1 Introduction

In recent years, historians, philosophers, and sociologists of science have told a horror story about the co-option of scientific expertise to undermine regulation in the public interest. In cases that range over tobacco consumption, asbestos exposure, the opioid crisis, the epidemic of honeybee deaths, and, perhaps most significantly, climate change, scholars have documented a dizzying array of strategies – perfected by industries under regulation, mobilising lobbyists, consultancy firms, think tanks and government agencies – that weaponize science to undermine the empirical basis for regulation (e.g., Fernandez Pinto, 2017; Franta, 2022; Freudenburg et al., 2008; Holman & Elliott, 2018; Kleinman & Suryanarayanan, 2013; Michaels & Monforton, 2005; Oreskes & Conway, 2010). The documented strategies vary in sophistication, from crude data manipulation and doctoring of graphs (e.g., Oreskes & Conway, 2010, pp. 186–190) to more subtle practices that appear cloaked in scientific norms and conventions, appealing to scepticism, criticism, and stringent evidential standards, but which effectively supress inconvenient knowledge (see Fernandez Pinto, 2017; Kleinman & Suryanarayanan, 2013). Such investigations form a substantial part of the emerging discipline of agnotology, dedicated to the study of factual ignorance and its aetiology (Proctor & Schiebinger, 2008). Their findings cast a dark shadow on contemporary science: public ignorance of crucial policy-relevant facts has been cultivated and maintained by ‘an increasingly politicized and commercialized science’ (Kourany & Carrier, 2020, p. 3).

Thanks to this literature, we have a much clearer picture of how our reliance on expertise leaves us vulnerable to empirical corruption: to being misled about empirical facts. Since a successful exercise of moral agency depends on knowledge of relevant empirical facts, our vulnerability to empirical corruption leaves us vulnerable to moral and political failure. What this literature neglects, however, are the ways in which scientific expertise can be deployed to undermine moral agency in ways that do not involve misleading audiences about empirical facts; indeed, which involve candidly presenting the relevant facts. The aim of this paper is to explore such wrongful deployment of scientific expertise beyond empirical corruption. To do so, I develop an account of morally subversive expertise, building on recent work in political philosophy (Howard, 2016). Expertise is subversive if it sets up its audience to fail morally, either intentionally or negligently. I distinguish three modes of subversion: empirical, normative, and motivational subversion. The three modes of subversion differ with respect to the mechanisms by which they induce moral failure. Roughly, empirical subversion transpires when an agent renders another ignorant of morally relevant empirical facts (the focus of agnotology); normative subversion transpires when an agent renders another ignorant of the content or role of morality; and motivational subversion transpires when an agent tempts another to act wrongly, e.g., by presenting them with practical incentives for wrongdoing. Scientific expertise, I argue, can be deployed to induce wrongdoing along all three dimensions – our reliance on expertise leaves us vulnerable to moral failure in broader ways than is often appreciated.

While the main contribution of this paper is to offer a broader account of wrongful expertise that accommodates but goes beyond agnotology, the paper makes a secondary contribution to the agnotology literature itself – by situating it within an explicitly moral framework: Whereas we have a highly developed literature on what makes, say, the practices of the tobacco industry wrongful epistemically, accounts of their moral wrongfulness lag behind. Existing moral critiques of the tobacco strategy point to the ultimate consequences of factual ignorance on population health or the welfare of marginalised groups as the wrong-making feature. By analysing the relevant practices as forms of subversion, this paper identifies what is wrong, in the first instance, with the strategies documented in the agnotology literature: they are wrongful firstly because they disrespect the moral agency of the subverted audience, and secondly because of their ultimate harmful social consequences – subversive expertise thus involves two injustices.

Ultimately, my aim in developing this framework of subversive expertise is to illuminate actual cases of political wrongdoing, and of climate deregulation in particular. As such, I use the framework to analyse a climate assessment produced by the Trump Administration to justify deregulating vehicle emissions (NHTSA, 2018), dismantling a key US climate mitigation policy. I demonstrate, contra ubiquitous charges of ‘bias’ and ‘anti-science’ levelled at it, that the Administration produced a highly reliable assessment: endorsing the scientific consensus on climate change and conceding that the Administration’s proposed policy will exacerbate climate change more so than all other feasible policies, on every timescale considered and on every metric of risk. The puzzle left by this conclusion – namely how such an apparently damning assessment could plausibly lend support to the Administration’s policy – is answered by shifting away from the empirical reliability of the assessment to its normative and motivational impact. The assessment supports deregulation, I argue, by deploying its empirically reliable content to assault the normative and motivational bases for mitigation: the assessment tempts its reader to accept greater emissions, noting that the harms of the policy will only marginally worsen an already bleak future, that the harms to which the policy contributes cannot be directly attributed to the policy, and that these harms will not disproportionately fall within the borders of the United States, among other subversive discourses. To understand the ways in which scientific expertise can be (and has been) deployed to prop up unjust policies, even Trump-era deregulation, we must look beyond empirical subversion towards other modes of morally subversive expertise.

I proceed as follows. Section 2 introduces the basic concept of moral subversion, distinguishing its three modes of empirical, normative, and motivational subversion. Section 3 develops the framework of subversive expertise, building on the basic concept. Section 4 uses the framework to analyse a case study from the Trump Administration’s climate assessments, exemplifying normative and motivational subversion in expert communication. Section 5 concludes.

2 Moral subversion

At its core, moral subversion is an intuitive concept: it refers to activities that induce others to fail morally. The aim of this section is to unpack this concept in more detail, developing Jeffrey Howard’s (2016) recent account, which forms the starting point for my discussion.Footnote 1

2.1 Howard’s account of subversion

Howard (2016) offers an account of moral subversion in the context of analysing what goes wrong, morally speaking, in cases of criminal entrapment: where a state agent induces a person to commit a crime in order to prosecute them. E.g., when a police officer bribes, persuades, or otherwise incentivises a suspected gang member to commit a criminal offence as part of an investigation leading to their arrest. Rather than locating the wrongness of entrapment in the prosecution and eventual punishment of the entrapped defendant, as prevailing accounts of entrapment do (ibid., pp. 25–28), Howard focuses instead on the initial act of incitement itself, arguing that entrapment is wrong, in the first instance, because it involves moral subversion, defined as follows: ‘To subvert an agent’s moral capacities is to interfere with the agent’s practical reasoning in ways that increase the likelihood she will culpably choose to act wrongly’ (ibid., p. 25). For the purposes of this paper, I will abstract away from the context of entrapment that inspired Howard’s discussion and focus purely on understanding the underlying notion of moral subversion, ‘what goes wrong when any moral agent induces another moral agent to act wrongly’ (ibid., p. 29).

