiscuous male primate and the ‘coy’ female primate. This bias started to be remedied only when a larger proportion of female researchers entered primatology (Small Reference Small1984), who discovered female competition for reproductive success and female sexual assertiveness. Female promiscuity includes extra-pair mating that enhance the female's fitness while she maintains the support of her existing partner, or engaging in sex with multiple males who all contribute to caring for the offspring. The significance of female bonding and matrilineal networks has been discovered and come to be studied. It was likewise revealed that engaging in special social relationships with females can have a higher impact on a male's reproductive success than his rank in a male dominance hierarchy. In baboons, it is females who provide social stability to groups (while males move across groups), and older females who determine foraging routes. The theoretical framework of sociobiology to view each animal as choosing among several strategies depending on what currently is fitness enhancing has made it possible to view even seemingly passive female behaviour as discriminatorily exhibited in this context only and as an active, opportunistic choice. Furthered by novel theoretical perspectives, these various findings about different primate species led to the recognition that females’ behaviour is richer, more diverse, and of more biological impact than previously assumed.

I have briefly reviewed theories from the recent history of archaeology and primatology that were empirically flawed and had sexist connotations. One point that these empirical cases show is that to arrive at an adequate scientific account, it is not sufficient to rely on an evidential basis no matter how true. (Recall Anderson's claim that there can be ‘unworthy representations of a phenomenon even if they contain no falsehoods’.) In fact, when for instance only male primates happen to be studied, enlarging the amount of observations would lend better support to what is actually an inadequate, because gender-biased, theory of primate sociality. This is an issue obscured by accounts of inductive risk, which focus on reducing uncertainty by increasing the amount of evidence.Footnote ¹⁶ Demanding that the observational base include all facts in the domain is not a satisfactory philosophical move, as a scientific model may legitimately abstract away from many individual details. Instead, in addition to conformity to some evidential basis, there are further conditions of adequacy that a scientific theory has to meet. Such conditions of adequacy can include standards of what it means for the theory to be unbiased and complete, and on my account, both epistemic and social values determine a theory's conditions of adequacy. This is not to deny that social values are also important for specifying what kind of evidence is to be gathered. But my approach is to move philosophical attention to the properties of theories, including what kind of theory is intended and what such a theory's conditions of adequacy are (which then ground what kind of evidence is needed). Conditions of adequacy are standards that always have to be met by a theory, so their role is not diminished when evidence accumulates or scientific uncertainty decreases. So if I am right that social values influence a theory's adequacy conditions, this yields indeed a stronger role for social values than recognized by inductive risk (and other) approaches. I will articulate my account in more detail and defend it, based on the above biological case, in the following section.

5. Beyond the inference-from-evidence framework: a theory's purpose and adequacy conditions

Many philosophical accounts of how social and other values matter for theory acceptance operate within what I call an inference-from-evidence framework, i.e. they focus on the relation between evidence and theory, typically proceeding from evidence so as to wonder whether it inferentially warrants theory acceptance. Consequently, how values matter is framed in terms of how values bear on the evidence-to-theory inference. This framework is obviously used by inductive risk approaches, which address evidential thresholds – what counts as a sufficient amount and strength of evidence – and invoke the concomitant notions of uncertainty and risk in inductive inference. But it also holds for underdetermination arguments for values, given that their starting point is that theory choice is underdetermined by evidence alone (Section 2), so that values have to bridge the gap between evidence and theory. I generally want to go beyond a philosophical framework that conceptualizes these issues exclusively in terms of making inferences from evidence, and motivate my broader picture as follows (also creating a connection to the issue of scientific explanation).

