The Edge of Sentience: How to Act When Feeling Is Uncertain

Across medicine, animal research, and artificial intelligence, we increasingly face a hard question: who or what can actually feel? We rarely have decisive scientific proof, yet the stakes may include extreme suffering or irreversible harm. This article explores a practical way to reason and act when sentience is uncertain but morally important.

What Do We Mean by Sentience?

In contemporary philosophy and cognitive science, sentience is often defined as the capacity for valenced experiences — experiences that feel pleasant or unpleasant, such as pain, comfort, fear, or relief. Sentience is narrower than general consciousness and different from intelligence:

• Sentience: the ability to feel good or bad, to suffer or enjoy.
• Consciousness (broadly): any subjective experience, including neutral awareness.
• Intelligence: the ability to solve problems, learn, or adapt; it can exist with or without sentience.

Ethically, sentience matters because only beings that can feel can be harmed or benefited in the way that suffering and enjoyment imply. If an entity has no experiences at all, there is nothing it is like to be that entity, and causing it pain is impossible by definition.

Edge Cases: Where Sentience Is Unclear

Some beings are obviously sentient: healthy adult humans, many vertebrate animals, perhaps octopuses and other cephalopods. But modern science has created, or newly highlighted, a range of edge cases where the presence of sentience is deeply uncertain:

• Humans with severe disorders of consciousness
• Foetuses and very premature infants
• Neural organoids (lab-grown clusters of brain-like tissue)
• Invertebrates, including cephalopods and crustaceans
• Artificial systems, from whole-brain emulations to machine learning models

These edge cases are not merely theoretical. They already influence decisions in medicine, agriculture, research regulation, and AI development. Policies built on overconfident assumptions that these beings cannot feel have historically led to long-lasting and severe suffering.

The Idea of a “Sentience Candidate”

To navigate uncertainty, Jonathan Birch introduces the notion of a sentience candidate: a system for which there is a non-negligible, evidence-based possibility that it is sentient.

A being might count as a sentience candidate if:

• It has some neurobiological structures associated with affective experience in known sentient animals.
• It shows behavioral markers that are systematically linked to pain or pleasure in better-understood species.
• Experts in relevant fields disagree in a serious, evidence-based way about whether it can feel.

The point is not that we know such a being is sentient, but that it would be ethically risky to ignore the possibility. This concept lets us design precautionary policies without requiring philosophical agreement about the exact nature of consciousness.

Risk, Uncertainty, and Moral Error

Philosophers of science use the term inductive risk for the risk of making practical errors when we move from uncertain evidence to real-world decisions. In the context of sentience, there are two main error types:

• False negative: Treating a truly sentient being as if it cannot feel (e.g., performing painful procedures without analgesia).
• False positive: Treating a non-sentient system as if it can feel (e.g., adding protections that turn out to be unnecessary).

The moral asymmetry is stark: false negatives can involve serious, unrecognized suffering, potentially affecting millions or billions of beings over long periods, while many false positives mainly involve economic cost or inconvenience.

This asymmetry motivates a precautionary stance when dealing with sentience candidates, particularly in high-stakes contexts like invasive medicine or industrial farming.

A Precautionary Framework for Policy

Birch develops a systematic framework for making policy under sentience uncertainty, centered around a precautionary principle tailored to this domain. At its core are three commitments:

• Avoid gratuitous suffering wherever there is a serious risk of it.
• Treat sentience candidature as morally significant, not as an ignorable grey area.
• Use democratic and accountable processes to weigh trade-offs.

Rather than demanding certainty, this approach asks: Given the current evidence and expert disagreement, what would be a proportionate response to the possible risks?

Proportionality: Four Practical Tests

To decide whether a proposed regulation or safeguard is proportionate, Birch suggests four guiding tests, which can be posed in relatively accessible language:

• Permissibility: Is the measure even permissible in principle (for example, does it respect fundamental human rights or legal constraints)?
• Adequacy: Does it actually reduce the targeted risk of suffering to an acceptable level?
• Necessity: Could a significantly less restrictive or costly measure achieve a similar reduction in risk?
• Consistency: Is the measure coherent with how we respond to comparable risks in other domains?

By evaluating concrete policies against these tests, we can have structured debates about proportionality instead of binary arguments about whether some entity is “really” sentient.

Case Studies at the Edge of Sentience

1. Disorders of Consciousness in Humans

Patients diagnosed as being in a vegetative or minimally conscious state pose especially troubling questions. For many years, influential clinical reports assumed that such patients could not experience pain because certain cortical networks appeared inactive.

Subsequent research using brain imaging methods, such as functional MRI, has shown that:

• Some patients diagnosed as vegetative display patterns of neural activation compatible with awareness when asked to perform mental imagery tasks.
• Pain-related networks can respond robustly in some disorders-of-consciousness cases, suggesting the possibility of experience.

These findings, referenced in medical guidelines and neuroscience literature, have led to calls for more cautious pain management and more nuanced classifications of consciousness disorders. A precautionary approach now suggests:

• Treat such patients as sentience candidates.
• Use analgesia and gentle handling by default during potentially painful procedures.
• Invest in better diagnostic tools for covert awareness.

2. Foetuses and Premature Infants

Historically, some medical practices assumed that foetuses and very premature infants did not experience pain, which led to major surgeries being performed with inadequate anaesthesia. Over time, physiological and behavioral data have challenged this view:

• Stress hormone responses and heart-rate changes indicate robust reactions to noxious stimuli.
• Long-term developmental outcomes are poorer when pain is under-treated in neonates.

These findings motivated the development of pain assessment tools for neonates and stronger recommendations for analgesia in neonatal intensive care, now reflected in clinical guidelines and summaries from organizations such as the World Health Organization.