Howard’s objection to subversion is grounded in a duty of respect for moral agency (ibid., pp. 29–31). Howard contends that respecting a person’s moral agency involves taking up appropriate attitudes towards their agency, in light of its value, and regulating one’s conduct by norms that reflect the relevant attitudes (ibid., pp. 29–30). The relevant attitudes appropriate towards moral agency, i.e., that convey respect for moral agency, are recognition and support: recognition of others’ abilities to reason about right and wrong, and support for the exercise of their abilities, helping them ‘live up to morality’s demands’ (ibid., p. 30). Setting up others to fail morally is thus incompatible with this second attitude of support for the successful operation of others’ moral capacities (ibid., p. 25). Subversion, in short, violates a duty of respect that we owe to each other qua moral agents.Footnote 2 The upshot of the respect-based analysis of subversion is this: When we think about the wrong of subversion, e.g., in standard cases of incitement to violence, we typically point to the ultimate consequences and victims of violence, the harms caused by the subverted agent’s actions. However, on the respect-based analysis, the subverted agent themself is also wronged – subversion thus typically involves two injustices. Importantly for Howard, though the subverted agent has a complaint against their subverter for tripping them up morally, since subversion does not involve coercion, or other exculpatory factors, their own responsibility for wrongdoing is not diminished; hence Howard’s insistence that the subverted agent ‘culpably choose[s] to act wrongly’ (ibid., p. 25, see also pp. 30–33).

Having identified disrespect for moral agency as the wrong-making feature of subversion, Howard proceeds to distinguish two broad categories of subversive activity: motivational subversion and epistemic subversion. Motivational subversion transpires when an agent ‘either creates a reason for another agent to act wrongly, or strengthens the practical force of a temptation the agent faces to act wrongly’ (ibid., p. 31). Motivational subversion can take the form of material incentives for wrongdoing, e.g., money, or communicative practices that render a wrongful act more attractive or, conversely, render a moral duty less attractive. Epistemic subversion, by contrast, transpires when an agent ‘causes the formation of false beliefs in another agent’ (ibid.). The beliefs in question include beliefs about the substance and role of morality. E.g., when a person convinces another to adopt norms of racial supremacy, or to rank moral reasons on par with non-moral reasons, ‘rather than decisive and regulative’ (ibid.). They also include beliefs about morally relevant facts, such as empirical facts about climate change, factory farming, drug safety, etc., that are relevant to the application of moral principles. In short, subversive practices work to undermine others’ moral agency by either targeting their beliefs or motivations.

Let us now consider the scope of activities that properly qualify as subversive on Howard’s account. There are plenty of activities that increase the likelihood of others’ wrongdoing which we would not want to condemn as subversive. Easy cases include those where an agent does not know and cannot reasonably foresee the effects of their actions on others. For example, when I sincerely believe that I am alone, I have a habit of talking to myself out loud. Suppose unbeknownst to me, my friend is in the other room and overhears a claim that leads them to act wrongly. Such cases fall outside the scope of subversion on Howard’s account; subversive activity is that which ‘foreseeably increases the likelihood’ of others’ wrongdoing (ibid., p. 37, emphasis in original), a rider that importantly covers negligent subversion (ibid., pp. 33–37), where an agent ought to have known the subversive effects of their actions, but excludes the easy cases exemplified above.

There are, however, harder cases where our actions foreseeably increase the chance of others’ wrongdoing, but which we would also want to exclude from the scope of subversion. To see this, consider two examples that Howard offers: Jones, a black man, walks down a street frequented by white supremacists, thereby increasing the likelihood they will harm him; a government justly raises taxes on the rich, increasing the likelihood they will wrongfully evade them (ibid., pp. 31–32). It seems perverse to subsume such cases under subversion, as that would amount to saying, e.g., in the first case, that the white supremacists have a complaint of justice against Jones for disrespecting their moral agency. To exclude such cases, Howard distinguishes between morally protected and unprotected activities (ibid., p. 32). To say that an activity is morally protected is to say that the agent has a moral right or duty to engage in the activity, e.g., Jones has a morally protected right to freedom of movement, and the government has a moral duty to pass just tax laws. Morally protected activities fall outside the scope of subversion, even if they foreseeably increase the chance of others’ wrongdoing. Thus, we arrive at Howard’s more precise definition: ‘Subversion, then, is otherwise morally unprotected activity that foreseeably increases the likelihood that others’ first moral power will culpably operate defectively’ (ibid., p. 33).Footnote 3

These three elements – the analysis of the wrong of subversion in terms of disrespect of moral agency; the different modes of subversion; and the scope of activities that qualify as subversive – form the core of Howard’s account of moral subversion. Before applying the concept to the analysis of wrongfully deployed expertise, I will strengthen the element that will figure most centrally in the subsequent discussion but is underdeveloped in Howard’s account: the analysis of the different modes of subversion.

2.2 Three modes of subversion

Recall that Howard distinguishes epistemic and motivational modes of subversion. While I will borrow Howard’s definition of motivational subversion unchanged, I will modify the definition of epistemic subversion, and further distinguish the category into empirical and normative varieties, giving us three modes of subversion in total.

For Howard, epistemic subversion transpires when an agent ‘causes the formation of false beliefs in another agent’ (ibid., p. 31, emphasis added). This definition, however, is too restrictive – true belief has other ‘contraries’ beyond false belief (Goldman, 1999, p. 5). Following Guerrero (2007, p. 63), let us distinguish three states of ignorance where an agent lacks true belief: (I1) A person is ignorant because they have never thought about an issue, and therefore hold no explicit beliefs about it: ‘has no beliefs about F or only has unexamined ‘implicit’ beliefs about F’. (I2) A person is ignorant because, having thought about an issue, they have come to hold false beliefs about it: ‘believes that not-F when in fact, F’. (I3) A person is ignorant because, though they have thought about an issue, they do not know what to believe and have come to suspend judgement: ‘doesn’t believe that F or that not-F’.Footnote 4 Note that only I2 is captured in Howard’s definition of epistemic subversion. Clearly, however, agents can be induced to act wrongly by being rendered ignorant in the first sense, keeping them unaware of an issue; or in the third sense, being made ignorant through engendering doubt and suspension of judgement on an issue. Indeed, one of the main contributions of the agnotology literature regarding the manufacture of ignorance in science has been to show how industries successfully halted regulatory action in more subtle and effective ways than the propagation of outright falsehoods, but rather through inducing the public to suspend judgement on scientific matters despite the existence of a scientific consensus, as I will review shortly. As such, let us broaden the definition of epistemic subversion from false belief to ignorance: epistemic subversion transpires when an agent renders another ignorant of the content or role of morality or of morally relevant empirical facts. Furthermore, let us distinguish epistemic subversion into two sub-modes, empirical and normative, to separate the analysis of factual versus moral beliefs going forward, giving us three modes of moral subversion:

  1. 1.

