2. Some Orientation

Considerable scholarly attention in recent years deals with how AVs should be programmed to respond in emergencies when situations of inevitable accident arise. In emergency situations human drivers sometimes have to make decisions that engage ethical considerations; when AVs come to ply the roads similar choices will have to be pre-encoded in the vehicles’ software. Our focus is on dilemmatic circumstances where harm or damage will occur no matter what course of action an AV adopts. Typical scenarios involve what a machine should do if a young child darts into the street in front of it, leaving it no time to brake. This presents a choice between staying the course and hitting the child or swerving into a tree or another vehicle, thus posing a threat to persons in the primary AV and in any other vehicles they may hit. Here, whose interests properly enjoy priority make this a particularly challenging scenario. When the stakes involve the lives of all of those involved, the appropriate solution may evade moral consensus.

Discussions of such dilemmatic circumstances in the emerging literature are partly informed by their resemblance to a thought experiment that has been a staple of academic philosophy for half a century: the trolley problem.Footnote ⁴ In one well-known version of this, the driver of a trolley whose brakes have failed is confronted with the choice between doing nothing and letting the trolley continue on its course, in which case it will hit and kill five persons on the tracks ahead, or taking action by switching it to a side track where only one person will be killed. The impending advent of AVs, and in particular the dilemma sketched above, has given new relevance to the trolley problem.

Some have responded to this new literature by pointing out that there are disanalogies between the trolley problem and the ethical choices that will arise from adoption of AVs.Footnote ⁵ Others have pointed out that AVs will be so much safer than human-driven cars that situations like that set out above will seldom arise,Footnote ⁶ so the problem remains an arid academic one. A third reaction to the claim that literature on the trolley problem can help us in programming AVs is that society’s response to the sort of dilemma set out above will not be derived from philosophy but rather from existing law, both criminal and civil.Footnote ⁷

We adopt no position on the first of these objections. The dilemmata arising from the use of AVs may or may not amount to versions of the much-discussed trolley problem. However, they certainly constitute “whom-to-harm” problems, and that is our focus here. As for the second objection, we reject it. In an AV dominated world, even if a whom-to-harm (or whom-to-spare) decision of the sort set out above arises only once in every 10 billion miles driven, that will still amount to about 1,000 times per year.Footnote ⁸ That is sufficient to make the problem one that cannot be ignored and will have to be dealt with in AV software. Unavoidable questions of law and public policy arise here.

As for the third objection to thinking that scholarly writing on the trolley problem contains the solution to how to program AVs, it has some merit. Canadian responses to the AV whom-to-harm problem will be made against the background of both criminal responsibility and rules of civil liability, in particular the tort of negligence. This background includes as-yet unanswered questions about who bears legal responsibility for harms arising from crashes of AVs—the vehicle’s owner, its retailer, its manufacturer, its programmer, or the vehicle itself?Footnote ⁹ This legal framework may tend to push ethical analysis to the side. However, while existing legal structures, not to mention the input of the insurance industry, will provide important contributions to the resolution of the whom-to-harm problem posed by AVs, those seem unlikely to completely control the conversation. Some confirmation of this can be found in the response of the country that appears to have devoted the most thought to this subject to date, Germany. In 2016 the German Federal Minister of Transport and Digital Infrastructure appointed an ethics committee to come up with a draft code of responses to ethical questions presented by AVs. That committee comprised law professors, ethics professors, experts in various technical disciplines, a bishop, consumer groups, industry input, and a prosecutor.Footnote ¹⁰ We think that it would be unethical to deploy AVs on Canadian roads without resolving the question of how they should operate in whom-to-harm situations and that the Canadian debate leading to that resolution will involve participation at least as diverse as the German one.

Indeed—and here is where animals come in—we suggest it should be even wider. It should include animal advocates, or at least ethicists attuned to the animal problem. The shift to AVs will have a number of consequences for nonhumans, including a potential reduction in the rate of vehicular slaughter of wildlife along highways. AVs will be superior to human drivers in detecting and reacting to animals who venture onto roadways. Yet despite this possible reduction in car-caused animal slaughter, there will still be situations of unavoidable harm. Animals will be involved in whom-to-harm situations similar to those described above. Those setting the rules for AV software will have to decide where they fit in—if their interests matter at all.