In the article on which I relied above, Anderson (Reference Anderson1995) sets up her discussion by criticizing a 10-step argument for the idea that values cannot play a role in the justification of theories, which she broadly ascribes to Haack (Reference Haack1993). (See also Haack Reference Haack, Nelson and Nelson1996, which shows that Haack's concern is to prevent politicized science.) The first two steps are most relevant for my purposes, so I restate only them:

This aligns with an idea we have already encountered, the traditional tenet that while considerations of significance – including social relevance – may well be used when choosing research problems, in the context of theory acceptance only fit with evidence matters, but significance does not. But philosophical accounts of scientific explanation – and thus considerations independently of social values – already show premise 2 to be flawed. It is well known that not every true representation qualifies as an explanation (Craver Reference Craver2007). A scientific representation may merely describe a phenomenon or capture correlations among features without explaining. More than truth and fit with observations is needed, to wit, an explanation has to reveal explanatorily relevant features, such as genuine causal relations. Many scientific theories are expected to offer explanations, so that pace Haack such theories are accepted only if the account meets considerations of (epistemic) significance, such as explanatory capacity.

Explanatory accounts can be very complex. Love (Reference Love2008) uses the notion of a ‘problem agenda’ for a complex scientific problem that consists of many individual but related questions. He views a problem agenda as being tied to various (implicit) criteria of explanatory adequacy, which determine what conditions an adequate explanatory account has to meet. In the scientific case he considers, the explanation of the evolutionary origin of novelties (e.g. the jaw, originating in the transition from jawless vertebrates to jawed vertebrates), the criteria of explanatory adequacy entail that explanatory contributions are needed from genetics, cell and developmental biology, from systematics and paleontology, and from ecology, so that a multidisciplinary account is required. Love argues that any adequate explanatory framework of the origin of novel traits must address not only structural but also functional aspects of traits, among other things (Brigandt and Love Reference Brigandt and Love2012; Love Reference Love2013a).

This biological case likewise shows that criteria of adequacy are far from straightforward (even when only epistemic-empirical issues are at stake). The considerations are much more complicated than the guideline of putting forward an explanation by adducing causes or mechanisms; in fact, there are serious biological disputes about the relevant criteria of explanatory adequacy in the case of evolutionary novelty. Some evolutionary developmental biologists favour explanations in terms of changes in gene regulatory networks, while others maintain that a broader framework is required, which addresses the developmental interaction of genetic and non-genetic features on several levels of organization. And some mainstream evolutionary biologists reject the aspirations of evolutionary developmental biology and claim that a traditional population genetic explanation is all that is needed (Love Reference Love and Kampourakis2013b, Reference Love2013c; Love and Lugar Reference Love and Lugar2013). Explanatory standards are often merely implicit assumptions, which may well be in need of scrutiny. For instance, Robert (Reference Robert2004, ch. 1) grants the fruitfulness of the strategy in developmental biology of investigating the causal impact of genes, while holding other factors constant. However, he rightly insists that some biologists’ implicit assumption that an adequate explanation of development can be given in terms of the activity of genes only may well be prompted, but cannot be justified by this experimental methodology – which does not even investigate the causal role of non-genetic factors (given that they are held constant).Footnote ¹⁷

Now the connection to the topic of social values should become clear. Prior philosophical accounts surrounding scientific explanation (independently of any engagement with social values) have recognized that conformity with an evidential basis is not enough, but that many scientific theories have to fulfil standards of scientific significance, such as conditions of explanatory adequacy. Depending on the scientific purpose for which a particular theory is being developed and used, there are certain conditions of adequacy the theory has to meet, which apart from being explanatory (required of only some theories) can include conditions of what makes this theory unbiased, complete, and/or practically applicable. My account now maintains that a theory's conditions of adequacy are determined not only by empirical considerations and epistemic values, but also by social, environmental, and other non-epistemic values, if the theory is intended to capture empirical features that matter by the light of the non-epistemic values, or if the theory's use has social or environmental implications.

