Non-human animals feature prominently in all areas of neuroscience research, ranging from drosophila to higher order primates. It is a problematic belief among many researchers that nonhuman animals are outside the scope of moral consideration, but that we ought to “reduce, refine, and replace” their use where possible.1, 2
Drawing on Peter Singer’s interpretation of utilitarianism,3 I first seek to demonstrate that it is capacity for suffering that qualifies beings for moral consideration in a utilitarian framework, as opposed to intelligence, language, or any other arbitrary trait. I will then argue that various species’ capacities to suffer can be reasonably estimated and expressed as a relative numeric “c-value.” Based on these values, I propose a utilitarian model for animal experimentation which incorporates seven key considerations: capacity for suffering (c), degree of suffering inflicted (s), number of animals used (n), probability of positive experimental results (Pe), probability that practical medical treatments will arise from positive findings (Pb), value of individual human benefit from treatment (B), and number of humans benefitted (N). If the product of csn>PePbBN, I argue that the experiment in question is morally wrong, whereas the inverse signifies an experiment which is morally right (and ought to be done).
Within classical utilitarianism, the action resulting in the highest net utility (one’s happiness, or the ability to further one’s needs or desires) is the morally right action to take, while choices resulting in comparatively less utility are morally wrong. 4 Due to imperfect knowledge, the utility of a given outcome is often multiplied by the probability of that outcome occurring (as evaluated to the best of one’s current knowledge).5 Utilitarianism is an appealing normative model for its simplicity and its fundamentally egalitarian premise. No one individual’s well-being is weighted more than another, as every individual in the moral equation counts as one.4, 6 But some would argue that the decision to only include human beings in moral considerations is an arbitrary one.3, 6 Peter Singer and Jeremy Bentham elegantly propose that one’s capacity to suffer is the essential quality that ought to qualify a being for utilitarian consideration, not simply one’s arbitrary status as a human being.3,4 If adult monkeys are undoubtedly smarter, more conscious and more expressive than day-old human babies, how could we possibly prioritize all humans over all animals on the arbitrary basis of intelligence or language?3
Nevertheless, non-human animals are incredibly varied in complexity – it seems intuitive that torturing a monkey and torturing a fruit fly should not have equal moral weighting. It seems highly plausible that capacity to appreciate suffering, like all cognitive processes, exists on an evolutionary spectrum, and ought to be weighted in a utilitarian equation accordingly. I put forward two simple, candidate metrics that might serve as reasonable metrics for a species’ capacity to suffer (in order to derive a theoretical “c-value”): a) cortical thickness and b) complexity of social behaviour.
It has long been proposed that cortical thickness is a general indicator of general intelligence — one’s ability to detect patterns and solve novel problems.11 While it is problematic to equate intelligence with capacity to suffer, it seems plausible that more intelligent animals have a better understanding of and memory for instances of pain, thus possibly adding dimensions to suffering that go beyond immediate stimulus-response pairings. For instance, complex pattern analysis might be a prerequisite for feeling anxious of future pain, and a sophisticated memory might be necessary to experience post-traumatic stress. This metric provides intuitively sound rankings, with C. elegans = Aplysia = Drosophila < mice < rats < squirrels < dogs < cats < rhesus monkeys < horses < gorillas < chimpanzees.10, 11
Secondly, degree of social complexity may be able to provide a similarly plausible set of rankings for a species’ capacity to suffer. It seems highly probable that empathy (including the ability to understand and abstract the suffering of other conspecifics) developed out of pure evolutionary necessity in social animals.21 Such adaptations increase the chances of survival of individuals via reciprocal altruism, but likely also increase the capacity for understanding and abstracting one’s own suffering, in addition to the suffering of others. Most regulations protecting higher-order animals are mere guidelines, open to a wide range of subjective interpretations by any given review board.8, 9
To illustrate how a more rigorous utilitarian approach can be applied on review boards based on theoretical c-values, let us examine the 2003 experiment by Carmena et al., in which two rhesus macaques were taught to control a closed brain-machine interface (BMIc). Experimenters read electrical activity in the frontoparietal cortex via surgically implanted electrodes. Repeated training with the BMIc allowed the monkeys to use visual feedback to reach and grasp with the robotic arm, without moving their own arm. These findings may contribute to technologies that would allow paralyzed patients to bypass their spinal cord injury to elicit voluntary, machine-mediated movements directly from the brain.12
Capacity for suffering (c). Rhesus monkeys have 480 million cortical neurons, compared to 11 500 million in man and 160 million in dogs11. The social organization of this species is complex and well-documented. For instance, they are capable of producing at least five distinct types of scream vocalizations during agonistic encounters to elicit support from conspecifics, each denoting particular kinds of threats and levels of aggression.16 Most notably, rhesus monkeys which were excluded from their social group quickly died in the absence of support, protection and resources from their conspecifics.17, 18 Fear and suffering caused by social exclusion is therefore likely an important feature for survival in this species (as it has been in humans), and it seems plausible that they share with humans the deep pain associated with rejection and isolation. On this basis, let us assume a c-value of 0.75, which is to say the average rhesus macaque might have about 75% of the typical human’s capacity for suffering.
