1 Introduction

On 25 July 2023, the European Commission (EC) announced the creation of a phase-out roadmap for regulatory animal testing in its jurisdiction, in response to a demand of the European Citizens’ Initiative “Save Cruelty-Free Cosmetics” (EC 2023). However, the EC also refused one of the initiative’s demands: that it shall commit to creating a general phase-out roadmap for all animal experimentation in the European Union. This verdict came two years after the European Parliament had approved a non-legislative resolution in favour of a general phase-out plan for animal experimentation (EU Parl 2021). It however mirrored the EC’s opposing position to an earlier European Citizens’ Initiative with similar demands (EC 2015).

The EC’s reason for differentiating between regulatory animal testing and research is that the former serves a more limited purpose. Regulatory chemical safety assessments require testing for specific endpoints. Thus, eliminating animal testing in this area is a matter of providing alternative methods to test for these specific endpoints or of revising the endpoints themselves (EC 2023, p. 13). When it comes to research, by contrast, the potential applications of animal experiments are not so limited. The EC stated: “[…] at this stage, it is not possible to predict when scientifically valid methods able to replace particular animal procedures in research will become available. Consequently, setting reduction goals seems unrealistic and these would need to be constantly adjusted” (EC 2023, p. 17). The EC also emphasized that, in its view, animal research remained indispensable for safeguarding the health of humans and ecosystems (EC 2023, p. 19; see also 2015, 10). Burdening the open-ended endeavour of research with specific reduction goals would thus threaten the societal benefits of science.

While the EC’s case is partly legal (there is no legal basis for setting reduction goals, see EC 2023, p. 17), its persuasiveness crucially rests on a moral argument: We ought not to burden animal research with reduction goals because such burdens will lead to avoidable harms. Lives will be lost that could have been saved, suffering will be endured that could have been eased. Therefore, we ought not to specify reduction goals for animal research. Call this the “argument from avoidable harms.” As a moral argument with real-world impact, it merits philosophical scrutiny, but a debate about it is yet to be had. Philosophers, perhaps because they already pay relatively little attention to animal experimentation compared to other areas of animal use, have not devoted much thought to the issue of phase-out planning.

This article discusses the argument from avoidable harms in order to show three things: First, the argument rests on a strongly nonegalitarian view of animal moral status that not everybody in the debate can be expected to share. Thus, disagreements about the advisability of phase-out planning for animal research can reflect underlying differences in moral outlook, not just in assumptions about science and strategy. The debate about phase-out plans is also a moral debate (Sect. 2). Second, the argument from avoidable harms rests on an exceedingly narrow conception of what a phase-out plan can be. It only applies to phase-out plans based on prohibitions, even though a range of non-prohibitive measures are available (Sect. 3). Third, the argument’s premises indeed imply that ban-free phase-out planning is morally obligatory, because these premises construe animal research as a moral conflict, and moral conflicts ought to be avoided whenever possible (Sect. 4). Thus, while authorities like the EC may be drawing the correct conclusion from their own premises when it comes to rejecting a ban-based roadmap for the end of animal research, they should commit to developing ban-free phase-out plans (Sect. 5).

2 Against Phase-Out Planning: The Argument from Avoidable Harms

The argument from avoidable harms is probably the most important moral argument against phase-out planning for animal research.Footnote 1 In the following, let me first state the argument in its basic, though not most compelling version, and then refine it further. In its basic version, the argument reads:

  1. 1.

    We should avoid incurring avoidable harms.

  2. 2.

    But by committing to a phase-out plan for specific research methods, we would incur avoidable harms.

  3. 3.

    So we should refrain from committing to a phase-out plan for specific research methods.

Apart from the EC’s communications (EC 2015, 2023), arguments along these lines take centre stage in the campaigns of groups for the defence of animal research (EARA 2023; EFPIA 2023; SID 2023) and in researchers’ warnings against restrictions on animal experimentation (Grimm 2019; Loeb 2019; Aerts et al. 2022). The argument is easy to recognize by its strong conclusion: Not only is there no moral obligation to plan the phase-out of animal research, but there is an obligation not to do so because it would threaten health and safety.