In the case of a theory of human evolution or primate social behaviour, among other things the conditions of adequacy stipulate that to be complete, the theory include the social contributions of women (to the extent that they impact human evolution) or the social role of female primates – both the social influence that aligns with men's/male primates’ and that is distinctive to women/female primates. Some may grant that past theories in these domains were indeed inadequate, while arguing that social values are not implicated given that the theories were inadequate with respect to purely epistemic considerations, e.g. the capturing of all explanatorily relevant social activities. One problem with this construal is that it does not quite mesh with the considerations of some of the scientists involved. Even though several of the for the most part female biologists who criticized and rectified past accounts denied to have employed an explicit feminist agenda and in some cases may not even have identified using the label ‘feminist’, it is clear that at the very least an implicit gender-sensitivity has played a role in their theoretical reflections (Fedigan Reference Fedigan and Hager1997; Kourany Reference Kourany2010; Schiebinger Reference Schiebinger1999; Strum and Fedigan Reference Strum, Fedigan, Strum, Lindburg and Hamburg1999). The same holds for past work in archaeology (Conkey and Gero Reference Conkey and Gero1997; Conkey and Spector Reference Conkey and Spector1984; Hanen and Kelley Reference Hanen, Kelley and Embree1992; Wylie and Nelson Reference Wylie, Nelson, Kincaid, Dupré and Wylie2007).

But regardless of what values motivated (some) scientists in the past, these scientific accounts can and should be scrutinized also based on such social values as gender equity. I indeed claim that the very value of gender equity is relevant. This is possible as my account distinguishes between conditions of adequacy, which are scientific standards to be met by a theory (e.g. ‘include the social role of females’), and the values and other considerations that guide the formulation of adequacy conditions. In this particular scientific case, the value of gender equity does not entail that the same social roles or sex-related behaviours are to be ascribed to males and females, but that the social contributions of females (be they similar or different from males’) are to be captured by the theory.

One possible reason for this is that gender-biased and empirically flawed accounts of human evolution – and even of non-human primate sociality – can make current human gender stereotypes or inequities seem to reflect a general natural reality. This would justify adequacy conditions based on the social consequences that the promotion of an inadequate theory would have, analogous to how inductive risk accounts appeal to the consequences of endorsing a false theory (although I have indicated that there is more is to be considered than the falsity of scientific representations). But my account permits the formulation of adequacy conditions – based among other things on social values – even without the condition's violation having any tangible social consequences. After all, adequacy conditions stem from a theory's purpose or what the theory is intended to represent; and in case of a theory being meant to offer a scientific explanation (the context in which I introduced the idea of adequacy conditions), such an adequacy condition obtains regardless of whether putting forward a non-explanatory scientific account would have any social consequences. In the case of a primatological theory, in addition to other considerations (e.g. capturing causal features impacting primate sociality), in our current societal context, considerations about gender equity likewise inform the adequacy condition of including the social role of female primates (regardless of consequences on humans). Social values are in this scientific case crucial, because not only do they contribute to specifying what a complete theory has to include, but they are needed to determine that omitting female contributions counts as a particularly egregious kind of incompleteness.Footnote ¹⁸ In summary, if an account of primate sociality or human origins is deemed to be inadequate for ‘scientific’ reasons, this is to a relevant extent due to the fact that a similar account would be unacceptable when used in accounts of extant humans, for social reasons.

A social value need not have the same implication for what the adequacy conditions are in any context; rather, my claim is that ‘gender equity’ has this impact in our current social and scientific context. The actual conditions of adequacy of a primatological theory are more specific than what I mentioned above, and making them explicit would require fleshing them out in detail. Indeed, in addition to values, empirical background knowledge is needed to arrive at adequacy conditions. For instance, in Section 4 I mentioned how Lloyd (Reference Lloyd1993) as a feminist philosopher objects to studies of female primate sexuality that investigate only sex with males (so as to ignore widespread female–female sexual interactions).Footnote ¹⁹ A potential objection to Lloyd is that given the studies’ aim of arriving at evolutionary explanations, only sexual behaviour increasing reproductive fitness matters. However, recent theoretical work by Joan Roughgarden (Reference Roughgarden2009) shows how various non-standard gender behaviours found in the animal kingdom contribute by means of an ecological layer of complex social interactions to the individual's fitness as the ultimate evolutionary currency. Thus, novel empirical work can be used to show that the equating of female sexuality with reproductive sex is inadequate even for the aim of an evolutionary theory. A consequence of this influence of empirical beliefs on adequacy conditions is that a change in empirical assumptions may well lead to a revision in adequacy conditions used, so that a theory once endorsed can no longer be deemed to be adequate.

