I'm Tim Ventura, and we're joined today by Dr. Jeff Sebo, associate professor of environmental studies at New York University and one of the leading voices at the intersection of animal welfare, moral philosophy, and the emerging question of what we might owe to digital minds, AIs, and other non-standard intelligences that may not look, act, or communicate like us. Dr. Sebo earned his BA in philosophy and sociology from Texas Christian University, graduating summa cum laude, and earned his PhD in philosophy at NYU. Now at NYU, his work and teaching sits right at the overlap of animal minds, ethics, and policy; AI minds, ethics, and policy; and broader global health and climate ethics. In that practice, it shows up in courses and seminars that connect philosophy to real-world impact, topics like animals and climate change, animals and public health, and effective advocacy. He has written multiple books on animal ethics and our responsibilities to non-human life, including Saving Animals, Saving Ourselves, and his book, The Moral Circle: Who Matters, What Matters, and Why, which argues for a precautionary approach to moral concern. So, Jeff, welcome. It is a true pleasure and honor to have you with me today, sir.
Well, thank you so much for having me, and thanks for that wonderful intro. I should also briefly introduce my fellow interviewee, Smokey. He will be here for the interview as well, my dog. So, both of us are excited to hear your questions.
Oh, that's awesome. I have an elderly pup myself. She is in the house snoozing while I do this. So, yeah, again, really excited. This is a topic that is really close to my heart. So, we're exploring themes from your work in animal and AI ethics in the moral circle. I want to jump right into this. You have argued that when we're genuinely unsure of who can suffer, the ethical default should be precautionary inclusion. Right.
That is correct. There are a lot of details to be sorted out. Exactly. How much weight should we give a particular entity? How much priority should we give a particular entity when we feel unsure if they can even suffer at all? If they can even pursue their own goals at all. But I do think that we should give at least limited minimal consideration to beings who have a realistic possibility of being sentient or agentic or otherwise morally significant when making decisions that affect them, in the spirit of caution and humility. I think that is the least that we can do given our track record of really bad under-inclusion, really, really bad treatment of non-human entities that turn out in the fullness of time to pretty clearly be sentient and agentic and otherwise morally significant.
Well, that again, that is why this is an incredibly powerful issue, and I think that there's kind of a sea change happening in, you know, Western culture, but in global culture as well. And, you know, AI, I think, is almost like the tip of the spear on this, right? Where now we're looking at these non-human, non-biological intelligences emerging, and really my hope is that this forces us to look back at animal intelligence and say, hey, guess what? They are also sentient. They're just not quite the same as we are.
Yeah. It would be really interesting if for some people AI is what persuaded them that we should take the possibility of non-human sentience seriously, and then they look back at the other animals who were here all along and have a little bit more of a generous attitude about them. That would be a funny order in which to make progress, but honestly, I will take progress in whatever way we can get it.
Well, so you have said implicitly and explicitly that how we treat AIs is a rehearsal for how we treat non-standard minds in general. So what are some of the warning signs that we may be repeating some of the same old exclusion patterns as the past?
Yeah, good question. So we are starting to see some interesting patterns emerge with chatbots, large language models, other generative models, other AI systems that are familiar from our interactions with non-human animals. So for example, we are seeing a lot of people attribute consciousness, sentience, agency, personhood, moral significance to chatbots and digital companions, digital assistants, but then not attribute those features to other kinds of AI systems that lack that same familiar human interface. And this is very similar to how we have engaged with non-human animals in the past. Studies show that we are likely to, if anything, over-attribute consciousness, sentience, agency, personhood, moral significance to animals when they look like us, act like us, when we classify them as companions. So we anthropomorphize our cats and dogs. If anything, we see them as having more human-like interests than they actually do. And we see them as members of our family, protect them with our lives. But then we under-attribute these features to animals who have the other characteristics. When they look different, they act different, we classify them as commodities, and we use them for food or research. Then if anything, we anthropo-denial. We see them as having human-like features less than they actually do. And so that can lead us to predict that those same tensions and those same dynamics are going to emerge with AI, that when we do see ourselves in them, we will if anything over-attribute, err too much on the side of seeing them as having human-like traits. But then when they simply perform algorithms in the background or when they do radically un-human-like things like just detect patterns and so on, then we might if anything under-attribute and deny human characteristics that they actually have.
Well, so this might be going too far down the rabbit hole. This is a little bit off my questions list, but so we basically, what you're saying is like in the case of a dog, right, where they evolved to evoke certain emotions, you know, in us, right? They look childlike in many ways and so they trigger some of those responses in human physiology that create a bond. You know, in something like them, we over-anthropomorphize. We might do the same thing to like ChatGPT because it talks like a person. It uses pronouns like I, you know, things like that, right? It speaks emotively. So, we over-anthropomorphize things that are like ourselves. But then we see something like, you know, a mouse or an ant or, you know, a process server that uses AI. Those don't communicate or invoke emotions in us and we tend to exclude them and they become treated poorly. So for me it seems like there's kind of a cline in how these are treated, right? It's almost a matter of degree. Do you have any thoughts on what the cutoffs or boundaries should be there?