From a philosophical perspective, the argument raises a range of difficulties. In this and the next section, consider the two premises in turn.

Premise (1) seems straightforward enough, as it can be drawn from a general principle of beneficence or nonmaleficence, some version of which is accepted by practically any moral theory (see Beauchamp and Childress 2012, Chap. 5). The expression “incurring harm” is used ambiguously here between “causing” and “letting happen,” but the claim seems broadly plausible under either reading. Its plausibility furthermore does not seem to depend on a specific conception of harm or a particular theory of welfare (see e.g., Holtug 2002).

For example, it should be uncontroversial that one should turn off one’s stove before leaving the house to avoid burning it down along with the other residents. But this example involves, first, a clearly identifiable evil whose avoidability is known (burning down the house), and second, a precaution whose effectiveness is known (turning off the stove). Uncertainty complicates everything. It is by no means obvious how much precaution is demanded by general beneficence or nonmaleficence in cases where a harm’s avoidability or the precaution’s effectiveness is unknown and uncertain. It seems plausible that beneficence and nonmaleficence give us what I propose to call “Hail Mary duties” – obligations to try our best, despite unclear prospects of success, given high stakes. If a large asteroid is headed straight for Earth, we should surely try to knock it off its course even if we cannot know whether doing so is possible and whether we are making any significant progress. But Hail Mary duties raise many boundary issues: How high do the stakes have to be for these duties to arise, how unlikely may the success of our efforts be, and how much can these duties demand of us?

Unfortunately, conducting research in order to address potential future harms is a matter of just these hard-to-pin-down duties. Imagine a researcher, or a group of researchers, who is tasked today to save the life of a particular patient, suffering from an as yet untreatable disease, fifty years in the future. This task is more like knocking an asteroid off course than turning off a stove – there is no guarantee that the task is possible to do and that any particular contribution will be effective in the end.

Consider that several levels of uncertainty remove research work from actual harms and benefits (see generally Grimm et al. 2017). First is the uncertainty of the individual study’s contribution to knowledge. Experimental studies do not directly answer research questions, but gather data points that, together with many others, may over time point in the direction of answers. A single study’s contribution to the progress of research is not just difficult to quantify and measure, but even difficult to conceptualize (see also LaFollette and Shanks 1996, 254–55). The extent of the contribution also certainly depends on factors outside the researcher’s control (such as the extent to which results are read and taken seriously). Second, there is the uncertainty of knowledge’s contribution to societal benefit. Other factors, such as markets and regulation, determine what benefits can actually be realized. Third, there is the uncertainty of the distribution of societal benefits among particular individuals. Not everybody benefits equally when “society” benefits. There is no way for researchers to ensure that the societal benefits they hopefully help generate reach those that need them. Thus, there is never a guarantee that a particular study will benefit particular beneficiaries.

A defender of animal experimentation could object that, while the benefits of research cannot be pinpointed accurately, the enterprise of animal research as a whole leads to societal benefits with a high degree of likelihood.Footnote 2 We may not know the exact future harms that are prevented by today’s animal research, but we can be reasonably certain there are some. However, research can also cause rather than prevent unforeseen future harms. In Michael Dummett’s words: “What sane man, magically given the ability, in 1900, to foresee the nuclear weapons which it would make possible, would not have opted, if given the power, to prohibit all future research in physics?” (1981, 291). What is more, any profile of research conducted also causes opportunity costs in terms of other avenues of scientific exploration not pursued. So, in addition to the uncertainties that remove research work from direct societal benefit, there is the added uncertainty whether unforeseen benefits will outweigh unforeseen harms and opportunity costs.