Animals have so far played little part in discussions of AV whom-to-harm scenarios.Footnote ¹¹ As was the case with academic discussion of the trolley problem over the last several decades, recent analysis of the whom-to-harm problem presented by AVs has generally avoided scenarios involving endangered nonhumans. Yet the creature who darts in front of an AV might not be a small child but rather a duck or a dog, particularly but not exclusively in urban environments. Depending on the ethical framing, this might give rise to the question of whether the AV should be programmed to forge ahead and kill the nonhuman, or instead swerve, posing some increased risk of harm to the vehicle’s occupants, other drivers, nearby pedestrians, or valuable property.

We suggest that the discussion of how to program AVs to respond in whom-to-harm situations should not continue to ignore the problem of where nonhumans fit in. As mentioned earlier, we are moving into a time where we must rethink how we live on this planet. We face two global crises of our own making: climate change and what some have christened, however prematurely, the sixth extinction.Footnote ¹² Our biosphere is under threat and it is in our interests, as well as the interests of animals whose habitats and lives we are endangering, to find a less destructive way to live on our land and in our territories. If this is to be a comprehensive re-envisioning of how we live in this world, we must address how we are to program AVs with animals in mind. This means that animal advocates should devote some thought to formulating positions on this question and ensuring that those views are advanced.

Of course, it is possible that those who have made no explicit mention of animals in recent analyses of whom-to-harm problems might nevertheless regard themselves as having addressed the animal question; they might have done so simply by proceeding as though animals were included in their consideration of property. That is, some discussions of how AVs should be programmed to respond in emergency situations have addressed the question of how harm to property fits in. Since most legal systems regard animals as property, or at least as being eligible to be turned into property, it is conceivable that these discussions of how AVs’ emergency algorithms should regard property were intended to embrace the question of how they should regard animals.

This seems unlikely. Not all the animals whose interests should be present in such discussions are straightforwardly thought of as property, wildlife for instance. Moreover, these days some animals commonly bought and sold, or transferred as gifts, sometimes fit only uncomfortably into traditional notions of property. A recent instance of this is a decision of the Court of Appeal of Newfoundland and Labrador where, in the course of deciding which party should get possession of a couple’s beloved dog following their break-up, one judge commented at length on the need for the law to treat animate property differently than inanimate things, at least for some purposes.Footnote ¹³ And a 2015 amendment to the Civil Code of Québec states: “Animals are not things. They are sentient beings and have biological needs.”Footnote ¹⁴ We suspect that analyses of AV emergency algorithms, even those that expressly mentioned property, simply overlooked them.

Either way, we suggest that animals merit express and distinct consideration when these questions arise and that their treatment should be more nuanced than simply lumping them in with inanimate property. The lives and suffering of sentient nonhumans are considered important in other areas of law and policy as well as ethics. Indeed, as we will see in our ethics discussion, according animal interests equal consideration to relevantly similar human interests is a live debate and such a position is not easily dismissed merely on the grounds of species identity.Footnote ¹⁵ To seriously countenance such a position is to countenance something in deep tension with viewing the relevant animals as property. Even where animals are ranked as property, many areas of Canadian law recognize that some of them are a special class of property—one whose members have a moral status that dictates that they may not be subjected to suffering or distress without justification. With that in mind we offer some thoughts on how animals might be treated in AV whom-to-harm problems and AV software more generally.

3. How Animals Might Fare Under Law as it Currently Stands

In considering how animals’ interests might figure in AVs’ whom-to-harm software it will be useful to start by glancing at how those interests have been treated in Canadian law. In the past, whom-to-harm situations have arisen with human-driven vehicles, presenting drivers with agonizing decisions over whether to swerve to avoid a small creature in the road ahead—thus imposing increased risks on passengers and pedestrians—or maintain course and kill the animal. In cases where drivers have swerved to spare the animal they have sometimes harmed humans or their property. That has given rise to both criminal and civil court cases in which the legality of the drivers’ evasive action had to be assessed. Those assessments, found in reported decisions, may be of some use in informing the design of AVs’ pre-set algorithms for whom-to-harm situations.

