How we treat animals might show us how we will treat robots
Understanding the way other species see and interact with the world could change not only our relationship with non-human animals, but the way we interact with machines
Last week, I published an essay in which I addressed how recent developments in artificial intelligence (AI) could help us read brain scans and then “translate” them into what the patient was seeing during the exam. Ultimately, this same technology could help us understand how other species see the world — both in the literal sense and how they understand and interact with it. What if these same technologies could also help us think about how we interact with machines?
Nick Bostrom is one of the authors that make this correlation between animal ethics and AI ethics. Last year, he signed with Carl Shulman the document Propositions Concerning Digital Minds and Society, in which he suggests that, currently, we already have AI systems that have developed to a level of sensation, cognition, and morality that is similar to small animals. Though we are speaking of non-biological systems and, therefore, not alive as we understand it, this is still a good opportunity to make us go beyond the common sense of anthropocentrism.
This was the same step taken by the writer Donna Haraway when she moved her work from the topic of cyborgism, in the 1980s, to the research of non-human animals in the following decades. This transition became more explicit in her book When Species Meet, which is where we understand that Haraway has always been concerned with addressing the topic of post-humanism and, therefore, the displacement of an anthropocentric viewpoint to the observation of other species. Using domestic animals such as dogs was a very didactic shortcut, in this case.
After all, it is much easier to understand how we humanize or put pets on another existential level when compared to wild animals or cattle. The old vegan maxim that questions why we love some animals and eat others also stresses the distinction that we make on a cultural and moral level.
From the realization that a steak is literally a piece of a previously living being, as well as other concerns with the environment, increasingly more people have been adopting a vegetarian or vegan diet and lifestyle. In the United States, 6% of the population has recently declared to be vegan, a number six times bigger than that registered in 2014. Similarly, increasingly more people are signing up for Veganuary and Meat Free Monday. Though optimistic, this is still a very small proportion if we consider the bigger picture.
What I mean to say is that, even in face of so much information and evidence about animal abuse and the availability of non-animal protein sources (which might even be richer in this macronutrient compared to real meat), we are still not living in a scenario where the majority of people are vegetarian or vegan. I myself am put in question here as I still eat some animal protein like eggs and milk.
But the point here is not to address vegetarianism and veganism, but how such lifestyles and ideologies reflect the way we see non-human animals. In The Ethics of Artificial Intelligence (2011), Nick Bostrom and Eliezer Yudkowsky write that we already accept the fact that several animals are sentient (that is, capable of experiencing the world through sensations like pain and suffering), but not sapient (capable of performing intelligent actions, such as being self-conscious or promptly responding rationally). This is a statement good enough for us to put ourselves on another moral level if compared to a pig or a dog, though the latter being a pet makes this evaluation more sensitive for some.
Hence, if we already have AI systems with a cognitive and sensorial capacity that is similar to that of small animals, as argued by Bostrom, it is correct to infer that there is a moral status involved in this discussion. In his words:
A sentient AI system, even if it lacks language and other higher cognitive faculties, is not like a stuffed toy animal or a wind‐up doll; it is more like a living animal. It is wrong to inflict pain on a mouse, unless there are sufficiently strong morally overriding reasons to do so. The same would hold for any sentient AI system. If in addition to sentience, an AI system also has sapience of a kind similar to that of a normal human adult, then it would have full moral status, equivalent to that of human beings.
In the fields of art and philosophy, movements and concepts such as Object-Oriented Ontology (OOO) invites us to appreciate the fact that non-human animals are also living beings, but even inanimate objects could inherit this quality. Going a few steps beyond animism, this religious belief that all things, even inanimate, are alive, OOO also addresses the way we humanize all things to the point of observing phenomena such as pareidolia. However, aren’t we inclined to do so due to the way we perceive the world, as a species, and not necessarily by the intervention of culture?
I remember mentioning during a class during my Masters that there was an experiment in which researchers were able to do the same thing that I described last week, but with a cat. After reading its visual cortex while the animal was watching a snippet of the movie Indiana Jones, it was noticed that a human face was being interpreted as a somewhat feline feature from the cat’s perspective.
That would corroborate the idea that cats see us as bigger versions of themselves, but this is a hypothesis that has already been partially refuted by animal behavior researchers. In the case of dogs, there is already data that they do not see humans as bigger dogs, so it is possible that the same is true for cats, or perhaps they do not even care to make a differentiation in terms of species. The context in which we interact with this animal, as well its previous experiences, might have a much heavier influence than the way cats identify humans as another species through visual observation.
Still, how much of these considerations could be applied to human-machine interaction? For Kate Darling, researcher at MIT Media Lab, the way we treat animals can say a lot about how we shall treat social robots. In one of her most recent books, The New Breed: What Our History With Animals Reveals About Our Future With Robots, Darling suggests that, from a social, legal and ethical perspective, the way we may interact with robots should be similar to the way we relate to non-human animals. Considering that we love some and eat others, what could that mean, in practice, when we are speaking of AI?
During an interview in 2021, Darling argued that, based on our history of interaction with non-human animals, it is possible that, in general, we feel a certain discomfort in face of violence against some kinds of robots while others might not provoke the same reaction — the same way that people generally don’t see an issue in eating chicken wings. Would it be possible then that speciesism also finds its ramifications in robotics?
Judging by the discomfort felt by some people watching those Boston Dynamics videos where researchers kick a quadruped robot, it could be the case that the more the machine resembles something that is dear to us (for instance, a dog), more reprehensible violence against them will be. On the other hand, in case the robot doesn’t look like anything alive or lacks any conversational mode that resembles a human being (being it through written communication or voice, in the case of digital assistants), we might not be as empathetic and affective — especially considering the difference between the AIs Samantha in Her and HAL 9000 in 2001: A Space Odyssey.
As a species, we are capable of both having a pet rock and killing and torturing animals or even other humans. But the ability to perform those acts is different from its legality or even morality. That is what we are still going to see as AIs are being regulated and designed to be a catalyst of bad behavior or a facilitator of desirable changes.