    Empirical subversion transpires when an agent renders another ignorant of morally relevant empirical facts.

  2. 2.

    Normative subversion transpires when an agent renders another ignorant of the content or role of morality.

  3. 3.

    Motivational subversion ‘transpires when an agent either creates a reason for another agent to act wrongly, or strengthens the practical force of a temptation the agent faces to act wrongly’ (Howard, 2016, p. 31).

To close this discussion, let us ground these distinctions in an everyday example before moving to the context of expertise. Consider a person rightly convinced that eating meat – or, at least, eating some form of meat, e.g., factory farmed chicken – is wrong. A subversive ‘friend’ may induce them to consume meat along any of the following routes: by misleading them about morally relevant empirical facts: e.g., claiming that a new study shows that chickens do not feel pain (empirical subversion). Or they may go directly for the norm in question, misleading them about the content or role of morality, e.g., arguing that animals, despite feeling pain, are not worthy of moral consideration (normative subversion). Or they may tempt their friend to consume meat by chipping away at their motivation to live up to morality’s demands, by saying something as simple as, ‘But you used to love eating chicken!’, or by placing their vegetarian friend in a context that breaks their resolve, e.g., by inviting them to a dinner where refusing meat would shame them (motivational subversion). Finally, note that although these modes are conceptually independent, as the example illustrates, they are often entangled in practice, with the same subversive act undermining moral agency along multiple modes, e.g., subverting both moral convictions and motivations simultaneously, as will become clear in the case study (Sect. 4).Footnote 5

3 Morally subversive expertise

With the basic concept in place, I now turn to a particularly potent form of subversion, where scientific expertise is deployed to undermine moral agency; what I call, morally subversive expertise, analysing it along the three modes of subversion distinguished above. By ‘deployment of expertise’, I mean the use of one’s own epistemic advantage (specialist knowledge or ability) or that of others. I.e., I mean to cover both cases where experts appeal to their own authority as well as cases where non-experts appeal to the authority of experts, e.g., when spokespersons for an industry appeal to the expertise of friendly scientists.

Before proceeding, I note that the focus of this section is limited in two ways. Firstly, I limit the discussion of subversive expertise to scientific expertise. Of course, other forms of expertise can be deployed subversively, e.g., a journalist or historian, deferred to by readers for their expertise on a political issue, can empirically subvert their audience. Second, I focus on subversive uses of science in the regulatory context. Again, scientific expertise can be used to subvert agents in contexts that do not concern state policies and regulations. A self-help book may normatively subvert readers by, say, drawing on evolutionary biology to normalise unjust social hierarchies in human society. This choice of focus reflects the aims of the paper – namely, to illuminate the role of science in cases of regulatory failure, and of climate inaction in particular – while noting that the concept of subversive expertise extends beyond my focus.

3.1 Empirically subversive expertise

Expertise is empirically subversive if it is deployed in ways that foreseeably render an audience ignorant of morally relevant empirical facts, increasing the likelihood that they will choose to act wrongly.Footnote 6

This mode captures the familiar epistemic wrongdoing of industries identified in the bias/agnotology literature that dominates analyses of regulatory failure at present. I will review the main insights of this literature before noting its limitations and the need to attend to other modes of subversive expertise to account for political inaction on issues such as climate change.

Scholars of scientific assessment for policy note that, in the twentieth century, complex regimes of assessment were institutionalised as an integral component of public policy in modern states (see Oppenheimer et al., 2019, Chap. 1). Pamuk puts this point succinctly when she writes, ‘It is a core assumption of modern states that claims to truth in policy contexts must be scientific’ (Pamuk, 2021, p. 151). As such, for industries under regulation, science represents ‘both a danger and an opportunity’ (Franta, 2022, p. 556), a source of power that must be aligned with an industry’s pronouncements, a fact keenly noted by industry strategists, as an early industry playbook for avoiding regulation makes strikingly clear:

Regulatory policy is increasingly made with the participation of experts, especially academics. A regulated firm or industry should be prepared whenever possible to coopt these experts. This is most effectively done by identifying the leading experts in each relevant field and hiring them as consultants or advisors, or giving them research grants and the like. This activity requires a modicum of finesse; it must not be too blatant, for the experts themselves must not recognize that they have lost their objectivity and freedom of action. (Owen and Braeutigam 1978, p. 7, cited in Franta, 2022, p. 556)

In recent years, scholars have mapped this landscape of co-option, uncovering strategies of empirical subversion that deploy scientific expertise to render publics and policymakers ignorant of morally relevant facts. The tobacco industry, a paradigm example in this literature, created what they referred to as a ‘stable’ of expert witnesses (Galison & Proctor, 2020, p. 31), marshalled to defend their product in public and in court. Scientists hired by the tobacco industry worked for decades to dispute the empirical basis for tobacco regulation, manufacturing uncertainty on the science linking smoking – and second-hand smoking – to lung cancer, heart disease, and other illnesses, and attacking the reputation of scientists who demonstrated such links (Fernandez Pinto, 2017; Michaels & Monforton, 2005; Proctor, 2011). Typically, the deployment of external experts was done covertly, via third-party agencies and law firms who handled the recruitment of experts and their training (Fernandez Pinto, 2017, p. 58), so that the tobacco industry could claim them as independent experts ‘without anyone knowing who was pulling the strings’ (Proctor, 2011, p. 207). In addition to mobilising external experts, the reliable but inconvenient results of in-house studies, evidencing, e.g., the addictive properties of nicotine, were withheld from the public – by shutting down entire research divisions to supress leaks, in extreme cases (Proctor, 2011, pp. 263–267). Beyond challenging the science linking cigarette smoking to cancer and other diseases directly, the industry also funded vast swathes of legitimate research in virology, genetics, biochemistry, and other disciplines that might be expected to reveal alternative causes to the diseases in question – over seven thousand papers were published with grants from the Council of Tobacco Research, the main research arm of the tobacco industry (Galison & Proctor, 2020, p. 30). By funding such ‘decoy’/‘distraction’ research (Proctor, 2011, pp. 261–263), the tobacco industry could point to an endless list of alternative causes for the diseases linked to smoking, placing the onus on regulators to rule out such alternative causes, thus perpetually raising the burden of proof for regulation. Since regulatory agencies are typically required by law to respond to submitted evidence (Michaels, 2008, p. 73), this strategy of flooding the system with ‘noise’ proved highly effective in buying time and creating a public impression of a highly contested scientific controversy which had in fact long been resolved (Galison & Proctor, 2020, p. 31).Footnote 7