We digress here to note that we presume that when driverless vehicles are deployed their whom-to-harm algorithms will be made mandatory—that is, not amenable to driver choice or modification but instead dictated by law. It might be thought that owners of AVs should be free to select their own software to deal with the whom-to-kill problem. One can imagine a world in which manufacturers and retailers of AVs offer purchasers a variety of software options for whom-to-harm scenarios: selfish ones for ethical egoists, which would opt never to harm the driver;Footnote ¹⁶ another for altruists; one for Kantians; another for Utilitarians;Footnote ¹⁷ perhaps even one for drivers who renounce human supremacy.Footnote ¹⁸ While this remains conceivable, it seems more likely that public policy will dictate a uniform solution, at least for most private vehicles.Footnote ¹⁹ Of course, currently human drivers do have choice. They elect how to respond in whom-to-harm situations. However, these choices sometimes result in litigation where the legality of those decisions is assessed ex post facto. In an AV world where considerations pertinent to such decisions can be set in advance, the legal system can impose its desired outcomes at the design stage. That suggests that there will at least be legalFootnote ²⁰ constraints to manufacturer and owner choice in selecting AV whom-to-harm algorithms. A further reason for compulsory software is that both in emergency and non-emergency situations it will be advantageous if AVs know how other AVs are programmed. This would be difficult if there were owner choice in the matter.Footnote ²¹ Finally, even if such uniformity is not imposed by law, it may effectively be dictated by the insurance industry.

In designing such software what lessons, then, might be gleaned from judicial decisions in whom-to-harm cases involving human drivers and nonhumans—in particular, from cases where drivers swerve or panic brake to avoid hitting and killing a small animal in the roadway and injury to humans ensues? From our review of the reported cases the first thing to note is that drivers often do take evasive action to avoid colliding with small animals, even when doing so increases risks to themselves and other humans. Presented with a small animal appearing in the road before them—an animal they could opt to hit with minimal consequence to themselves or their vehicle—they often either swerve to avoid a collision or jam on the brakes.Footnote ²² When harm to humans has resulted the law has generally been unforgiving: it has regarded swerving to be unreasonable, giving rise to liability under the tort of negligence. Indeed, steering to avoid a collision with a small animal has even been adjudged presumptively negligent, on a par with driving on the wrong side of the centre line.Footnote ²³ In the words of the Chief Justice of British Columbia in a case where the driver had swerved to avoid a dog, “If the choice is between an animal and human safety then … there is no real choice and the animal must suffer.”Footnote ²⁴

Where the driver’s alternative is between death to the animal and non-fatal injury to humans, the harm to humans is the greater wrong.Footnote ²⁵ A nice illustration of this appears in a case where a bus driver was held to have erred in successfully braking hard to avoid killing a dog but at the cost of relatively minor injury to a standing passenger who fell over.Footnote ²⁶

The holdings in these cases, and others, is especially notable in light of the fact that courts typically exercise a measure of leniency with respect to decisions made in emergency circumstances. For example, in exonerating the driver in Canadian Pacific Ltd. v. Gill (not an animals case) the Supreme Court of Canada noted that it is “trite law that faced with a sudden emergency for the creation of which the driver is not responsible he cannot be held to a standard of conduct which one sitting in the calmness of a courtroom later might determine was not the best course.”Footnote ²⁷

Tort law has a name for this: the “agony of the moment” doctrine. Here is a description from a leading Canadian text:

So, for example, in a case where a driver caused personal injury to humans by swerving to avoid a harmless inanimate object that suddenly appeared in the road in front of her the court found her action reasonable in the circumstance: “The average motorist will flinch from driving over a cardboard box or even a large sheet of paper; there is something in all of us that prompts reaction, by braking or swerving, to avoid any unnatural material in the road.”Footnote ²⁹

However, the doctrine does not seem to get applied in animal cases. The fact that drivers who harm humans by swerving to avoid inanimate objects are accorded greater forgiveness than those who do so when swerving to avoid killing animals gives rise to a suspicion that courts are anxious to avoid ever being seen as valuing animal life as in any way comparable to human life, or even human health.

