In The Edge of Sentience (2024, 2025), Birch develops a precautionary framework for crafting public policies about potentially sentient systems. Underlying this framework is a commitment to the democratic principle that the public should authorize all significant moral judgments that arise during the policymaking process. In this commentary, I suggest that Birch’s framework fails to live up to this commitment by requiring scientists without public accountability to make consequential moral judgments. I suggest a friendly amendment to Birch’s framework by replacing the idea of a scientific meta-consensus with the method of formal expert elicitation, a well-tested method of gathering expert opinions on complex topics in a way that makes their value judgments clear and accountable to democratic oversight.
What does The Edge of Sentience tell us about insect sentience and consciousness? In this commentary, we reflect on Jonathan Birch’s framework for defining and scientifically investigating sentience in animals. We discuss the methodological challenges of studying subjective experience in evolutionary distant species, and consider trace conditioning, affective states, and neural architectures as potential windows into consciousness in insects. We argue that uncertainty, whether concerning definitions or methodology, should not prevent experimental investigation; rather it should motivate new approaches and experimental designs for exploring the diversity of conscious experience.
The Edge of Sentience offers an ambitious and timely institutional framework for proportionate precaution under uncertainty about sentience. Its systematic treatment of empirical and theoretical uncertainty across diverse cases is a major contribution. Focusing on the problem of identifying and weighing different risks when designing proportionate precautions, this commentary suggests three areas in which Birch’s framework could be developed further: a more domain-sensitive justification of the priority given to avoiding under-attribution; a clearer distinction among competing forms of risk aversion and their divergent policy implications; and a fuller explanation of how the framework generates different recommendations across cases.
Oldfield & Bonano (2024) (O&B) document how far fish welfare research lags behind welfare research in other animals. We share O&B’s feelings-based concept of welfare and agree that aquaculture seeded this field; but research volume has not scaled with animal numbers: Welfare attention per farmed fish or decapod remains negligible. Along the lines of preceding commentaries (Maia 2024, Mather 2024), we treat farmed aquatic animals as individuals rather than tonnage. We then raise a point so far absent from the commentary on O&B so far: reducing the number of sentient animals entering intensive systems. Coastal seaweed farming, which cultivates non-sentient primary producer organisms, is the marine analogue of the plant-based shift on land – and it is credible economically.
Birch's framework manages uncertainty but overlooks the likelihood that citizen panels have speciesist biases that reduce precaution to mere harm-management even in species that undoubtedly feel.
Birch aims to remain neutral about whether life is necessary for sentience. We argue that life can be given a role while maintaining neutrality. Markers of sentience have been identified in living systems, where some contribute to the system's self-maintenance. When such a marker is reproduced in an artificial system that is not self-maintaining it is not clear that it is as reliable as in living systems.
Jonathan Birch’s The Edge of Sentience promises a precautionary framework that guides ethical action without taking sides in the disputed metaphysics of consciousness. I argue that it cannot do both. Birch admits “biopsychism” (that all living organisms are conscious) into his "zone of reasonable disagreement," yet his operational criteria exclude plants and unicellular organisms by requiring evidence that biopsychism holds to be irrelevant. This leaves a dilemma. Keeping the “broad” standard preserves neutrality but extends candidacy to all life, guiding nothing; restricting to scientific evidence guides action but is itself the theoretical commitment the framework was meant to avoid.
Birch's (2024) The Edge of Sentience identifies sentience candidates primarily through the pain-focused LSE Criteria. Yet, as Birch himself notes in Proposal 17, sentience is not pain. We argue that statutory welfare protections must therefore reach beyond pain, and that the right tool for doing so is Birch et al.’s (2020) dimensionality framework, which characterises conscious experience across five domains. Used together, the LSE Criteria establish that welfare is at stake, while the dimensionality framework maps what is at stake beyond pain to provide a taxon-specific profile of where captivity impoverishes experience. We apply this approach to invertebrate welfare and call for statutory, expert-informed, and regularly reviewed care standards for cephalopods, decapod crustaceans, and other sentience candidates.
Jonathan Birch's (2025) LSE criteria for sentience include motivational trade-offs, exemplified by C. elegansweighing hunger against desiccation risk. Yet once the neural circuit is known, this "trade-off" reduces to a single inhibitory synapse—an ‘elemental’ account discrediting the ‘molar’ one. My commentary illustrates the same discrepancy in Drosophila: dopaminergic ramping during courtship as it progressively suppresses threat-detecting neurons; so persistence past danger reflects recalibrated valuation rather than a weighed decision. As neurobiology increasingly discovers such discrepancies, all criteria resting on behavioral classifications risk being defined away without altering the behavior's validity.
Sentience evolved to allow organisms to move through their environment avoiding harms and achieving benefits. Sentience did not evolve from computation. It evolved from valenced feelings. AI (even robots) are not embodied and are not the kind of thing that needs to be sentient. There is no way to get from computations to feelings. They are qualitatively different things. We should avoid being lulled into thinking that an inanimate object designed to engage our propensity for anthropomorphism, like AI, is actually a mind.