The ‘tobacco strategy’, as it has since been dubbed (Oreskes & Conway, 2010, p. 6), has been deployed by numerous industries to undermine the empirical basis for regulation, obscuring everything from the risks of pharmaceutical drugs to the industrial causes of environmental degradation, e.g., acid rain and ozone depletion and, most significantly in recent years, the link between fossil fuels and climate change (see, e.g., Fernandez Pinto, 2017; Franta, 2022; Michaels, 2008; Oreskes & Conway, 2010).Footnote 8 ‘Doubt is our product’, and ‘time to come up with more confusion!’, are among the leaked industry memos that encapsulate the aims of the strategy (Michaels, 2022, p. 20; Oreskes & Conway, 2010, p. 34). Despite their diversity, the strategies in question all belong to what I am calling empirically subversive expertise: the deployment of expertise along the lines of the tobacco strategy undermines moral agency by misleading audiences with respect to empirical facts. Clearly, understanding this mode of subversive expertise is critical to an understanding of regulatory failure, particularly in the case of climate change. Yet, the near-singular focus in the literature on empirical subversion neglects other ways in which scientific expertise can and has been mobilised to contest and dismantle protective regulation, limiting our understanding of deregulatory phenomena (see Mansfield, 2021). As such, it is crucial to recognise how empirically unimpeachable expertise can still be deployed to undermine just regulation by shaping normative commitments and moral motivations directly.

Before moving on, however, I will cash out an important insight that follows from situating the tobacco strategy within an explicitly moral framework. Although a great deal of scholarly attention has been devoted to the tobacco strategy – and forms of empirical ignorance-making involving scientific expertise, more broadly – the focus of the literature has been on providing epistemic accounts of wrongdoing. I.e., on explaining what makes the various elements of the tobacco strategy – doubt mongering, strategic dissent, character assassination, etc.– wrongful in terms of their detrimental effects on the production of knowledge within expert communities or the uptake of knowledge among non-expert audiences. Such epistemic accounts are highly developed, drawing on everything from virtue epistemology (e.g., Biddle et al., 2017) to agent-based models (e.g., Weatherall et al., 2020) and much more. By contrast, moral accounts of the tobacco strategy have lagged behind, and have tended to focus on the ultimate risks posed to certain communities who are disadvantaged by the social and political consequences of manufactured ignorance/dissent (e.g., Leuschner, 2018; Leuschner & Fernández Pinto, 2021). By analysing the tobacco strategy, and similar empirically subversive strategies, as forms of moral subversion, this paper strengthens existing moral accounts. Such empirically subversive strategies are morally wrongful not only because of their harmful consequences for ultimate victims, but because, in the first place, they disrespect the moral agency of the subverted agents (recall that subversion involves two injustices, an immediate injustice against the subverted, and a final injustice against those harmed by the subverted agent’s actions). E.g., in addition to those harmed by unmitigated climate change, subverted publics in developed nations, say, the US, have a complaint of justice against those scientists who did the bidding of the oil industry for ‘undercut[ing] their quests to be just persons’ (Howard, 2016, p. 25).Footnote 9

3.2 Normatively subversive expertise

Expertise is normatively subversive if it is deployed in ways that foreseeably render an audience ignorant of the content or role of morality, increasing the likelihood that they will choose to act wrongly.

In familiar cases where an industry attempts to defend its product, or where a regulatory agency defends its policy, basic moral assumptions are typically uncontested: e.g., industries typically contend that their products are safe when used responsibly by consumers, granting that if proven harmful it would be wrong to sell them. Tobacco companies routinely made such commitments, e.g., when a tobacco company executive announced that his company would ‘stop business tomorrow’ if ‘we had any thought or knowledge that in any way we were selling a product harmful to consumers’ (Proctor, 2011, p. 257). The existence of such moral common ground is precisely why industries tend to focus, at first, on empirical subversion, contesting claims that their products are harmful, as surveyed above.Footnote 10 Note, however, that industries could take a different route: they could contest the moral ground directly, and scientific expertise can furnish the empirical premises in such moral arguments. They could say, e.g., ‘Granted, we’re exacerbating climate change, but our scientists have run simulations showing that the majority of those harmed by climate impacts are people outside our borders and, since national interest is what matters, mitigating climate change isn’t a priority for us’. Arguments of this kind defend a product/policy not by obscuring or denying inconvenient scientific findings, but by deploying such findings as premises in arguments that generate moral ignorance. This, in essence, is what normatively subversive expertise denotes. To get a better sense of this mode in action, consider the following passage from a 1995 advertorial in the New York Times by the oil giant Mobil Corporation (to become ExxonMobil following a merger):

[…] to those who think [the fossil fuel] industry and nature cannot coexist, we say show a little respect for Mother Nature. She is one strong lady, resilient and capable of rejuvenation. The environment recovers well from both natural and man-made disasters. When Mount Saint Helens erupted in 1980, it destroyed 200 square miles of land and 19 million old-growth trees. Just 15 years later, much of the area has recovered. New generations of elk have learned to feed in new-growth forests rather than old growth. Coho salmon, thought unable to survive in the now unnaturally warm waters of the area, are thriving. Even Alaska’s Prince William Sound survived the 1989 oil spill and today is biologically alive thanks to the remarkable powers of nature. (Mobil, 1995)

In this passage, Mobil does not defend fossil fuels by contesting their role in climate change, as it has done elsewhere (see Supran & Oreskes, 2017). Rather, Mobil grants that, at least to some degree, its products are responsible for climate and environmental degradation, and attempts instead to reshape readers’ moral conception of what an attitude of respect for nature enjoins. Activists desperate to protect Mother Nature from fossil fuel emissions fail to properly respect her, Mobil contends, since doing so involves recognising her profound resilience and acting in a manner consistent with that recognition – namely, continue emitting. The advertorial cites a range of ecological findings to support its key premise: nature is resilient. I suggest that the deployment of expertise in this passage strikes us as subversive not primary because it misleads readers about empirical matters – though the specific claims about Mother Nature bouncing back from oil spills and the health of Coho salmon, etc., are surely contestable. The deployment of expertise is subversive, in the first instance, because it marshals a set of ecological findings with the aim of reshaping audiences’ moral conceptions, persuading them to endorse and guide their actions in accordance with a false environmental ethic, and this is what is distinctive about the example.Footnote 11

Let us now turn to the final mode of subversive expertise, where a speech act involving the expression of a scientific claim incentivises wrongdoing.

3.3 Motivationally subversive expertise

Expertise is motivationally subversive if it is deployed in ways that either create a reason for another agent to act wrongly, or strengthen the practical force of a temptation the agent faces to act wrongly.

Beginning with a simple example, consider a demographics expert advising an official on where to spend a development budget. The expert presents the politician with the following information: ‘We’ve crunched the numbers from the previous election. We find voter turnout in deprived neighbourhoods is low. If you spend the money on richer areas instead, you’ll win the next election’. Assume that this information is accurate. What objection might we have to this form of expertise? Strictly speaking, the expert is not attempting to reshape the politician’s moral commitment, they are not claiming that those in deprived neighbourhoods are less worthy of moral consideration, so this is not normative subversion. They are, rather, creating a practical incentive for the politician to act wrongly. This speech act functions essentially like a bribe.