All of this suggests that if law is to be the chief guideline for AV whom-to-harm algorithms then animals, or at least small ones, will be losers. On the roads most evasive action or emergency braking poses some increased risk to humans, even if only minimally, but running over a cat or squirrel does not. A whom-to-harm algorithm requiring that AVs treat harm to animals’ lives as lesser than any risk of injury to humans would effectively require AVs to run over a lot of animals, even where the risk (to humans) of trying to avoid doing so was small.Footnote ³⁰ This may not be shocking given how other aspects of Canadian law view animals’ lives. For instance, it subordinates those lives to humans’ gustatory interests. To offer another example, the regime that regulates biomedical experimentation in Canada regards the lives of laboratory animals as worth less than the slim chance of finding improved treatments for human afflictions, and this goes not only for fatal diseases like cancer and Alzheimer’s but for less serious conditions such as male pattern baldness or tattoo removal. That being the case, some persons may regard the AV whom-to-harm algorithms just described as entirely appropriate.

However, the possibility that a new technology which makes the world safer for humans might simultaneously make it less so for animals should at least give us pause. Currently, human drivers frequently brake or swerve in an attempt to avoid running over, say, a cat or a duck—often succeeding and preserving an animal’s life without harming any humans. In replacing humans with technology, it is important not to make the relevant decision space less humane. To do otherwise is to abandon what amounts to a rather conservative and basic moral orientation governing human-animal relations. If we are to support the view that AV whom-to-harm algorithms should reflect a greater regard for the impulse not to drive cars over small animals, we need to look beyond the law as it is currently understood and applied.

4. A Turn to Ethics

If we are right about the law as it currently stands, perhaps ethics provides a more fruitful domain in which to interrogate and trouble the kinds of values and evaluations that support or legitimize prioritizing the interests of humans to the exclusion of other animals. Though ethical judgments are not straight-forwardly relevant to law (except, perhaps, criminal law), they can be relevant to those creating laws or those, in liberal democracies like Canada, voting for others who, at their best, represent their constituencies within institutions where laws are created. Engaging ethics well does not require a commitment to foundational moral judgments or agreeing with socially prominent or dominant value commitments that support prioritizing some to the exclusion of others. Ethics need not be anthropocentric, unlike much of law, and so can provide a space for careful reflection that challenges the status quo.

Two fundamental approaches to our treatment of other animals that will concern us here talk in terms of either indirect or direct duties. Indirect duty views do not consider unnecessary harms to be morally wronging the animals concerned. Rather, who is wronged is either the agent themselves or others enjoying moral status (usually other agents taken to be relevantly similar to the agent acting wrongly).Footnote ³¹ Direct duty views do consider unnecessary harms to be morally wronging the animals concerned, even where the animals do not enjoy equal moral status.Footnote ³² Where animals do not enjoy equal status, human interests can take priority and so can reflect an anthropocentrism familiar to law.Footnote ³³ Greater specificity about direct duty views complicates such a claim. Direct duty views which index moral status to the capacities of those who can be morally wronged are unlikely to hold that all humans enjoy a moral status that exceeds that which is accorded other animals (e.g., human embryos or first trimester human fetuses).Footnote ³⁴ Even here, however, there will be humans who enjoy greater regard than both some other humans and other animals. We submit that the inclusion of animals in discussions of the future use and position of AVs in whom-to-harm scenarios can be defended using either fundamental approach, though one approach is much less likely to require a significant rethinking of how animals are regarded in law. Let us take each in turn.

Perhaps the best-known advocate of indirect duties to other animals in the modern era is Immanuel Kant.Footnote ³⁵ For Kant, animals lack moral status because they are not rational agents who purposefully act in ways that cohere with their values, desires, and reasons and who can subject their decisions to act to logical and ethical scrutiny. To use language adapted from Kant, animals are not ends-in-themselves.Footnote ³⁶ Again for Kant, animals cannot be wronged by being unnecessarily harmed, but rational agents can harm themselves and others if they act in ways that ignore relevant similarities between themselves and animals, similarities that support particular duties to fellow rational agents and require a cultivated sensitivity and responsiveness to such grounds to adequately fulfill those duties to fellow rational agents. This last part was put in terms of an avoidance of an insensitivity that negatively affected one’s moral agency or ability to act morally in the world, and could harm others in the process.Footnote ^37, Footnote ³⁸

So, Kant is not indifferent to what we take to be the mistreatment of animals, though for anthropocentric reasons. To mistreat relevantly similar animals risks developing a comfort with mistreating fellow humans (understood as fellow rational agents). Without good reason to act cruelly, or without regard for the relevant animal’s possible stress or distress, we should not so act. To do otherwise risks harming our own humanity; it is in our own self-interests not to do that and, to top it off, it is required by moral law.Footnote ³⁹ This commitment of Kant’s will be put to use momentarily, but first let us get some other self-interested reasons out of the way that can nevertheless impact how other animals fare in a world dominated by AVs.