Other philosophers may acknowledge the relevance of social values, while trying to argue that my case does not demonstrate a stronger role for values in the context of theory acceptance, given that the social considerations are relevant to the choice of research questions and investigative methodologies. For instance, rather than ‘investigating primate social life’, the research question be more specifically ‘investigating the role of males and females in primate social life’ and observational methodologies be mindful of androcentric bias. However, from my perspective, such social values matter to research problem and methodology choice because the values originate from the intended theory's conditions of adequacy. In Section 3, I already argued that when it comes to the impact of values, different steps of research cannot be segregated. There the focus was on how a value underlying research question choice can sometimes be seen to be problematic only once its effect on other research steps, including theory acceptance, are considered. Now my point is that value considerations in the steps of research question and methodology choice are grounded by the theory's scientific purpose and the theory's conditions of adequacy, and in this sense the context of theory acceptance. Thereby the notions of a theory's purpose and its adequacy conditions, which my framework invokes, have an impact across various aspects of research. In our primatology case, gathering a diversity of evidence so as to be mindful of gender variation and androcentric bias is indeed a desideratum for observational methodology, but the underlying values stem from the theory's purpose. In any case, even if the social values are also operative in the early steps of research, the conditions of adequacy – based in part on these values – must always be met in theory acceptance.

My account also shows why social values make a positive contribution, rather than it being sufficient to remove problematic social values, such as androcentrism (see also Goldenberg Reference Goldenberg2015; Kourany Reference Kourany2010, ch. 3). Conditions of adequacy are scientific standards that have to be set up in the first place; and in the case of the evolutionary origin of novelty epistemic values are needed to defend what qualifies as an adequate explanation in this context. In the case of archaeology and primatology, in addition to epistemic values and empirical considerations, feminist values fruitfully contribute to delineating what counts as a complete account of human evolution and the social behaviour of non-human primates.

Longino (Reference Longino, Nelson and Nelson1996) has prominently scrutinized the dichotomy between epistemic and non-epistemic values (see also Longino Reference Longino1995). On her account, in a situation where established biological theories actually have a sexist bias, using the epistemic value of consistency with other theories will work against the acceptance of new, non-sexist accounts, so that in this context this epistemic value also has a social valence. Conversely, values that are social or at least were proposed by feminist discussions of science, e.g. ‘novelty’ (significantly differing from current theories), ‘ontological heterogeneity’ and ‘mutuality of interaction’, can in certain contexts further the endorsement of theories that are empirically more adequate. I likewise do not endorse a dichotomy between epistemic and non-epistemic values, as on my account epistemic values, social values and other values have the same role of (jointly) determining a theory's conditions of adequacy. The contributions of epistemic and social considerations may also be hard to disentangle, for example, covering the role of females in human evolution is in part an epistemic consideration about causal completeness and in part a feminist concern about capturing the social significance of women.Footnote ²⁰ Rather than merely calling for diversity of evidence and investigative methods, Longino rightly views feminist values such as ‘ontological heterogeneity’ as applying to theories – just like the epistemic values of ‘simplicity’ and ‘consistency with other theories’ are properties of theories. Yet, my account offers a clearer articulation of the impact of values, given that I distinguish between values and conditions of adequacy, where only the latter have to be met by a theory. Values are often quite generic and domain-general, as are the epistemic values of simplicity and consistency with background knowledge, and the social values of gender equity and non-androcentrism. In contrast, the conditions of adequacy that the values determine (in combination with empirical considerations relevant to the scientific domain) are specific standards that a particular theory has to meet.