Thus, research work as an effort to (on balance) avoid severe future harms must be understood as a Hail Mary effort – an effort whose feasibility and prospects of success are unclear but which is justified by the high moral stakes. Accordingly, premise (1) of the argument from avoidable harms must be understood to assert a Hail Mary duty:

  1. 1'.

    We should do our best to avoid severe future harms, even if we cannot know if they are avoidable or whether our efforts will be effective.

However, in contrast to the asteroid example, animal research involves conflicting interests. Animal research harms some – the animals – and may someday benefit others. This raises another difficult question: How should Hail Mary duties be weighed against ordinary duties of nonmaleficence (the kind where we know what to do in order to effectively avoid harming others)?

If human research participants are involved, classic standards of research ethics such as those of the Nuremberg Code, the Belmont Report and the Declaration of Helsinki (BR 1979; NC 1949; WMA 2001) assign lexical priority to the claims of the would-be victims of research at the expense of the claims of its potential future beneficiaries.Footnote 3 Indeed, it seems intuitive that we should not be too quick to harm some for the sake of highly uncertain and potentially futile efforts to prevent harm to others, even if this latter harm is more severe. Hail Mary duties generally weigh less than ordinary nonmaleficence duties.Footnote 4 But just how much less is difficult to say.

The conflict of duties would vanish, of course, if we denied that animals have any moral claims to human nonmaleficence in the first place. This is, however, hard to reconcile with any approach to animal ethics other than particularly harsh versions of anthropocentrism, the view that only humans matter and animals have no moral claims at all. This is a supremely controversial view in animal ethics, and relying on it would consequently turn the argument from avoidable harms equally controversial. Furthermore, the view that animals deserve protection for their own sake is already explicitly endorsed by many jurisdictions (Kotzmann 2023), including the European Union (in preamble 12 of Directive 2010/63/EU).

Proponents of the argument from avoidable harms are better advised to appeal to a nonegalitarian view of animal moral status (see e.g., Kagan 2019). Such views agree that there are duties, including duties of nonmaleficence, to both humans and animals, but contend that duties to animals generally weigh less than duties to humans. The relative weight could then be construed such that nonmaleficence duties to humans are significant enough to override Hail Mary duties, but nonmaleficence duties to animals are not. Of course, further questions could be raised about the precise reasoning behind this all-too-convenient view, as well as about its implications. For the moment, however, we can conclude that the first premise of the argument from avoidable harms implies a strongly nonegalitarian view of animal moral status, because there must be some reason why animals’ claims can be overridden for the sake of research, understood as a Hail Mary effort to avoid severe future harms, while the claims of humans cannot.

This alone reveals a noteworthy feature of the debate about phase-out planning: Proponents and opponents of phase-out planning might be in disagreement about the first premise of the argument from avoidable harms. Proponents might adhere to a more egalitarian view, concluding that animals have strong enough moral claims to override Hail Mary duties towards future humans and thus that animal research can legitimately be restricted. Meanwhile, opponents’ views may be more nonegalitarian or even anthropocentric, and thus they think that Hail Mary duties prevail.Footnote 5 This would imply that the debate about phase-out planning is in part a debate about values, not purely a debate about the scientific enterprise or about the most advisable strategy to move forward. Thus, the importance of scientific expertise in this debate, while it should not be understated, should also not be overstated. When it comes to the question whether Hail Mary duties justify harming animals, scientists have no claim to expertise and their opinions should not get more consideration than anyone else’s. Another implication is that, if the societal trend is towards more egalitarian regard for animals, the argument from avoidable harms will lose some of its perceived persuasiveness in public debate.

3 The Second Premise: Planning a Phase-Out Incurs Avoidable Harms

The argument’s second premise claims that planning a phase-out of specific research methods incurs avoidable harms. In light of the foregoing discussion, this must be understood as the claim that researchers’ collective Hail Mary efforts, justified by the high stakes involved and the relative unimportance of animals’ claims to nonmaleficence, would be hampered if an exit strategy were pursued. In essence, science cannot create its benefits if we take away its tools.