Strongly supportive overall of Birch’s broad tent Sentience Project, I offer an alternative dyadic, single-class, two-stage process based on estimating the probability for sentience, and the associated confidence, for any individual or taxonomic group. I criticize Birch’s narrow computational framework when addressing the question of algorithmic sentience; there are alternative approaches.
Several commentaries on the Oldfield & Bonano (2024) target article proposed that natural behavior should play a larger role in assessing the welfare of fishes in zoos and aquariums. However, the natural-behavior concept of animal welfare has been widely criticized because natural behaviors do not necessarily improve welfare, and unnatural behaviors do not necessarily reduce it. I review the three major concepts of animal welfare — physical health, natural behavior, and affect/feelings — and argue that behavior patterns, whether natural or otherwise, should be evaluated according to their effects on health and subjective experience. Using the affect/feelings concept and the Five Domains Model, I then re-evaluate several of my previous studies on aquarium fish behavior to illustrate how welfare interpretations can be made more explicit and conceptually consistent.
In The Edge of Sentience, Jonathan Birch (2024) provides a thought-provoking synthesis of current perspectives on animal (and artificial) sentience. He addresses the question from conceptual, ethical, regulatory and empirical viewpoints. Here we focus on some of the scientific issues raised, including defining and operationalising the concept of sentience and the implications for policy and regulation. Despite unresolved conceptual challenges, the book makes a compelling case for the continued scientific investigation of animal sentience.
This commentary assesses the pragmatic framework proposed in Jonathan Birch’s The Edge of Sentience (2024). Celebrating the book's exceptional clarity, scientific rigor, and profound real-world impact on UK legislation, I nevertheless have some questions about the method by which Birch arrives at his policy recommendations for animal protection. In seeking an overlapping consensus between deeply divergent ethical and credal viewpoints, the framework risks defaulting to a minimalist, lowest-common-denominator standard of protection for animals. Birch also risks drawing the standard even lower by including serious academic positions in animal ethics that deny moral status for animals altogether. A rights-based approach offers a standard for animal protection that is more robust, though less likely to be adopted.
There is another side to the story of the account of artificial sentience — one that makes the possibility of artificial sentience unlikely: the fundamental difference between living bodies and artificial systems. Sentience occurs in strongly embodied, living entities, and not in disembodied artificial systems, nor in robots. The appearance of sentience is exacerbated by deceptive behaviour and anthropomorphism. The risks of keeping an open mind about the required underlying substrate are in wasted efforts to protect systems that cannot suffer, and in placing inappropriate trust in their abilities.
Birch’s book is a remarkable exploration of the nature of sentience, its moral salience, evidence for its presence, and how to arrive at humane policies regarding our treatment of human beings, animals, brain organoids, and AI systems of uncertain sentience. Although Birch endorses the view that sentience is of fundamental moral importance, he endeavors to remain neutral among reasonable moral views and work out how to navigate through uncertainty—both ethical and epistemological—to arrive at defensible policies. While making common cause with Birch’s project, my own recent book advances a developed case in favor of the “sentientism” Birch endorses.
Oldfield & Bonano’s (2024) [O&B] target article discusses the wellbeing and sociality of bony fishes. Many fish species, such as cichlids, are not only social but territorial. Aggression is normal, but territorial aggression in a captive environment can have a negative impact on welfare. Aggression-related injuries and deaths are prevalent in pet stores as well as research settings. We discuss the need to develop effective methods to minimize the negative consequences of aggression.
Jennifer Mather has provided a comprehensive and fascinating overview of the various cognitive accomplishments of cephalopods and their sensory, behavioral, and neurophysiological attributes. The advances in our understanding of a clade that has been separated from vertebrate evolution by well over half a billion years is a testament to advances in both methodology and the questions we now feel free to address. Mather embeds her review in the issue of consciousness, which is certainly a perennial issue in mental evolution, going back to Romanes and earlier. I argue that the term “consciousness” has such a checkered history and is used in such diverse ways even today that, despite current popular enthusiasm, it is scientifically misleading.
Who is conscious and how do we know? Complexity in behavior and cognition is certainly a clue that cannot be ignored in identifying this trait, and as Mather’s target article shows, it is hard not to imagine consciousness arising in cephalopods. But how complex is complex? An alternative may be gleaned from the fact that all living things rely on negative feedback to navigate their worlds, and this requires all sentient beings knowing what is to be stabilized in reflection to themselves. If so, consciousness is a universal trait.
Because cephalopods are so different from the vertebrate lineage, their possible consciousness must be assessed on the basis of their heritage and abilities. Cephalopods have evolved in water, a denser medium than air, and their perception is adapted to it. They have inherited the molluscan mantle cavity and hydrostatic muscle type, which has led to a more distributed motor control system despite a centralized brain, and have evolved a completely novel skin display system. Using these abilities, octopuses can form concepts, plan for the future, generate a cognitive map, and self-monitor for apparent pain and manipulate communication. Most cephalopods use the skin display system fairly automatically to deceive potential predators, although cuttlefish and squid also communicate with conspecifics. They do not appear to have the social sophistication to form a Theory of Mind when using it, but some squid can divorce the skin output from its motivational background sufficiently to produce two separately motivated displays in different directions at the same time. Thus, the cephalopods do appear to be moving towards consciousness, but on their own terms rather than our vertebrate terms.