The size of animals will matter when considering whether an AV should avoid or run them over. The bigger the animal the more clearly prudential it is to for the AV to avoid collision. A moose or deer is best avoided; to do otherwise risks human life and limb.Footnote ⁴⁰ But, depending on the size, a domestic pig is also best avoided. Presumably, AVs will come in various sizes, so even a pig can do enough damage to a vehicle to warrant avoiding them on roadways. We can rehearse this point where horses are used as transport, in or outside of the city (think of police horses versus recreational riding), though other humans now enter the decisional context. We can go a little deeper.

Religious practitioners are not impervious to demands to regard how other animals fare when they act in the world. Though this can be an other-regarding element of the relevant religion, it need not. Certain cosmologies attach negative significance to harming animals or positive significance to benefiting them (e.g., Buddhism and Hinduism). Where either significance involves lengthy residence in unpleasant or pleasant locales after death (e.g., hells or heavens), or unpleasant or pleasant future consequences for the actor, the self-interest is unmistakable. Even if only during a period of transition from current vehicles to AVs, depending on how responsibility for animal harm is understood in situations where machines make the decisions, religious practitioners may well be advised, or morally required, to attend to how AVs are programmed to avoid harming animals and weigh-in on those discussions before market or other forces become decisive.

Returning to Kant’s approach to animals, we now know that the connection between cruelty to animals and mistreatment of humans is complicated and not quite as Kant understood it. There is a growing consensus, for example, that animal cruelty can be a precursor to mistreating humans when the cruelty is not socially sanctioned or tolerated. Where it is socially sanctioned or tolerated, the connection is much less clear.Footnote ⁴¹ If true, this is an important discovery because what qualifies as “risky cruelty”Footnote ⁴² will differ, even if only slightly, across the world. In other words, even if only rarely, “risky cruelty” may be different depending on where you are in the world and so some forms of cruelty avoid being inescapably risky. Regardless, applying this approach to our current focus will be difficult. Unless the running down of animals on roadways qualifies as a socially proscribed expression of animal cruelty, programming AVs to do so will not clearly connect to Kant’s view of our indirect duties to animals. Kant’s position may remain relevant in other ways, however. There is increasing evidence that work in abattoirs can have a significant adverse effect on the mental health of abattoir labourers.Footnote ⁴³ For those labourers, then, it would be better to avoid work that involves the slaughter and butchering of animals. This may be enough to connect back to Kant’s view,Footnote ⁴⁴ as such work can negatively impact the humanity we are all under obligation to cultivate if mental health issues experienced by some abattoir workers negatively impact their ability in the future to act autonomously or morally. Perhaps, then, the intentional killing of animals on roadways can enjoy proscription should such killing adversely affect those in the vehicles.

An obligation to avoid small-mindedness, mentioned by Kant when talking about our obligations to animals, does not seem to be reducible to cruelty and so can apply even where cruelty is not an issue. He states, “[a] master who turns out his ass or his dog because the animal can no longer earn its keep manifests a small mind.”Footnote ⁴⁵ Earlier in the same lecture he states, “if a dog has served his master long and faithfully, his service, on the analogy of human service, deserves reward, and when the dog has grown too old to serve, his master ought to keep him until he dies.”Footnote ⁴⁶

For Kant, small-mindedness harms the humanity of the actor, encouraging them not to universalize certain caring or respectful responses that should be universalized. The kind of caring or respectful response motivated by long and faithful service surely affects the steps taken to protect the relevant animals from unnecessary death. For example, Kant does not think that it is appropriate to simply “euthanize” the dog after their service is over. As long as those animals live among us, we have reasons to program AVs so as to avoid harming them, at least at first glance. This can include more than faithful dogs. We can talk of faithful horses, cats, birds, and even rats, so long as the companionship these animals provided their guardians can be reasonably construed as a service, as we think it can. More formal service animals can include more than dogs, and, when they are there for mental health support, relevantly resemble companion animals in socially bonded, interspecific family units. Besides, service animals provide a clear service and should be rewarded for it.