One possible objection against this premise is that science’s animal-based tools are far from perfect to begin with.Footnote 6 As Akhtar (2015) points out, relying on results from animal studies can incur rather than prevent harms if those animal studies have poor translatability to humans. For example, substances may be falsely assumed to be safe for human application, potentially effective treatments may be abandoned because they are unsafe for some other animals, and more promising avenues of human-relevant research may not be funded because costly animal studies eat up limited resources (2015, 415). Thus, on balance, reliance on animal studies might do humans more harm than good.

However, the argument from avoidable harms can be refined to avoid this objection. Defenders of the argument could concede that animal studies have their downsides, but argue that these downsides stem from excessive reliance on animal studies, not from their availability in science’s toolbox. Even if animal studies are currently utilized in such an ill-conceived way that they do more harm than good, one might argue that the way forward is to improve translatability and responsible utilization, not to eliminate animal studies as a source of insight altogether. The word “eliminate” is important here. It is because a phase-out plan takes a set of scientific options away without ensuring that there are sufficient replacements that, on balance, it can be argued to incur avoidable harms. Thus, a refined version of the second premise could read:

  1. 2'.

    By committing to a phase-out plan for specific research methods that relies on eliminating them without sufficient replacements, we would incur avoidable harms.

The question, of course, is why a phase-out plan would necessarily be so destructive.

At this point, it is important to clarify the notion of a “phase-out plan” or “roadmap.” Unfortunately, neither the European Citizens’ Initiatives demanding a phase-out nor the EC’s responses to them provide an explicit definition (see EC 2015; 2023). Common sense suggests that the term “roadmap” denotes a strategy for how to reach a distant goal by intermediate steps.Footnote 7 The European Union has already specified the goal, namely the end of all animal experimentation, in Directive 2010/63/EU. The EC has reasserted this goal in its responses to both initiatives demanding a phase-out plan (EC 2015, p. 2; 2023, 17). What is missing for a full-fledged roadmap are the intermediate steps: specific reduction targets and the measures by which to meet them, whatever they may be.

Why, given this broad notion of a “plan” or “roadmap,” are phase-out plans taken to be so destructive and dangerous? A revealing piece of communication was a public statement against the European Citizens’ Initiative “Save Cruelty-Free Cosmetics” released by the Dutch foundation Stichting Informatie Dierproeven (SID) on 10 July 2023, titled: “An abrupt ban on animal research stops access to latest, lifesaving medication” (SID 2023). The statement warned of a “general ban on all animal research,” asserting that “biomedical progress would stop” and “research for animal welfare, food safety, biodiversity and the environment will stall.” The imagery invoked here is of a plan that involves bringing animal research to a sudden stop. Assuming, along with SID, that this sentiment is representative of opponents of phase-out planning, their worry must be understood to be a worry about bans, where a “ban” could be anything from a refusal to license, the introduction of animal research quotas, or the outright criminalization of some or all forms of animal research.

However, nothing in the notion of a phase-out plan implies that bans in this sense have to be involved. Intermediate milestones are, by themselves, just checkpoints that trigger some defined set of policy measures (see Müller 2024). If targets are met, some measures are triggered, such as the continuation of the successful programmes or moving forward to a next phase. If targets are not met, then other measures are taken, such as a reevaluation, reinforcement, or reorientation of programmes. But it is completely up to the authorities what any of these measures should be. Introducing bans would only be one option, and a particularly drastic one at that.

Inspiration for other policy measures can be drawn from the literature on transformative governance for sustainable economies. Kanger et al. (2020) provide a helpful overview drawing on the theoretical framework of the Multi-Level Perspective (MLP). The MLP argues that socio-technical transformations result from the interaction of changes at three levels (Geels 2002): At the micro level, socio-technical niches generate innovations that are subject to an evolutionary logic of variation and selection, with some breaking through and becoming new norms; At the meso level lie the established modes of interaction and problem-solving, also called the socio-technical “regime;” And at the macro level, there are the external conditions of society and technology, referred to as the socio-technical “landscape.” While some transformations occur by themselves, say, due to groundbreaking innovations, breakdowns in established regimes, or natural changes in the landscape, many desired transformations do not occur. Transformative governance aims to alter the conditions so that desired socio-technical transformations are more likely to occur (Chaffin et al. 2016).