Moving on to the climate context, consider what Lamb et al. (2020) have recently dubbed, ‘discourses of climate delay’– forms of speech that accept the reality of climate change but demotivate effective regulation. These discourses are compelling, because they tap into ‘legitimate concerns and fears’ (ibid., p. 5), and are often promulgated by actors with a genuine concern for climate change, although they can be cynically deployed. They include discourses of ‘surrender’, e.g., where an agent describes in excruciating detail the challenges posed by climate change, the vanishing chances of success, the historical record of failed negotiations, the widespread greenwashing by governments and corporations, the prospect of having already crossed tipping points, etc. Surrender discourses are typically accurate in what they describe, faithfully citing the latest scientific assessments. Yet, in certain contexts, such discourses reliably trigger a sense of despair in audiences, sapping their motivation for action.

Before moving on let me say something more about which surrender discourses in particular I consider problematic, as their presentation in Lamb et al. (2020) is quick, and it is important not to condemn all such discourses as motivationally subversive. In a paper entitled, ‘Hope from Despair’, Huber (2023) argues persuasively that certain expressions of despair are critical in guarding ourselves and our listeners from succumbing to false hope, and can ultimately help us hope well. Huber distinguishes between ‘episodic’ and ‘fundamental’ despair, where the former is compatible with ultimate hope and action, and is the form of despair deployed by activists, culminating in action – think, e.g., of Greta Thunberg’s famous speech: ‘I don’t want you to be hopeful. I want you to panic. […] And then I want you to act’ (Thunberg 2019, cited in Huber, 2023, n. 3)– whereas fundamental despair destroys ‘our basic underlying sense that the future is open to our interventions’, undermining ‘practical agency as such’ (Huber, 2023, p. 82). The sort of surrender discourses that ought to be condemned as motivationally subversive are those that engender fundamental, rather than episodic, despair.

Beyond surrender discourses, Lamb et al. (2020) describe other discourses of delay that include the use of elaborate statistics comparing carbon footprints of different nations or projects, demotivating those with relatively low carbon footprints from acting first, emphasising the unfairness of current burden sharing arrangements and the likelihood that other actors will free ride on one’s efforts, etc. I will not unpack the motivationally subversive effects of each, but simply note their salient feature: they typically include an appeal to reliable science, framed and presented in ways that motivate a wrongful act or, conversely, demotivate a moral duty.

With this framework of subversive expertise in place, I now turn to analyse a central case of climate deregulation under the Trump Administration. As I will argue, without recognising the many of ways in which expertise can be deployed to undermine moral agency beyond empirical subversion, we cannot account fully for the strategic use of science in deregulatory politics.

4 Beyond empirical subversion: the trump administration’s climate assessments

During its tenure, the Trump Administration made dismantling US federal climate and environmental laws a strategic priority (Davenport, 2020a; Jotzo et al., 2018; Pitt et al., 2020). According to various regulatory trackers, the Administration succeeded in rolling back over 100 environmental regulations set by previous administrations – regulations aiming at preserving natural habitats and wildlife, protecting human health from industrial pollutants and chemical toxins, and mitigating climate change (Davenport, 2020b; Popovich & Plumer, 2020).Footnote 12 As required by law, underpinning these deregulatory actions are detailed environmental assessments analysing the foreseeable impacts of the Administration’s proposals and justifying them accordingly. Despite their significant political legacy, however, the role of environmental assessments in supporting Trump-era deregulation remains understudied. In particular, ubiquitous charges of ‘bias’ and ‘anti-science’ levelled at the Administration’s agencies, charging that its deregulatory actions are based on empirically unreliable assessments, have gone largely unchallenged.Footnote 13 The following analysis cuts against the grain of these dominant charges, revealing a more complex– and more problematic – deployment of scientific expertise in the Trump Administration. As such, I will briefly rehearse the dominant empirical subversion charge to set the scene.

To many observers, the science-policy interface in the Trump era was marked by ubiquitous empirical corruption, with scientific facts traded for alternative facts, and doubt and denial of the scientific consensus on climate change affirmed at the highest levels of the Administration (Becker, 2019; Bomberg, 2017; Desikan et al., 2023; Lavelle, 2020; Warren, 2019). Beyond publicly dismissing inconvenient science and erasing it from government websites (Dillon et al., 2018, p. S92; see also Hook, 2018; Milman & Morris, 2017), the Administration actively undermined the ability of its own agencies to deliver reliable assessments by dismissing or marginalising hundreds of competent scientists, replacing them with industry-funded experts (Perls, 2021, p. 593); appointed industry lobbyists to head regulatory agencies (Selby, 2019, pp. 471–474); defunded critical research projects (Jotzo et al., 2018, p. 813) and interfered with independent scientific assessments (Davenport & Landler, 2019). In short, the Trump Administration’s relationship to science represents, as Frickel and Rea (2020, p. 66) vividly put it, an ‘anti-science disaster’, to be studied ‘as other disaster researchers might study the impacts of a drought, hurricane, or wildfire’. A natural assumption to make, in light of these well-documented political pressures on science, is that the alignment between the Administration’s environmental assessments and deregulatory policy is straightforwardly due to empirical subversion: the Administration’s assessments prop up its policies by ‘cooking the books’ (Lavelle, 2020), or otherwise manipulating, doubting, or ignoring relevant facts. As I will argue, the story is more complex: other modes of subversive expertise played a central role in propping up the Trump Administration’s deregulatory policies.

To make this case, I analyse the Environmental Impact Statement (EIS) (NHTSA, 2018) developed by the Administration to justify rolling back emission standards for passenger cars and light trucks, a major source of global CO2 emissions, dismantling the Obama Administration’s keystone climate mitigation policy. Against reflexive charges of ‘anti-science’, I will show that the EIS in fact accurately represents the state of scientific knowledge on climate change. Furthermore, the EIS accurately captures the climatic harms of the Administration’s proposed policy: comparing it to a broad range of alternatives, the EIS concedes that the Trump Administration’s ‘preferred alternative’ exacerbates climate change more so than all other feasible policies, on every timescale considered and on every metric of risk. As such, empirical subversion utterly fails to capture the way in which the EIS lends support to climate deregulation – yet it was commissioned by the Trump Administration precisely for this purpose. The alternative analysis of the EIS I present reveals the role of normative and motivational modes of subversive expertise in supporting climate deregulation.

I start with some background, contextualising US regulation of vehicle emissions which the Trump Administration sought to dismantle (Sect. 4.1). I then examine the scientific assessment – the EIS – produced by the Administration to justify deregulating vehicle emissions despite their climate impacts, rejecting the empirical subversion diagnosis (Sect. 4.2), arguing instead that the EIS supports deregulation by deploying its empirically reliable content to assault the normative and motivational bases for mitigating climate change (Sect. 4.3).