Perhaps we can continue to broaden the sense of service we are drawing from here to include the maintenance of our environment (including the health of our biosphere), the facilitation of our food production, and even the health benefits we reap from periodically exploring national parks, protected areas, or wilderness. If we can, the indirect duty view that we have been outlining can be used to motivate programming our AVs with a sensitivity to the presence of many more animals than might otherwise be the case.

There is another avenue which perhaps we could take, using Kant as our guide. In his Grounding for the Metaphysics of Morals, Kant provides one argument against the morality of suicide that foregrounds the contradictory nature of a universal law of nature that leads those obeying it to destroy life.Footnote ⁴⁷ Though this may be pressing Kant’s point too far, the growing concern to stay the decline of global free-living animal populations, to conserve wilderness, animal habitats, and rainforests, and maintain a to-be-specified level of biodiversity can all be understood in terms of human self-interest. Indeed, this is often how these considerations are put in public media.Footnote ⁴⁸ To universalize maxims that threaten the stability or survival of our biosphere, and so ourselves, seems to violate the kind of ethical analysis that Kant would bring to bear. If something like this move is right, it could once again motivate programming our AVs with a broad sensitivity to what they are free or not free to run over.

Of course, there are limits to this line of reasoning. Considerations of whom-to-harm scenarios will not, on indirect duty views, favour other animals over humans. But all is not lost. The considerations we have canvassed would warrant making significant efforts to rethink how we move about in our world. Arguably, trivial reasons for travel would be critically scrutinized, we should redesign various spaces through which we travel to minimize lethal or harmful encounters with other animals, and perhaps prohibit the use of vehicular transport in ecologically sensitive areas.

However, this is not likely to satisfy those who would defend according greater moral significance to the animals who might be seriously injured or killed by AVs. It is to direct duty views of our obligations to animals that we now turn.

As we mentioned earlier, direct duty views can come in two versions. Under one version animals can be directly wronged by our actions, but the animals do not enjoy equal moral status with ourselves (broadly construed to include all humans who enjoy high moral status). The other version also holds that animals can be directly wronged by our actions but, in contrast to the first version, holds that a to-be-specified number of these animals enjoy equal moral status with ourselves.Footnote ⁴⁹

Direct duty views where other animals are not moral equals to ourselves (direct duty unequal views) can reflect what some have called a sliding scale of moral status. A sliding-scale approach holds that

It is possible that direct duty unequal views do not hold a sliding scale of moral status (e.g., perhaps all sentient animals who are not persons, or who cannot form and prioritize higher-order desires, goals or preferences, are both all equal to each other but not equal to persons or those who can form and prioritize high-order desires, goals, or preferences),Footnote ⁵¹ but we will not discuss them here for several reasons. First, where moral views hold that we have direct duties to animals but resist proscribing the use of animals in the food, research, or testing industries, they tend to reflect a sliding-scale approach.Footnote ⁵² We think it is safe to claim that such views outnumber either indirect duty views, direct duty unequal views which hold that all other animals enjoy a moral status equal to each other, or direct duty views where other animals are our moral equals. Second, policies governing the use of animals in, say, the research or testing industry tend to offer special protections to some animals based on their greater social importance (e.g., cats, dogs, horses, and nonhuman primates).Footnote ⁵³ These can be readily understood as reflecting direct duty unequal views committed to a sliding scale of moral status, though such a scale is informed by the social significance of animals, not just their morally significant psychological capacities. A third reason arises out of a very common view in applied ethics that interests reflect the capacities of animals who possess them, so that we can reasonably talk of differing interests across taxa relative to differing capacities among members of these taxa.Footnote ⁵⁴ It is possible that animals with interests in, say, liberty of movement, liberty of social choice, and control over their immediate environments will enjoy protections that for various reasons incline those using animals to use those who either lack those interests or lack them to a lesser or greater degree. This effectively supports a sliding scale where the use of animals lacking or possessing “burdensome” interests to a lesser or greater degree is consistently favoured and normalized as animals first-to-be-used.Footnote ⁵⁵

We can expect that direct duty unequal views will favour AV programming that is sensitive to the presence of socially significant animals, be they companion animals, primates, or working animals. Presuming that the visual acuity of these vehicles is less than ours, and/or programmers are (required to be) responsive to the variety of animals who can qualify as companion animals, such programming sensitivity should favour a number of animals living among us.Footnote ⁵⁶