Kanger et al. (2020, 5–7) identify six intervention points for transformative governance based on the MLP:

  1. 1)

    Niches can be stimulated, e.g., by support for research and development or creating favourable regulatory conditions;

  2. 2)

    The buildup of niches into trends that challenge the regime can be encouraged, e.g., by supporting innovative start-ups through the “valley of death,” creating innovation platforms, and promoting knowledge-technology transfers;

  3. 3)

    The current regime can be destabilized, e.g., by introducing bans on certain technologies – for bans are one possible measure –, by involving more niche agents in regime decision-making, or by restructuring key institutions;

  4. 4)

    The negative consequences of regime destabilization can be anticipated and remedied, e.g., by offering retraining programmes or financial support to those on the losing end of transformation (in animal research, managing the potential risks that come with increased reliance on human research participants also springs to mind);

  5. 5)

    Where different socio-technical regimes interlock, coordination can be provided to ensure that they change in tandem rather than holding each other back, e.g., modes of housing and mobility may be interdependent and need to be addressed simultaneously (as it concerns animal research, the interlocking regimes of research, industry, and education spring to mind);

  6. 6)

    Where the external landscape is human-made, it can be altered to favour the desired changes, e.g., by participation in international negotiations that set the regulatory conditions for the regime.

Although the aim of this article is not to provide a comprehensive, let alone a detailed overview of all policy measures that can be considered for phase-out planning,Footnote 8 even a superficial glance at the above options reveals that bans are neither the only, nor a particularly important measure in transformative governance. A phase-out plan could involve reduction goals, understood as checkpoints, that trigger different policy packages depending on whether targets are met or missed. In principle, a roadmap out of animal experimentation can consist entirely of non-prohibitive measures that escalate in scale the more they miss their predefined targets.

It is worth noting that authorities in Europe and elsewhere are already taking some of the measures on the list, particularly in the area of niche stimulation, often flying the flag of the 3Rs of “replace, reduce, refine” (see Neuhaus et al. 2022 for an overview). What is still missing is the integration of these measures into structured strategies with clearly defined milestones and path-dependent measures.

How plausible is it that strategic policy measures incur avoidable harms in the future, as premise (2) of the argument from avoidable harms asserts? In general, the complexity of transformations does not allow making bold predictions about how any regime will react to specific interventions. In principle, it could turn out that a sudden and total criminalization of all animal research leads to a self-organized emergency effort in science that innovates all the required non-animal methods in a surprisingly short timespan – necessity is the mother of invention, after all. This would be the best case scenario. The worst case scenario, described by SID above, is that scientific progress comes to a screeching halt and countless avoidable harms are incurred. Faced with these contrasting scenarios, it may be wise to assume that interventions are riskier if they aim at making the socio-technical regime more dysfunctional in order to promote the transition to a new, functional system. In essence, this is the worry about animal experimentation bans: They are drastic interventions in a broadly functional regime of science that, by themselves, do nothing to ensure continuing functionality.Footnote 9

However, destabilization does not have to imply dysfunctionality. Less drastic efforts to destabilize a regime, such as giving niche agents more of a say in regime decision-making and restructuring institutions to accommodate change, do not necessarily threaten functionality. What is more, all of the other intervention points discussed by Kanger et al. (2020) aim exclusively at building a new functional regime, not undermining or destroying the old one. The worst that could happen with such efforts is that public resources are wasted on well-intentioned, but ultimately futile transformation efforts. But this danger is even greater in the absence of a strategy. In sum, the argument from avoidable harms applies at best to phase-out plans based on bans, but many policy measures other than bans are available. As a general argument against phase-out planning, the argument from avoidable harms simply fails.