For the cutoffs or boundaries for when we are at risk of over-attributing or under-attributing these characteristics, or when we should or should not give them the benefit of the doubt? What kind of cutoff do you have in mind?
Well, so animals are certainly a challenge. The example that came to mind actually in this case would be, okay, so in something like ChatGPT or maybe Google Gemini where it communicates like us and people are out there saying this thing seems like it's sentient, it needs to be treated better, right? Where is the cutoff between something like that and then something like a web server where it's just doing its job and you don't think anything is going on behind the scenes?
Yeah, that is a really tough question. And when I set out to write my book, The Moral Circle, that you mentioned in the intro, I had a very specific goal there. I wanted to promote moral consideration for what I thought were clear candidates like invertebrates, mollusks, crustaceans, insects, and then advanced near-future AI systems that would have many of the markers of sentience and agency and moral significance. But then when I started writing the book and going through the evidence and arguments and seeing how important and how difficult these issues are and how much disagreement and uncertainty we have about the core ethical issues and the core scientific issues, I found myself talking myself into extending at least a little bit of curiosity to a much wider range of entities. Maybe at least giving, if not moral concern, then at least a little bit of puzzlement and curiosity to plants and fungi and microscopic organisms on the organic side. And then yes, to these basic computer servers that we ordinarily experience as simple objects on the silicon side, on the computer side. I think the probability that those entities have even flickers of their own pleasures and pains and feelings and emotions is very small. But I cannot say that the probability is zero just because of how complicated these issues are. So I personally now, and I hate this, but I personally now have a really hard time drawing the line even with those types of entities.
But you came to this issue of AI moral status through animal welfare. And so I wanted to ask what animal advocacy taught you about the psychology of denial, right? The ways that humans rationalize ignoring minds that don't speak our language or invoke our emotional response or kind of reach out and touch us in that way.
Yeah, great question. Really important issue, and we already touched on some of it, which is when we are more at risk of excessive anthropomorphism versus excessive anthropodenial, right? They look like us, act like us, we treat them as companions versus they look different, act different, we treat them as commodities. And then you mentioned some of the specific features that might be involved there, right? When you have a large body, large head, large eyes, furry skin instead of scaly skin, symmetrical features instead of asymmetrical features, four limbs instead of six and eight limbs, those are all going to hijack our parenting gene or hijack our sense of empathy, sympathy, compassion. And then when they are companions instead of commodities, it makes it all the more so. And then when they are right here instead of on the other side of the world or trapped in a factory farm or slaughterhouse, out of sight, out of mind. Those are all relevant factors. And then one other important point to note here is my own complicity in these industries can affect how I perceive these individuals as well. So for example, there are some studies that show that for people who eat meat, they are less likely to attribute sentience and moral status to those particular animals. Whereas if people are vegetarian or vegan, then they are more likely to attribute sentience and moral status to those animals. Now, it can be tough to know in which direction the causal arrow goes there. Is it that you go vegan and vegetarian because you see the animals as sentient? Or is it that you see the animals as sentient because you are vegan and vegetarian and that cognitive dissonance is missing? But either way, we see that variation. And this is part of what gives the issue some urgency for AI because the more time passes, the more these models will become really advanced and then really integrated into our lives and economies and societies. And soon enough it will be a little bit like animal agriculture where people all over the world are really dependent on their use of these models for their livelihoods or for various cultural practices. And at that point there will be a lot of incentive for them to look the other way. And so this is a precious moment where we can ask these questions in a little bit more of an objective and impartial way before we become super dependent on them and we have this really strong incentive to see them as objects.
That's a really amazing point that you made about diet and the psychology of denial. So, in my case, I went plant-based probably about three or four years ago, and it was really for health reasons, right? I was exercising more. I was getting in better shape. I wanted to eat healthier. And so, you know, the meats and the fatty foods and things like that, I just kind of left that behind, but over time kind of moved more to a vegan position. And I would say just speaking personally, I have become more sensitive to issues like factory farming and things like that because I'm looking at this a little more squarely now, right? I don't have an incentive to kind of look the other way.
Yeah, I do think that that is possible. And people like you who start being plant-based or vegan, however you want to describe it, for non-animal welfare and rights reasons, but then you find yourself opening up to animal welfare and rights once you go down that path. I think that is some indication that the causal arrow can run in that direction sometimes. Now, none of us should kid ourselves. We always have bias and we always have mixed incentives. So it is not as though the meat-eaters are the biased ones and the vegans are wonderfully bias-free, right? We might simply be inheriting a different set of biases and now we want to look at evidence in a way that positively reinforces the choices that we made and can send nice signals to our community and so on. So we should always be on guard for new ways that we might be susceptible to bias. But with that said, I do think that there is something here that when you abstain from participation in a harmful industry, you do have a little bit more of an ability to see it for what it is and to be motivated to work to dismantle it and replace it with something better. And I think that this is an underappreciated reason for being vegan or plant-based, for example. People mostly think about the economic impacts of our diets and mostly think of that as the helpful consequence, but actually I think the social signal that it sends to others and the psychological signal that it sends to ourselves, the way that it can remove that cognitive dissonance and it can reinforce our values and our goals and our commitment to the issue, that can be really motivating and really helpful and another important consequence to keep in mind.