Degree of suffering inflicted (s). In this particular experiment, the painful surgical implantation of brain electrodes coupled with the stress of social isolation, captivity, and forced physical restraint12 might reasonably warrant an s-value of 3 out of a possible 4.1
Number of animals used (n). X utiles lost by 5 animals is a five-fold worse outcome than X utiles lost by only 1 animal. In this case, two adult female macaques were used.12 Thus the total score for animal outcomes might be c*s*n = (0.75)(3)(2) = 4.5.
Probability of positive experimental findings (Pe). Experiments which seek to fine-tune already well-researched treatments will be far more likely to succeed in their goal than those which seek to pioneer new treatments from scratch. The Pe term ought to reflect the probability that the experiment proposed is a reasonable one and will not simply result in negligibly beneficial negative findings, evaluated by the impact and number of related previous experiments in the field. The hypothesis that “drug X is not harmful for primate consumption” might have quite a high Pe if there is extensive pre-existing evidence that drug X is not harmful to mice or reptiles. In contrast, BMIcs were relatively uncharted territory at the time of this experiment (with only about three preceding studies of this kind having ever been done on primates),13-15 yielding a low Pe, reflecting a relatively high chance of failure. Let us retrospectively assume a Pe of 0.1, multiplied by 0.95 to adjust for type I statistical error. 2
Probability of benefit resulting from findings (Pb). It can be argued that there is intrinsic value in scientific knowledge itself (knowing for the sake of knowing), but utilitarianism would hold that this knowledge is only valuable insofar as it can be used to derive practical treatments or 1 McGill’s 2013 Animal Use Report (see Appendix A) categorizes animal experiments by degree of suffering, ranging from category B (“experiments causing little or no discomfort or stress”) to category E (“procedures involving inflicting severe pain, near, at, or above the pain threshold of unanaesthetized, conscious animals”), providing a useful standard for numerical categorization. 2 We typically allow for a 0.05 (5%) margin of type-I statistical error; thus, at best, even if the hypothesis was corroborated by experimental findings, there is typically only up to a 0.95 probability that the findings indeed reflected true differences in the population at large benefits for actors at some point down the road.3 For instance, experiments on animals for the purposes of educating undergraduates may have ramifications on student knowledge of biology, and some of these students may go on to benefit society as doctors or researchers. However, such links are tenuous and ought to reflect a relatively low Pb — there are a number of interfering factors between a given educational experiment and a student giving back to her community. Conversely, experiments testing the safety and efficacy of pre-clinical stage pharmaceutical drugs may have a high relative Pb, as testing for adverse effects in animals can very immediately and directly save human lives. Experiments which seek to contribute to basic understanding of fundamental biological structures and functions without a direct practical benefit in mind ought to have a Pb value somewhere in between. In the case of the macaque experiment, there remain a number of scientific and financial barriers impeding the development of mechanical limbs for day-to-day use among paraplegics, and far more research will undoubtedly need to be conducted before this benefit can ever be practically realized.12 Let us therefore assume that this particular study advanced the possibility of brain-machine interfaces for paraplegics by only 5%, or 0.05.
Treatment or benefit (B). To match the 4-point scale for animal suffering, one might propose an analogous 4-point scale for the value of the human benefit. In the field of public health, QALYs (quality-adjusted life years) attempt to assign numerical utility to particular ailments in a similar manner.19, 20 Let us assume the B-value of a mechanical limb for the average paraplegic to be 3 out of a possible 4.
Number of humans benefitted (N). Of the 85000 patients with paraplegia in Canada,22 let us assume 1000 will be eligible and able to afford MBIs due to the high cost of the treatment. The probability-adjusted human utility for performing this experiment might therefore amount to Pe*Pb*B*N = [(0.95)(0.1)](0.05)(3)(1000) = 14.25. According to these estimations, it was 3 McGill’s annual Animal Use Report delineates 5 categories of purposes of animal use (PAUs) (see Appendix A) which may be useful in evaluating Pb. morally correct to approve this experiment. Nevertheless, upon a more careful analysis of factors, one can easily imagine an increase or decrease in any one of these values, which might tip the scale in one direction or another.
Overall, it is highly reasonable to assume that capacity to suffer qualifies beings for ethical consideration, and that animals are capable of suffering. It follows that inflicting pain on animals through experimentation comes at a moral cost. It seems intuitive that this moral cost ought to be weighted according to a species’ relative capacity to suffer, as fruit flies likely cannot experience the complex emotional suffering that monkeys can. “Capacity to suffer” times “suffering inflicted” times “number of animals” used is therefore a reasonable estimation for the negative utility produced in a given experiment. Moreover, ethics boards intuitively justify the suffering of animals by citing the possible societal benefits that might arise from scientific findings. “The probability of a given hypothesis being true” times “the probability of given findings resulting in a treatment” times “the utility of the treatment” times “the number of humans that will benefit from the treatment” is therefore a reasonable expression of net utility gained. While these values may never be objectively resolvable, such a model at the very least delineates the kinds of factors one ought to weigh when attempting to justify the moral status of an animal experiment. A standardized cost-benefit framework such as the present one ought to be employed as a decision aid to supplement intuitive decision-making on ethics review boards