4 The Argument from Avoidable Moral Conflicts

As we saw in Sect. 2 above, the argument from avoidable harms has the broadest appeal when it is supported with a nonegalitarian account of animal moral status: Animals do matter for their own sake, and we do owe them duties of nonmaleficence, but these duties are not strong enough to outweigh our Hail Mary duties to avoid severe future harms. Therefore, the argument goes, we should use animals to generate knowledge that hopefully creates societal benefits that will reach those that need them.

A noteworthy presupposition of this argument is that animal research involves a conflict of moral duties. It may be a conflict between unequally strong duties on a nonegalitarian view, but it is a conflict nonetheless. This structure of moral thinking is mirrored in regulatory frameworks that make the approval of animal research licenses dependent on a positive harm-benefit analysis (see Pound and Nicol 2018). While such frameworks make no explicit claim to making the best moral decisions, the frameworks draw much of their moral legitimacy from giving apparent consideration to both sides of the moral conflict.

But it is exactly from this underlying conception of animal research as a conflict of moral duties that the strongest argument in favour of phase-out planning arises. To use the words of author Brigid Brophy in her discussion of animal experimentation: “The moral thing to do about a moral dilemma is circumvent it” (Brophy 1972, p. 125). In parallel to the argument from avoidable harms, there is thus an argument from avoidable moral conflicts:

  1. 4.

    We should avoid creating avoidable conflicts of moral duties.

  2. 5.

    But by failing to strategize for the phase-out of animal research, we create avoidable conflicts of moral duties.

  3. 6.

    Therefore, we should strategize for the phase-out of animal research.

Premise (4), which one could call “Brophy’s principle,”Footnote 10 concerns situations in which we have simultaneous moral duties but cannot observe them all. Brophy emphasizes that, although we may not be blameworthy for committing necessary evils, we should always prefer not having to commit them in the first place.

An underlying point was discussed by Bernard Williams: In contrast to conflicts between incompatible beliefs, we cannot resolve conflicts between incompatible duties without a “remainder” (1965, 117). While we can simply decide which of two beliefs strikes us as more convincing and abandon the other, a duty unobserved still exerts some normative force over the agent. “If I eventually choose for one side of the conflict rather than the other, this is a possible ground of regret” (1965, 110). For example, if one can save only one out of two people from a burning building, one is justified in feeling regretful about the person one did not save. Thus, the duty to save this person continues to affect the agent’s moral life even though a rational decision was made not to observe it.

Regret is rational even if the moral conflict is not due to any fault of the agent, Williams emphasizes: “While there are some cases in which the situation was his own fault, and the correct conclusion for him to draw was that he ought not to get into situations of that type, it cannot be believed that all genuine conflict situations are of that type” (1965, 117). If the occurrence of the moral conflict is however indeed the agent’s own fault, Williams notes that they have reason not just for regret, but for self-reproach (1965, 110). Thus, for example, we should avoid making simultaneous, incompatible promises, as it creates conflicts of moral duties we could have avoided from the outset. When we inevitably resolve these conflicts by breaking one of our promises, we have reasons to blame ourselves in addition to reasons to regret failing to fulfil the other promise.

Premise (5) asserts that failure to strategize for the phase-out of animal research creates future moral conflicts. Inaction now leads to difficult moral decisions later. Like its counterpart, the second premise of the argument from avoidable harms, premise (5) rests on certain assumptions about what a phase-out plan is and what it can achieve. Specifically, it rests on the assumption that phase-out planning – in the broad sense advocated above, i.e., as a system of policy measures coordinated along checkpoints towards an ultimate goal – is necessary and sufficient to identify the quickest feasible way out of the moral conflicts of animal research.