4.1 Background: regulating and deregulating tailpipe emissions

Historically, federal legislation in the US aimed at curbing vehicle emissions developed in response to a host of national environmental and economic concerns. Global climate change emerged as a key driver of federal vehicle regulation only in recent years, during the Obama Administration (Nesbit et al., 2016, Sect. 2). Given the focus of this paper, this section aims solely at summarising the climatic rationale for reducing vehicle emissions, in order to contextualise the Trump Administration’s deregulatory action and its appeal to climate science.

According to the 2019 US greenhouse gas (GHG) inventory, the transportation sector is the leading source of US emissions, emitting 1.9 billion tons of CO2 equivalent (GtCO2e), or 29% of total US emissions (EPA, 2019). Significantly, unlike emissions from other key sectors which show declining trends, transportation emissions have increased 23% between 1990 and 2017 (ibid., p. 3–21), rising every year between 2012 and 2019 (Wang & Ge, 2019). Within this sector, passenger cars and light-duty trucks (hereafter, vehicles) contribute the lion’s share, accounting for 58% of US transportation emissions (EPA, 2019, p. 3–21). I.e., vehicles contribute roughly one fifth of US GHG emissions. To put their significance into global perspective, analysing data from 2017, Oppenheimer and Duffy (2018, p. 3) put it this way: ‘If emissions from US passenger cars and light trucks were considered their own country, they would rank 7th-largest worldwide: bigger than Germany and Indonesia and slightly smaller than Brazil’. As such, any serious attempt by the US to mitigate climate change must address vehicle emissions.

Beginning in 2009, the Obama Administration directed the Environmental Protection Agency (EPA) and the National Highway Traffic Safety Administration (NHTSA) – the two federal agencies with overlapping mandates to set vehicle regulation – to develop a national policy that would improve vehicle fuel efficiency and, in turn, reduce GHG emissions. On the first page of the joint EPA and NHTSA proposal, climate change was cited, for the first time in federal vehicle emissions regulation (Davenport, 2020a; Nesbit et al., 2016), as the main rationale for curbing emissions: ‘[…] responding to the country’s critical need to address global climate change and to reduce oil consumption’ (EPA & NHTSA, 2009, p. 49,454). Over the following years, the two agencies released rules that gradually increased fuel efficiency standards.Footnote 14 For our purposes, the relevant standards that were in place as the Obama Administration left office required car manufacturers to improve the fuel efficiency of their vehicles on average by 5% annually (EPA & NHTSA, 2012, p. 62,679), with increased efficiency (more miles travelled per gallon) translating into emissions reduction.

In 2017, an alliance of major automakers, as well as oil-industry lobbyists, urged the Trump Administration to reconsider the EPA/NHTSA rules of the Obama Administration in favour of less stringent emission standards (Becker, 2019; Puko, 2020; Tabuchi, 2018; White & Shepardson, 2018). The Trump Administration obliged. In 2018, Trump’s EPA and NHTSA proposed a new fuel economy standard, abandoning the requirement of 5% annual increase in efficiency (for vehicle model years 2021–2026) for a new standard of 0% (EPA & NHTSA, 2018). In other words, freezing emission standards indefinitely at, 2020 levels, removing federal incentives for the auto industry to produce cleaner cars going forward. The standard of 0% was a step too far for the auto industry, which had hoped for a less stringent target than the Obama Administration’s, not an abandonment of efficiency requirements altogether (Becker, 2019; Davenport, 2020a).Footnote 15 The Trump Administration ultimately settled on a standard of 1.5% (EPA & NHTSA, 2020), shifting the US auto market from one of the most ambitious fuel efficiency standards globally to one of the weakest (Davenport, 2020a). Crucially, as part of this rollback, the Trump Administration revoked California’s famous ‘waiver’ (EPA & NHTSA, 2019), which allows the state to set its own vehicle emission standards at a more ambitious level than those set by the federal government – a hard-won exemption dating back to the Clean Air Act of 1970, granted to California in recognition of the urgency of its air pollution crisis (Claiborne, 2019, pp. 11–13). Later amendments to the Clean Air Act extended the exemption to other states, allowing them to choose between either federal standards or California’s more ambitious standards (Nesbit et al., 2016, p. 17). Since then, over a dozen states have adopted California’s standards (Popovich & Plumer, 2020), covering more than 40% of cars sold in the US under stricter emission standards than federal policy (Shepardson, 2019). By revoking California’s waiver, the Trump Administration made the federal government the sole regulator of vehicle emissions, effectively forcing over a dozen states to abandon their independent efforts to mitigate climate change and limit urban pollution.

In summary, then, whereas the Obama Administration’s policy aimed at gradually raising federal vehicle emission standards to the standards of California, the Trump Administration’s policy aimed to do the opposite: to lower federal standards, and force California and its allied states to lower their standards in turn. The name given to this two-part deregulatory action is the Safer Affordable Fuel-Efficient (SAFE) Vehicles Rule. To many analysists, SAFE constitutes the most significant climate rollback completed by the Trump Administration (Davenport, 2020a; Davenport & Landler, 2019; Pitt et al., 2020; Puko, 2020; White & Shepardson, 2018).

US law requires federal agencies proposing actions ‘significantly affecting the quality of the human environment’ to prepare ‘a detailed statement’ addressing, among other things, the ‘environmental impact of the proposed action’, including a comparison with a reasonable set of policy alternatives (the National Environmental Policy Act of 1969, NEPA). Accordingly, the Trump Administration prepared a lengthy, 500-page Environmental Impact Statement (EIS) to defend SAFE’s environmental credentials (NHTSA, 2018). At a time when calls for radical emission cuts to mitigate climate change intensify globally, how did the Administration defend a deregulatory policy which, as we shall shortly see, results in a massive increase in emissions from its most polluting sector? Let us turn to this question.

4.2 Is the EIS empirically subversive?

Given the problematic nature of the science-policy interface in the Trump Administration, surveyed above, it seems reasonable to suspect the EIS of empirical subversion: to assume, as many of its critics did, that the Administration’s avowed climate scepticism and its deregulatory agenda distorted the assessment of SAFE’s impacts, resulting in empirically unreliable findings that vindicate its proposed policy as environmentally responsible, against the facts. In this section, I consider and reject the empirical subversion diagnosis. The conclusion raises a puzzle: namely, how a reliable assessment which acknowledges the risks of climate change and the harmful contributions of SAFE could plausibly be deployed in its support. Answering it requires an alternative analysis that looks to other modes of moral subversion (Sect. 4.3).