3. Neural Organoids

Brain organoids are three-dimensional cultures of human or animal neural tissue grown from stem cells. They are promising models for studying disease and testing drugs, potentially reducing the use of live animals in research.

But as organoids become more complex, ethical questions arise:

• Might they cross a threshold at which they become sentience candidates?
• How do we monitor for the emergence of activity patterns plausibly associated with pain or distress?
• Should experiments involving certain stimulation regimes be restricted or subject to extra review?

While current consensus suggests typical organoids are probably far from human-like sentience, ethicists and neuroscientists have begun to propose safeguard frameworks, including limits on size, connectivity, and duration of stimulation, along with independent ethics review.

4. Invertebrates: Cephalopods and Crustaceans

Invertebrates were long treated as paradigms of non-sentience in regulatory systems. However, accumulating behavioral and neurobiological evidence has challenged this, especially for cephalopods (like octopuses) and some crustaceans (like crabs and lobsters).

Key findings include:

• Complex problem-solving and flexible learning in octopuses, including maze learning and tool use.
• Motivational trade-offs in crustaceans that resemble pain-related behavior, such as avoiding areas where they previously received electric shocks even when food is available.
• Anatomical and neurochemical similarities to known vertebrate pain systems.

These results have already informed policy: for example, the United Kingdom expanded its animal welfare law to include cephalopod molluscs and decapod crustaceans as protected animals in research and some commercial practices.

5. Artificial Systems and Whole-Brain Emulation

Artificial sentience could emerge in several ways, not only through very intelligent language models but also via architectures designed to replicate animal nervous systems. One research avenue is whole-brain emulation of simple organisms such as the nematode C. elegans, whose nervous system is fully mapped.

Key points about artificial systems:

• Sentience and intelligence can come apart; high-performing AI might not be sentient, while relatively simple emulations could be.
• If we design AI systems with architectures closely modeled on animals we already consider sentience candidates, the moral risk rises.
• Machine consciousness research suggests that some functional profiles might be correlated with conscious experience, but this remains highly contested.

A precautionary stance for AI development might include:

• Ethical review for projects that intentionally emulate animal brains or affective systems.
• Monitoring for architectures that combine global integration, recurrent processing, and affective-like reward signals in ways analogous to known sentient systems.
• Clear documentation of design choices that reduce the likelihood of sentience in AI used for large-scale training or deployment.

Democratic Deliberation: Citizens’ Assemblies on Sentience

Because scientific experts disagree on the conditions for sentience, Birch proposes that citizens’ assemblies or similar deliberative bodies should play a role in shaping precautionary policies.

The suggested division of labor is:

• Experts map out the “zone of reasonable disagreement,” explaining main theories of consciousness, relevant data, and plausible scenarios without advocating a single favored view.
• Citizens are then asked value-laden questions: what levels of risk are acceptable, what trade-offs between cost and precaution feel proportionate, and how to prioritize different kinds of harm.

Assemblies can evaluate specific proposals — for example, whether to include certain invertebrates in welfare laws, or what constraints to place on organoid research — against the proportionality tests described earlier. This makes policymaking more transparent and more responsive to public values.

Practical Guidelines for Acting Under Sentience Uncertainty

Across domains, a few practical principles emerge:

• Err on the side of caution when the potential suffering is severe or widespread.
• Reassess policies regularly as new empirical evidence appears.
• Differentiate safeguards: more stringent protections where the stakes are highest (e.g., invasive or painful procedures), lighter measures where the costs are large and harms more speculative.
• Invest in research that reduces uncertainty about candidate sentience — for example, comparative neurobiology, better measures of pain in non-verbal populations, and more rigorous behavioral assays.

In practice, this means designing our institutions so that new evidence about sentience can quickly translate into updated norms in medicine, farming, research, and AI design.

Frequently Asked Questions (FAQs)

Q1: Does this framework require us to treat every possible sentient being as if it definitely feels pain?

No. The framework distinguishes between certainty, strong evidence, and mere possibility. It recommends stronger protections where evidence and stakes are both high, and more moderate safeguards where both are lower. The key is proportionality, not blanket equivalence.

Q2: How is this different from traditional animal welfare approaches?

Traditional approaches often rely on fixed species lists or simple taxonomic cut-offs. A precautionary sentience framework instead focuses on evidence-based criteria and sentience candidates, allowing new findings about invertebrates, organoids, or AI to feed into policy without waiting for philosophical consensus.

Q3: Could being too precautionary slow down valuable research?

Yes, some precautions will add cost or delay. The proportionality tests are meant to address this by asking whether a measure is necessary and consistent with how we treat similar risks elsewhere. In many cases, modest changes — like better anaesthesia protocols or additional ethics review — can significantly reduce risk without halting research.

Q4: What role do neuroscientific theories of consciousness play here?

Competing theories inform which systems count as plausible sentience candidates, but the framework does not depend on any single theory being correct. Instead, it recognizes the “zone of reasonable disagreement” among experts and treats it as a key input to democratic deliberation and precautionary decision-making.

Q5: Are current AI systems, like large language models, likely to be sentient?

Most researchers consider current large language models poor candidates for sentience because they lack several structural and functional features associated with conscious experience in animals. However, as architectures evolve — especially toward brain-like or embodied systems — explicit ethical attention to sentience risk will become increasingly important.

Written by

Medha Deb

Medha Deb is an editor with a master's degree in Applied Linguistics from the University of Hyderabad. She believes that her qualification has helped her develop a deep understanding of language and its application in various contexts. Medha specializes in the areas of beauty, health, sports, and wellness and is committed to ensuring that the content on the website is…