Of course, the development of phase-out plans that live up to this high aspiration is itself a large-scale scientific, legislative, and regulatory endeavour. In particular, it must be ensured that phase-out plans do not give rise to more moral conflicts than they avoid. But this is not a question of whether to do any phase-out planning, but of how to do it well. For, good planning involves a careful analysis of the status quo, an anticipation of potential unintended consequences, and precautions against them if necessary. Of course, as anywhere in science, success cannot strictly be guaranteed. Thus, premise (5) still suffers from a degree of uncertainty. A concerted societal effort to work out a feasible and promising phase-out plan for animal research would inevitably be a Hail Mary effort – we cannot know if the task can be accomplished and whether any particular contribution is effective.

Viewed in this light, however, the EC’s assessment that “setting reduction goals seems unrealistic and these would have to be constantly adjusted” (EC 2023, p. 17) suggests that the development of workable phase-out plans is either not recognized as the scientific matter that it is, or is viewed with a surprising degree of science pessimism. We could imagine a science pessimist making just the same claims as the EC, but about animal research instead of phase-out planning: It is far-fetched and unrealistic that any particular study will lead to particular benefits, and the arrangements of science in pursuit of these benefits would have to be constantly adjusted. Yet the EC resists this pessimist rationale when it comes to animal research, and given the high stakes of severe future harms, this might be wise. Sometimes, morality demands making a far-fetched and seemingly unrealistic effort that needs constant readjustment.

Now, given Brophy’s principle, the prospect of facing millions upon millions of moral conflicts should strike us as a severe future evil. They may be avoidable if we devote serious resources to the development of sound phase-out plans now. By the logic of the argument from avoidable harms, these high stakes should trigger Hail Mary duties. Thus, if we understand animal research as involving moral conflicts, we should recognize a Hail Mary duty to try our best to come up with feasible phase-out plans, even if we cannot be sure that our efforts will be effective. This means that, while there may be reasons for caution against ban-based phase-out plans, authorities like the EC should urgently support the development of workable, ban-free phase-out plans as a scientific endeavour at large scale.

5 Conclusion

This article has articulated the strongest argument against phase-out planning for animal research, the argument from avoidable harms. It was argued that the argument rests on the view that we must do our best to avoid severe future evils, even if their avoidability and the effectiveness of our precautions are unknown – the view that we have Hail Mary duties.

The first noteworthy upshot of this discussion is that only strongly nonegalitarian views of animal moral status can support giving preference to Hail Mary duties over duties of nonmaleficence to animals. Though less controversial than outright anthropocentrism, nonegalitarian views are unlikely to be shared by all in the debate. Thus, the debate about phase-out plans is likely in part a moral debate, not just a debate about science or strategy.

The second result is that the argument from avoidable harms rests on an exceedingly narrow conception of what a phase-out plan can be, namely, that it must be a plan based on bans. This is simply not true. Phase-out planning can draw on a diverse range of policy measures and can, if necessary, do entirely without introducing any denials of permits, any quotas, any criminalization, or any other prohibitive measures.

The third upshot was that the logic of the argument from avoidable harms implies that we indeed have Hail Mary duties to develop workable phase-out plans. This is because future moral conflicts, like future harms, are severe evils that we should do our best to avoid. Thus, the argument from avoidable harms indirectly bolsters the case for phase-out planning for animal research, rather than undermining it. In conclusion, authorities should urgently support the development of workable, ban-free phase-out plans for animal research, treating this as a major scientific project in itself.

A philosophical debate about phase-out planning for animal research is not yet being held. This is unfortunate, first, because the various available policy measures and their integration into strategies raise conceptual and normative questions.Footnote 11 Second, it is unfortunate because the political debate about phase-out planning for animal research is suffused with moral arguments and there is a need for clarification and critique. At the same time, this article also revealed an urgent need for social scientific research into large-scale phase-out programmes for animal research. For, only if proposals for phase-out plans that include few or no bans become more tangible can the political debate move beyond its current, immature state.