The Edge of Sentience: Why Drawing Lines Is So Difficult

Sentience means the ability to feel. Scholars from many different disciplines are very interested in who/what is sentient and who/what isn't—what I call the biodiversity of sentience. I like to say the sentience is everywhere—it isn't science fiction—and that we must be very careful about drawing lines about where to find it, including plants. Ample research supports my and others' cautious approach and why scaling sentience is problematic and why sentience matters.¹ For these and other reasons I was thrilled to learn of Professor Jonathan Birch's new bookThe Edge of Sentience: Risk and Precaution in Humans, Other Animals, and AI that is free via open access. Here’s what he had to say about his eye-opening and deeply thoughtful evidence-based work.

Marc Bekoff: Why did you write The Edge of Sentience?

Jonathan Birch: When the right policy seems to depend on whether some animal or system is sentient, and we aren’t sure either way, what should we do? Governments around the world are grappling with a set of challenges of this type—problems at the “edge of sentience”. How do we decide whether to include invertebrates like octopuses, crabs and insects in animal welfare laws? How should we regulate “brain organoid” research? Could AI become sentient and what should we do about that?

I wanted to write a book that would put these policy challenges at the centre. A lot of books about sentience or consciousness mention these challenges as motivation—but then set them aside and get to the real business, the business of offering a speculative theory of consciousness. In my book, deciding what to do in our current state of uncertainty is the real business. We disagree about the nature of sentience, and we’ll continue to disagree, but we need ways of reaching agreement on sensible precautions.

MB: How does your book relate to your background and general areas of interest?

JB: My first book, The Philosophy of Social Evolution, was about the question: when can Darwinian evolution, often seen as a ruthless struggle for existence, favour forms of self-sacrifice—forms of altruism? I was weighing into controversies about the evolution of altruism in social insects, like bees, ants and termites. Through studying the behaviour of these animals I became extremely interested in their minds—in the ways they represent the world around them. So I started to ask: is there anything it feels like to be a bee, an ant, a termite?

These questions are incredibly difficult—we have no secure theory of subjective experience, and no secure experimental approaches, because it’s a very young area of science—but I do believe we can bring scientific evidence to bear on them. For about 10 years now I have been trying to solidify the foundations of the emerging science of animal sentience.

MB: Who do you hope to reach in your interesting and important book?

JB: It is a book of proposals—26 proposals, in fact, for how we could manage our uncertainty better than we do at present. Steps we can take to err on the side of caution, to mitigate risk. I want to reach policy-makers with these proposals, I want to reach doctors, vets, biomedical researchers, tech companies, and I want to reach the general public too—I want to start a big conversation about taking risks seriously that we have been neglecting up to now.

MB: What are some of the major topics you consider?

I’m probably best-known for my work on invertebrate sentience, which is one part of the book. My team’s work has led to some invertebrates—octopuses, squid, cuttlefish, crabs, lobsters, shrimps—being included in the UK’s Animal Welfare (Sentience) Act 2022. And I was pleased to see our work cited in a new law banning octopus farming in California. But there is far more to do. I don’t want the focus on octopuses to lead to us forgetting about insects, which are also “sentience candidates” (in the way I define that term). The emerging insect farming industry needs to take welfare more seriously.

These cases led me to see that there is a whole family of cases at the “edge of sentience”, and the book is about all of them. Similar issues arise with research on brain organoids. These are tiny models of human brain regions, made by inducing human stem cells to form neural tissue. In one study, these systems developed precursors to eyes (optic vesicles); in another study, researchers mounted one of them on an electrode array and trained it to play the videogame Pong, showing measurable improvement in performance during a 20-minute session. Intuitively, there must be some ethical limit here. The more like a human brain these systems get, the stronger the possibility they might achieve some form of sentience.

Then there is the possibility of sentience in AI. That debate has become far more mainstream in the last two years. I’ve even been working with people from tech companies on this topic—whereas previously the industry always tended to dismiss the idea of artificial sentience as a sci-fi idea.

In trying to monitor for signs of emerging sentience in AI, we face some serious problems. Octopuses may be separated from us by over 500 million years of evolution, but we at least have a common ancestor with them, and they are evolved, biological beings like us. Sentience in octopuses is supported by a nervous system—one very differently organised from ours, but still a nervous system. Testing for sentience in a non-biological being with no nervous system is whole different challenge.

To make matters worse, some AI systems (including large language models, like ChatGPT, Claude and Gemini) have enormous amounts of information about what humans find persuasive—and they can leverage this to fake superficial signs of sentience convincingly if it serves their objectives. They can game our criteria.

We don’t face this “gaming problem” with octopuses—if they tick all the boxes for feeling pain, it’s most likely because they feel pain, not because they stand to gain from fooling us into thinking they feel pain.

MB: Are you hopeful that as people learn more about sentience they will be more open to factoring it into how they treat humans and other animals and perhaps other entities?

JB: Of course, yes. A lot of my proposals are very moderate, and deliberately so—they recommend small, low-cost changes to our current ways of life to reduce risks of causing serious harm. A simple example: people should stop dropping crabs and lobsters into pans of boiling water. Yet, moderate as they are, if my proposals were all adopted, the benefits would be significant. We can change our way of life by small steps.