Some may object that given that different values may legitimately be brought to bear on a theory, there are cases where my account has the implication that the theory is adequate (relative to one set of values) but at the same time inadequate (relative to other values). Note that something similar already obtains in the case of the widely accepted inductive risk accounts, given that a theory may meet one evidential threshold (that stems from only certain practical consequences of erroneous theory acceptance being deemed to be socially problematic) but not another evidential threshold (stemming from a different ethical judgement about the various consequences of acting on a false theory). But my response to the objection is bolder, in that it would be misguided to conceptualize the issue in terms of theories (isolated from any scientific aims and standards) being accepted given the evidential basis. From my perspective, a scientific theory is a tool developed by us for certain intellectual and practical purposes, so that a scientific account has always to be viewed together with the purpose for which it is used. Scientists recognize that models are not all-purpose tools and that different models are required for different epistemic purposes (where some epistemic purposes can even make reality-distorting idealizations legitimate). The same holds when the purpose includes social considerations; and there is nothing wrong about a particular theory being adequate for one human purpose but not being adequate for another purpose. My perspective involves a broad construal of theory ‘acceptance’, which in line with a good deal of scientific practice is not just the passive belief in individual propositions, but the active scientific development and use of theories (for a certain purpose).Footnote ²¹ In addition to using a scientific model for some aim specific of a research group, this includes the communication of an overall scientific account to the public – which is by no means an insignificant aspect of science, as scientists’ push for being able to speak about their findings without government censorship in the case of Canada's ‘death of evidence’ issue reminds us.

These considerations show that my account of the role of social values cannot be captured by a framework that exclusively conceptualizes theory acceptance in terms of making inferences from evidence. The broader philosophical framework that is needed for any discussion of values in science (though it can already be found in other philosophical contexts) includes that an important aspect of scientists’ acceptance and use of theories are conditions of adequacy, which are concrete, domain-specific standards of what it means for a particular scientific theory to be significant, explanatory, unbiased, complete or practically applicable. Such standards are based on the epistemic, social and/or environmental purpose for which a scientific theory has been developed and is being used. While such purposes are also germane to the issue of choosing research questions and methodologies for developing scientific accounts, they cannot be neglected in context of theory acceptance. In fact, the purpose for which a theory is to be used and the underlying values have an overarching impact on various aspects of research.

6. Conclusion: politicized science or socially responsible science?

While inductive risk arguments for social values (and likewise underdetermination arguments) operate with an inference-from-evidence framework, I have argued that a broader philosophical framework is needed, which recognizes that scientific theories are developed for certain intellectual or practical purposes and are used to serve human interests. Such a purpose – and the associated epistemic, social and other values – yield specific conditions of adequacy that a theory has to meet, in addition to merely confirming to an evidential basis. Traditional inductive risk accounts provide a compelling argument for social values being relevant even in the context of theory acceptance. However, an implication of this approach is that the role of values declines whenever evidence accumulates. My account yields not only an additional, but a stronger role for social and other non-epistemic values, because it maintains that a theory's conditions of adequacy are in part determined by such values, and conditions of adequacy always have to be met and thus their role does not decline.

Although the ‘death of evidence’ issue in Canada is prompted by interference in science from the outside, I indicated that Canadian scientists’ response is likewise motivated by political values, including concern for the environment and public health (Section 1). Especially in this latter context, my philosophical framework, which endorses that social and environmental values underlie a scientific theory's purpose (affecting even theory acceptance), may raise the spectre of politicized science. Indeed, whereas Heather Douglas's inductive risk account has been developed to provide a safeguard against politicized science, where the priority of evidence over social values is encapsulated by the idea that the impact of values decreases upon the accumulation of evidence, I reject this latter idea. However, my framework may not permit illicitly politicized science any more than other philosophical accounts, given that I have questioned the success of efforts to erect a clear boundary against politicized science, including Douglas's distinction between direct and indirect roles for values (Section 2), and any segregation of different steps of research that restricts the use of values to some particular steps (Section 3).