To evaluate the empirical subversion diagnosis, I focus on the EIS’s answer to three key questions relevant to political deliberation over SAFE in light of climate change: (1) What is the state of scientific knowledge on climate change? (2) What impacts will SAFE have on climate change? (3) How do SAFE’s climate impacts compare to a reasonable set of policy alternatives? While government agencies have a degree of flexibility in producing EISs with respect to the range of issues they address, contextualising a policy, detailing its predictable environmental impacts and comparing them against a reasonable set of policy alternatives is a statutory obligation (NEPA, Sect. 102(2)(C). I.e., these three questions address the basic empirical facts which set the legal bar for an EIS, and we should regard the EIS as empirically subversive if it generates ignorance with respect to any of them. E.g., we should regard the EIS as empirically subversive if, in answering (1), it casts doubt on the reality or severity of anthropogenic climate change, exaggerates uncertainties in climate knowledge, etc. And likewise, if, in answering (2) and (3), the EIS denies or underplays the magnitude of SAFE’s impacts, denies that feasible alternatives exist or that they have better climate credentials than SAFE.Footnote 16

Let us begin with the EIS’s answer to the first question, its assessment of the state of scientific knowledge on climate change.

Discussions of the evidence for anthropogenic climate change are threaded throughout the EIS, reflecting the fact that climate change is a crosscutting issue in environmental assessments and cannot be dealt with neatly in one chapter. All in all, the EIS devotes over 100 pages to the assessment of climate change, making it by far the most carefully investigated environmental consideration in the EIS. The main chapter where the findings of climate science are summarised, Chap. 5: ‘Greenhouse Gas Emissions and Climate Change’, begins with a note of epistemic deference to authoritative expert panel-reviewed reports on climate change, such as the IPCC’s (NHTSA, 2018, p. 5 − 1), affirming the trustworthiness of their conclusions, noting that they have ‘endured a more thorough and systematic review process than information on virtually any other topic in environmental science and policy’ (5–20). The EIS goes on to describe a broad set of detected phenomena that ‘provide evidence for rapid climate change’ (5–10): e.g., extreme temperatures; global and regional sea-level rise; drought; ocean acidification; retreat of glaciers and ice sheets; thawing of the permafrost (5–13–17). With the basic evidence for climate change in place, the EIS then proceeds to enumerate hundreds of specific global and regional risks predicted by climate science: e.g., mass species extinction (8–70); ‘catastrophic’ impacts due to the release of methane from the permafrost, among other tipping points (8–67–72); increased prevalence of disease (8–57); increased risk of drought, wildfires, and flooding (8–55); disruption of infrastructure essential for maintaining healthcare access (8–55); etc.

It is worth noting that the EIS does not merely list climate risks without context. Rather, it offers a rich discussion of their implications and distribution across various populations, threading considerations of climate justice throughout the EIS. ‘Across all potential impacts’, we are told, ‘disadvantaged groups such as children, elderly, sick, and low-income populations are especially vulnerable’ (8–55). Diminishing air quality due to climate change mostly affects ‘impoverished communities’ (8–57). Heat stress is particularly risky for ‘pregnant women and their fetuses’, as it may increase ‘preterm birth, decrease birth weights, and increase the rate of stillbirths’ (8–56). Burdens of ‘distress’ and ‘adverse mental health consequences’ from ‘climate-related disasters’ will predominantly fall on ‘first responders’, the ‘economically disadvantaged’ and the ‘homeless’, ‘women’, ‘children’, among other groups (8–58).

In addition to accurately laying out the first-order evidence, the EIS also offers policymakers important meta-reflections on the state of scientific knowledge, emphasising the tendency of the scientific community to underestimate the risks of climate change, especially when producing consensus reports, such as the IPCC’s (see 5–13, 5–40 and Figure 5.2.2-2). I.e., not only does the EIS offer an accurate answer to the question, ‘What is the state of scientific knowledge on climate change’, it also warns policymakers that the climate assessment of the EIS likely reflects the conservativism of the scientific literature it relies on.

Let us then set aside the empirical subversion diagnosis in relation to the first question. Still, for policymakers to appropriately assess SAFE in light of climate change, it is not enough for them to be well-informed on the state of climate science. They also need to know whether SAFE will exacerbate climate change (question 2) and whether there are reasonable alternative policies to SAFE with better climate credentials (question 3). Let us turn to these more specific questions.

In short, the EIS offers candid answers to these questions as well, concluding that SAFE is more harmful to the climate system than every alternative assessed on every metric of climate risk in all simulated timescales. Firstly, the EIS compares SAFE’s impacts to those of the Obama Administration’s policy. The EIS estimates that vehicles produced under SAFE will emit in their lifetime approximately an additional billion tons of CO2 compared to if they had been produced under Obama-era standards – about the sum of present annual emissions from Germany and the UK. Looking beyond the lifetime emissions of these vehicles to the long-term, cumulative impacts of SAFE on US vehicle emissions to 2100, the EIS predicts that SAFE will lead to an increase of approximately 8 GtCO2e– more than a decade’s worth of emissions from Germany (Ritchie et al., 2020) and ‘more than a century’s worth of emissions from Portugal’ (Oppenheimer & Duffy, 2018, n. 17).Footnote 17 Importantly, the EIS translates these technical numbers for the benefit of lay audiences: the emission increases in a single year from adopting SAFE vs. maintaining the Obama policy is roughly equivalent to the emissions resulting from adding 9 million vehicles to the road (NHTSA, 2018, p.14).

Secondly, in addition to comparing SAFE’s climate impacts to the Obama policy, the EIS also compares SAFE to 7 ‘reasonable alternative’ policies less stringent than the Obama policy but more stringent than SAFE (i.e., the EIS considers 7 intermediate fuel efficiency standards falling between the Obama Administration’s 5% and SAFE’s 0% standard). The EIS concludes that SAFE is more harmful to the climate system than every alternative assessed on every metric of climate risk in all simulated timescales (5–29–47). Importantly, such apparent indictments of SAFE are not hidden away in obscure appendices or footnotes, rather they are announced in the heart of the summary for policymakers, cashed out in comprehensible terms (e.g., number of additional vehicles equivalent to a quantity of emissions), and represented visually (see, e.g., Fig. 1).

Fig. 1
figure 1

Increase in emissions in terms of vehicle equivalence from SAFE (Alt 1) compared to the Obama policy (x-axis) and 7 intermediate alternatives (NHTSA, 2018, Figure 5.4.1-1)

Full size image

The Trump Administration issued the EIS to support SAFE. But how could such an empirically reliable assessment, with its ‘startling’ admissions on climate change (SEEC, 2018), possibly lend support to SAFE?