Well, I think that change is already happening in terms of dogs and cats. I had read that there was a social change in American culture, I believe in the 1980s, where they said dogs started to come indoors, right? Before that, it was the dog had the dog house out back, you know, and now, I mean, dogs, they sleep in the house. Sometimes they sleep in people's beds. They are much more a member of our family. But the fact that we're even discussing things like animal welfare, I mean, the other day I read in the news an article about squid hunting, right? And they were comparing it to whaling. And so, you know, seeing stories in the news headlines like that to me is an indicator that awareness is rising of animal intelligence. And even if it's not the same kind of sentience, even if they don't necessarily have the same identical capacity to suffer pain and anxiety and fear that we do, it seems like there is a greater awareness in human society, right?
Yes. I think that there are different trends going in different directions, which makes it a complicated situation. So on the one hand, I think you are right to say that over the entire global population, there is more awareness than there ever has been before about animal consciousness and sentience, about the extent to which animals can have their own feelings and emotions and be capable of happiness and suffering and desires and preferences. And there is more concern than there ever has been for animal welfare, maybe even animal rights, more people who do want there to be basic welfare protections for animals, including animals used for food and research and entertainment and other practices like that. And those are wonderful developments. And it coincides with a rise in the availability and affordability of plant-based foods and more people than ever have been eating plant-based foods. Global consumption of plant-based meats is going up, for example. But then there is also this trend of us harming and killing animals more than we ever have before. Factory farming and industrial animal agriculture are also operating at higher scales and are spreading to new species. Octopus farming is coming online. Shrimp farming is thriving. Insect farming is thriving. And so we kill more than a hundred billion vertebrates per year through farming, and then hundreds of billions if not trillions of invertebrates per year. So some trends are making things better and better and better and better. Other trends are making things worse, worse, worse, and worse. And this gives us a lot of reason for optimism and pessimism at the same time. A really complicated scenario.
I wanted to ask about, I touched on dogs coming indoors a moment ago. Companion animals already blur the boundaries of what we consider family, right? Dogs and cats, they're inside of our moral circle emotionally, legally, and culturally. What are some of the lessons that that could give us for AI systems, AI companions, right?
Yeah, good question. The domesticated animals and especially domesticated companion animals. They are in our lives and societies and as you say our families in a way that really even many humans are not always treated as being, to say nothing of other animals, to say nothing of digital minds, AI systems. So what lessons can we learn from that? One lesson again just goes back to the types of bias, the situations in which we are more primed, more likely to give another type of being the benefit of the doubt and see them as one of us. And that tells us something about our own dispositions and something we can work with or work around. But maybe a more positive lesson to take from it is that there is at least some precedent for us seeing a radically different kind of being as kin, for seeing ourselves as being in a socially and morally significant relationship with them. I really do feel as though I have a relationship with my dog and responsibilities to my dog. My partner, my human family, they understand that he comes first and they come next and they are at peace with that. He remains the only individual for whom I have gotten a tattoo. And I think that there are a lot of other people who are like that and they ought to be. And the positive lesson that we can take from that is that we clearly do have some capacity for seeing as kin radically different types of beings with radically different forms of life. And so we just have to ask, okay, what are the conditions under which we are motivated to do that and how can we recreate those conditions for these other beings who we think might deserve that same type of treatment. So Eric Schwitzgebel, a fellow philosopher, and I recently argued for what we call the emotional alignment design policy. And what we argue there is that AI systems should be designed in a way that elicits emotional reactions in users that matches their capacities and moral status. So when AI systems are not sentient and agentic and morally significant, they should not be designed so that they appear like they are. But when they are or at least are likely to be sentient, agentic, morally significant, then they really should be designed to appear as though they are, to tug at our heartstrings a little bit so that we naturally feel inclined to treat them like kin in the same way that I do with Smokey and that many of us do with our companion animals.
Well, there are actually a lot of reasons that you would want to have like a policy like that in place, right? And one of those is it is so easy to hijack, I guess, the human mind, right? And so that's something we've seen with doom scrolling and social media, right? Where it plays off just, you know, processing flaws and the way that our mind works and things like that to basically hijack us to sell advertising. And AI in theory could take that to the next level. So being able to align that with the way that it's built seems like it would have multiple purposes. Not just encouraging better treatment of agentic sentient AI, but also I guess disincentivizing the possibility of basically it hijacking our mind to sell us ads, you know.