Moreover, in contrast to clearly problematic instances of politicized science, my account likewise does not permit scientists to distort evidence or to ignore relevant evidence, and a scientific theory has to reflect reality. For instance, the adequacy condition that an account of human evolution includes the contribution of women does not mean that one may prefer a fictional account over a veridical account because the former would be less sexist. Instead, historical contributions of women are to be included to the extent to which there is evidence for them, and the kind of contributions actually made are to be captured no matter how similar or dissimilar they are to the contributions of men.Footnote ²² I do not endorse politicized science in that evidence matters to my account of theory acceptance. Still, conforming to evidence is not the only consideration in theory acceptance. For amongst the various representations of some aspects of reality, scientists choose one that suits their particular aims and standards, which often are certain explanatory aims and standards of explanatory adequacy, but can also include social and environmental aims that call for the theory to include relevant features of reality. Scientific models and theories are tools for us that also answer to our intellectual and practical purposes. Given this, just like ‘a theory is to be accepted solely because it aligns with the evidential basis’ in contrast to ‘a theory can be accepted solely because it aligns with one's scientific or social agenda’ is a false dichotomy, I reject the opposition that Douglas views between aiming at understanding about the world and aiming at understanding that suits one's interests:

In his discussion prompted by the ‘death of evidence’ issue in Canada, Stathis Psillos (Reference Psillos2015) grants a role for social values in science, but notes that the primacy of evidence and the objectivity of science can only be ensured if ‘the right kind of values’ are used. I agree with this, in fact, within my framework that invokes the notion of a theory's purpose, scientists developing and using a theory for some purpose neither means that this purpose cannot be questioned, nor that the values underlying a scientist's particular choice can be illegitimate. But the use of particular values, including social values, can be justified, where – as Goldenberg (Reference Goldenberg2015) argues – empirical inquiry contributes to the rational evaluation of values.Footnote ²³ Overall, my approach is in line with what Kourany (Reference Kourany2010) calls the ideal of socially responsible science, which, while committed to epistemically sounds science, considers what the proper social aims are and conducts research so as to further them.

To be sure, my proposal endorses a distinctive role of values, where social and other non-epistemic values impact a theory's conditions of adequacy. While some philosophers may worry that adequacy conditions resulting from an interaction of different types of values (epistemic and social alike) could not be laid out at all, I respond that this is not a problem created only with the addition of non-epistemic values, and that scientists already tackle it as a scientific issue. In the case of evolutionary developmental biology, I pointed out that there are ongoing scientific debates about epistemic standards for what counts as an adequate framework for explaining the evolutionary origin of novelty (Section 5). In the context of human evolution and primate social organization, social considerations about androcentrism have played a role in how the scientists involved have criticized past accounts and attempted to develop more adequate theories. Given that conditions of adequacy and assumptions about the shape of a proper scientific framework are often implicit, it is in fact important that scientists attempt to make such standards and the underlying aims explicit, so that they can reflect on and defend them. And for the purpose of scrutiny from outside of science, it is minimally necessary that scientists are transparent about the values they endorse and how they use them. Sure enough, this alone does not yield a social mechanism for the values of the stakeholders affected by the formation and application of particular scientific knowledge having an influence on the scientist's values. But how this can be achieved and the related question (also addressed by Goldenberg Reference Goldenberg2015 and Psillos Reference Psillos2015) of whose values matter – the scientists’, the affected stakeholders’ or the societal majority's – is to be discussed as part of an open dialogue that goes beyond a philosopher's treatise.