The environmental and ecological considerations, mentioned earlier in relation to indirect duty views, can re-enter our discussion here, as there will be many environmental advocates who take direct duty unequal views of our obligations to other animals quite seriously.Footnote ⁵⁷ Arguably, direct duty unequal views can provide support for stronger constraints on travellers who would risk seriously injuring or killing animals of great ecological or conservation significance. After all, on top of considerations of human self-interest there would be the interests of the relevant animals themselves, however they are weighted or ranked. Much the same could be said for religious constraints on harming other animals, where those constraints are understood to reflect other-regarding, species-inclusive considerations.

As with indirect duty views, however, direct duty unequal views are unlikely to favour animals over humans in whom-to-harm scenarios. They would offer greater support for even more changes in the way we travel than indirect duty views, but they are still likely to resemble attitudes of some of the judges mentioned in our earlier discussion of law.

For those yet to be satisfied by this discussion of the promise of ethics to substantively inform AV programming in animal friendly ways, there is still hope. Though direct duty unequal views are, we think, very popular, they come under increasing strain when they hold that no other animals can enjoy a moral status equal to ourselves (that is, those humans who enjoy high moral status). As one of us has argued elsewhere, the more we learn about the psychological capacities of other animals on this planet, the more a number of animals seem to enjoy a relevantly similar range of interests as ourselves. Chimpanzees quickly come to mind, but so do domestic dogs, dolphins, elephants, and crows.Footnote ⁵⁸

Some think that our species identity gives us special moral importance, but this has not survived critical philosophical analysis. Consider, personhood is often taken to be a marker of high moral status of the kind we commonly accord ourselves. But there is no definitive account of personhood in the philosophical literature,Footnote ⁵⁹ or to the best of our knowledge anywhere else, and no plausible account treats being a member of our species as a necessary condition.Footnote ⁶⁰ There are some simple reasons for this. First, such a restrictive account would not reflect its common historical use. Many religious stories and teachings are populated by nonhuman persons.Footnote ⁶¹ Even if some, many, or all of these stories and teachings stretch credulity, it is not on the grounds that only humans can be persons. Second, should there be sapient life on other planets whom we can detect or encounter, some will likely be persons in the sense we commonly use to describe ourselves.Footnote ⁶² At least, this is not an unreasonable view, but it could not be otherwise if only humans are persons. Third, there was a time in our species’ past when we co-existed with other members of our genus. There is even suggestive evidence that some early members of our species interbred with Homo neanderthalensis and Homo altai.Footnote ⁶³ Despite their off-spring not being fully members of our species, it does stretch credulity to hold that they were lesser (or ‘half’-) persons because of it. Indeed, it stretches credulity to think that their mates were lesser or non-persons too.

Once being a member of our species is ruled out as a necessary condition, we face the very real prospect that some other animals are persons. We are not going to offer an account of personhood that clearly includes other animals, as we do not have to for our discussion to trouble direct duty unequal views. The more inclusive your views of personhood are, the more animals are likely to qualify (e.g., chimpanzees, dolphins, domestic dogs, macaques). The less inclusive your views, the less likely that all humans, between birth and death, are persons. Even in such a scenario, formal justice considerations—that like should be regarded alike—will favour animals whose psychological and behavioural profiles evince a range of interests relevantly similar to those seen among protected humans, whether they are persons or not.Footnote ⁶⁴

An ethical framework that would continue to resist allowing that some animals are our moral equals, despite what we now know of their psychological or behavioural capacities and the kinds of critical considerations that can be brought to bear on views of human supremacy (see our quick discussion of personhood, above), looks so self-interested as to undermine its credibility. Important for our discussion here, none of us needs to admit that many animals enjoy moral equality with ourselves for that admission to require some radical constraints on how we program AVs. We can be as conservative as we like, as long as we are consistent. Any animals who are our moral equals (or the moral equals of humans between birth and death, whether they are persons or not) should enjoy the same kind of consideration as ourselves when thinking about whom-to-harm scenarios. They are no more properly sacrificed than a child or grandfather on a pedestrian crossing. To the best of our knowledge, no efforts have been made to develop programming with this constraint in mind. If we are right here, there is much work that needs to be done.