4.3 Normative and motivational subversion in the EIS

To answer this question, I turn now to the discursive strategies embedded in the EIS which organise its empirically reliable content into an assault on the normative and motivational bases for mitigating climate change. This assault on moral agency is achieved mainly through four connected discourses, which may be condensed as follows: ‘The world is heading for disaster anyway’; ‘We’re adding a mere drop to an ocean of impacts’;Footnote 18 ‘The most severe impacts will fall outside our borders’; ‘Specific climate impacts cannot be traced back to this policy – so no liability’. I will unpack each discourse in turn.

The EIS reviews efforts to mitigate climate change to date, as well as ‘reasonably foreseeable future actions’ based on the promises and pledges of various governments, and concludes that, most likely, they will fail to avoid extreme levels of climate change. The EIS begins its analysis by assuming a baseline scenario that takes the world as it is, assuming current levels of policy ambition, and asks what climate change will look like by 2100 on ‘the current climate trajectory’ (5–28). The EIS concludes that, by 2100, atmospheric CO2 concentrations will reach 789 parts per million (ppm), resulting in global warming of 4°C relative to preindustrial levels (5–31, cf. IPCC, 2023, p. 11). Next, the EIS adopts a more optimistic scenario based not on the current climate trajectory but on a trajectory that factors in pledges and international efforts ‘already underway or reasonably foreseeable’ (8–21–25). Still, the EIS finds these international pledges woefully inadequate to limit global warming to 2°C (5–30), let alone 1.5°C, projecting a global warming of 3.3°C by 2100 (8–28).

The EIS then calculates the increment that SAFE would add to this grim outlook, and finds that SAFE would add 1 ppm to CO2 concentrations, would deplete the remaining carbon budget consistent with 2 °C by a further 1%, add another 0.003 °C to global warming, etc. The EIS thus concludes that SAFE’s harms ‘are extremely small compared with total projected future climate change, they would only marginally increase the potential risks associated with climate change’ (S-20); that climate change will be ‘exacerbated but only to a marginal degree’ (S-21); that SAFE ‘would result in marginal increases in the already anticipated increases’ in climate risk (8–38, emphases added), etc. It is hard to overstate the pervasiveness of the first two discourses throughout the EIS: phrases such as, ‘very small compared to total’ (5–46), ‘extremely small in relation to’ (S-14), ‘marginal degree in proportion with’ (S-21) are threaded throughout the analysis, impressing upon its reader that ‘drastic reductions in all US sectors and from the rest of the developed and developing world’ are needed to bend the global emissions curve (5–30), tempting them to regard mitigation as futile and resign themselves to a policy that would marginally worsen an already bleak outlook (9 − 1). The most rhetorically powerful manifestation of these two discourses, however, is not in the text of the EIS, but in its figures. For each climate variable assessed, the EIS visually drives home the drop-in-an-ocean narrative. As shown in Fig. 2, the EIS ranks SAFE as the worst policy among all feasible alternatives with respect to its impacts on global warming (top panel); then, by adding the total anticipated warming on the world’s current level of ambition, the figure erases the differences between SAFE and its alternatives (bottom panel).Footnote 19

Fig. 2
figure 2

Top: SAFE’s (Alt 1) contribution to global temperature increase is highest among alternatives. Bottom: Impacts of SAFE, the Obama policy and 7 alternatives on global temperature ranked in terms of their comparative impacts on the global trend (NHTSA, 2018, Figure 5.4.2-3 and Figure 5.4.2-4)

Full size image

While the first two discourses – ‘The world is heading for disaster’; ‘We’re adding a mere drop to an ocean’– play the dominant subversive role in the EIS, they are further propped up by discourses that deploy other (true) features of the climate predicament to subvert its reader.

For example, the EIS reminds its reader, at various points, that causation in climate change is complex. Once released into the atmosphere, individual packets of GHG emissions mix together and jointly cause climate impacts (more accurately, jointly increase the likelihood of impacts). As such, tracing a specific impact back to a specific policy choice is, as the EIS puts it, ‘inappropriate’: ‘it is inappropriate to identify increases in GHG emissions associated with a single source or group of sources as the single cause of any particular climate-related impact or event’ (8–38, n. 13), since they ‘alone would neither cause nor prevent climate change’ (8–38). In case the ‘no liability’ message was missed in these quotes, taken from the technical chapters, the EIS delivers the basic point bluntly in the summary: ‘the emissions resulting from this rule cannot be tied to any particular climate impact’ (S-21).Footnote 20 Furthermore, the EIS notes that the impacts of climate change, to which SAFE will contribute, will disperse globally and will be felt unevenly, and will not fall mainly within US borders. Thus, the EIS rounds off its summary of its key findings with the sentence, ‘The impacts of the Proposed Action […] would occur on a global scale, and would not disproportionately affect the United States’ (S-14, emphases added).

Through these discourses, the EIS offers a description of the climate predicament, grounded in reliable expertise, that renders an injustice more attractive and undermines its readers’ normative commitments. The discourses summarised above trade on the ‘fragmentation of agency’, the ‘dispersion of causes and effects’ and the ‘institutional inadequacy’ of global mitigation frameworks – all the peculiar elements that make climate change a ‘perfect moral storm’, where a number of factors converge in such a way so as to predispose agents towards wrongdoing (Gardiner, 2011, p. 24). Needless to say, these discourses would not pass moral scrutiny: The fact that past mitigation efforts have failed and existing policies and pledges remain thoroughly inadequate gives us reasons to strengthen mitigation efforts, not abandon them; and recognising that the contribution of even the most harmful individual climate policy pales when cast against a global trend tells us only what we already know about climate change – that it is a global challenge requiring comprehensive mitigation from all. And there is little debate among climate ethicists that affluent nations have special obligations to aggressively mitigate climate change; on basic questions of this sort, irrespective of one’s conception of climate justice, ‘all roads lead to Rome’ (McKinnon, 2022, p. 55). Knowing this, however – and reporting on my own ‘research encounter’ with the EIS – does little to blunt the rhetorical force of these discourses which press, page after page, on one’s predisposition to fail morally.

5 Conclusion

My aim in this paper has been to develop a framework of morally subversive expertise and apply it to shed light on the science underpinning climate deregulation. I distinguished three modes of subversive expertise, capturing broader ways in which expertise can undermine moral agency beyond misleading audiences about empirical facts. The framework of subversive expertise accommodates the dominant, agnotological framework in the literature on industry bias and deregulatory science, with its focus on the use of expertise to generate factual ignorance in an audience, situating it alongside other ways in which expertise can be used to trip up others morally. To further explore the framework and show its fruitfulness, I analysed a case of climate deregulation under the Trump Administration. I argued that a broader analysis, attending to the normative and motivational dimensions of subversion, and not just the empirical, is necessary to capture the use of science in deregulatory politics. The lesson of the analysis is that dismissing the expertise that underpins climate deregulation as empirically corrupt ‘anti-science’ both obscures its actual role in the politics of climate change and understates its wrongfulness: it misses the breadth of the assault on moral agency that sustains climate injustice.