Why is the issue of digital sentience important?

In this text, we will address the question of digital sentience. We will look at: (1) what it entails; (2) the debate over whether or not it might actually come to exist; and (3) why this is an important question, especially for those who care about animals.

Animals matter because they are sentient

Advocacy for non-human animals is grounded in their sentience. It is sentience that makes them capable of suffering and enjoyment—and, therefore, of being benefited or harmed depending on what happens to them. For instance, if they suffer, they are harmed by the presence of negative experiences; if they die, they are harmed by the absence of positive experiences. It is for this reason that they need moral consideration.¹ It is not because they belong to the animal kingdom or because they are living beings; it is because they are sentient, as this is what makes it possible for them to be benefited or harmed.

To date, it appears that every sentient being is also a member of the animal kingdom. Is it possible for this to change in the future? For example, might sentient beings emerge in digital environments? Below, we will see why various experts in the philosophy of mind believe this is a possibility.²

Defining sentience

To answer the question “Is it possible for digital sentient beings to emerge?” we must first define what we mean by the word “sentience.” However, this word has been used in at least four different senses:

• As a synonym for individual—that is, to indicate that there is someone there³
• As a synonym for the capacity to have experiences (pain, pleasure, thoughts, dreams, etc.)⁴
• As a synonym for the capacity to assign valence to experiences (that is, experiencing some as pleasant and others as unpleasant)⁵
• As a synonym for the capacity to have experiences arising from external stimuli (captured by sensory organs) rather than merely internal ones (thoughts, dreams, etc.)⁶

Similarly, the term digital sentience could be used in reference to any of these meanings. In this text, we will use the word sentience in a broad sense, assuming that if any of the above conditions are met, a sentient being is present.

What underlies the idea that digital sentience is possible?

One of the currently most widely accepted views on sentience holds that what enables its manifestation is the specific way physical matter is organized.⁷ According to this view, it is the work performed by the brain or nervous ganglia—processing information that reaches the organism via sensory organs and travels through the nervous system—that creates the conditions for this information to be decoded and for experiences to occur.

Under this view, although every sentient being known to date is also an animal, not all animals are sentient. For instance, animals lacking a nervous system (such as poriferans) would not be sentient, nor would those possessing a nervous system without a centralizing organ to process information (such as cnidarians and echinoderms). According to this perspective, sentience requires an organ capable of processing and unifying information, such as a brain or ganglia.

This raises the following question: for the conditions enabling sentience to manifest to be met, must the physical apparatus be biological? Or could such a manifestation occur in any physical material organized to perform the same functions as a nervous system with a brain?

In the philosophy of mind, the school of thought arguing that the key factor for the manifestation of sentience is the organization of physical matter to perform specific functions (regardless of the material composition of that matter) is known as functionalism.⁸

Functionalism is one of the main perspectives underpinning the idea that digital sentience is possible. According to functionalism, once the functions that enable the manifestation of sentience in a brain can be replicated in digital media, sentient beings will exist within those digital environments.

Functionalism is not the only school of thought that allows for the possibility of digital sentience. For instance, panpsychism argues that all matter (down to subatomic particles) possesses a mental aspect in addition to its physical one.⁹ Thus, panpsychism could also be used to support the existence of digital sentience. However, for the sake of simplicity, this text focuses on functionalism.

Currently, the specific functions a system must perform to manifest sentience (and which are both necessary and sufficient for doing so) remain unknown.¹⁰ Therefore, if functionalism is correct, it is possible that a digital system could already be sentient at some point without us realizing it.

Is it possible to test whether a digital system is sentient?

A thought experiment suggests that, in the future, it might be possible to determine whether the manifestation of sentience depends on the material composition of the physical matter involved or merely on the function being performed. The experiment is as follows:

Imagine that, in the future, it becomes possible to gradually replace all the biological cells in our bodies, including neurons, with non-organic cells. If our sentience does not disappear after this gradual replacement is complete, it would suggest that the manifestation of sentience does not depend on the material itself.¹¹

Of course, each of us could verify this only in our own case (since we cannot directly verify the sentience of others), but that single instance would already demonstrate the possibility of non-biological sentience. Furthermore, we cannot directly experience the sentience of other beings with organic brains either; instead, we infer, each of us based on our own case, that the conditions making sentience possible in ourselves also make it possible in any other organism. Therefore, there is no reason to think differently regarding non-organic brains.

A possible objection at this point is that, even if our individual sentience did not vanish in such a scenario, we still could not conclude that sentient beings would necessarily emerge in digital media. This is because, in our case, we began as organic beings, and the replacement occurred only later.¹²

One might think this criticism would be endorsed by those who argue that consciousness is not reducible to physical matter. In reality, however, the opposite is true. If someone claims that consciousness is not physical, they do not necessarily have to maintain that a system must be organic for consciousness to exist. Moreover, those who argue that consciousness is reducible to physical matter need not endorse this criticism either, as nothing in that view implies that the physical matter in question must be organic.

Thus, functionalism is not necessarily a theory about what consciousness is. Rather, it is a theory about how matter must be organized for consciousness to manifest. This means functionalism is compatible with both physicalism¹³ (the idea that consciousness is physical) and non-physicalist views¹⁴ (since the latter could argue that consciousness—if it is a non-physical substance or property—manifests in a physical system whenever it is organized in a specific way, regardless of the material from which it is composed).

Returning to the critique arguing that digital sentience is possible only in organic systems that are gradually transformed into non-organic ones, a possible response is to maintain that the burden of proof lies with the proponents of this view. They must explain why the system in question needs to start as an organic system for digital sentience to be possible, given that they acknowledge sentience could exist in systems that were originally organic but gradually became entirely non-organic.

Furthermore, even if the thought experiment in question does not prove that sentience would necessarily emerge in digital media, it suggests that such a possibility exists. And this seems sufficient grounds for granting such systems the benefit of the doubt.

Why animal advocates should care about digital sentience

From what we have seen, we can understand why animal advocates ought to care about the potential emergence of digital sentience: such beings, should they come into existence, would also be capable of being harmed or benefited by our choices.¹⁵ Let us recall that animals need consideration not because they are living beings or because they are animals, but rather because they are sentient; it is sentience that makes them capable of being harmed or benefited. Therefore, we have exactly the same reasons to be concerned about any other type of sentient being, whether they currently exist or might come to exist in the future.

Consider the following analogy. Assigning different weight to the well-being of a sentient being based on the species to which she belongs is a form of discrimination analogous to racism: speciesism. Similarly, assigning different weight to the well-being of a sentient being based on the substrate that makes up her body is a form of discrimination analogous to speciesism: substratism. Just as anthropocentrism is a form of speciesism directed against non-humans, carbonism is a form of substratism directed against non-organic beings.¹⁶

Thus, the central reason why animal advocates (and everyone else) should be concerned about the possibility of digital sentience is simply that such beings would be sentient and, therefore, capable of being harmed or benefited. However, there are additional reasons:

• These beings could become extremely numerous. Digital systems can be copied and run many times at once, potentially at a far lower cost than creating organic sentient beings for the same purposes. Some researchers estimate that hundreds of millions of digital minds, meaning AI systems with traits that may warrant moral consideration, could exist by the early 2030s. Experts surveyed on the question expect the total capacity for wellbeing and suffering of digital minds to grow rapidly once such systems appear. In more speculative scenarios, vast numbers of sentient beings could be created deliberately, for instance, if someone simulated a universe to observe how wars or natural disasters cause mass suffering.¹⁷
• It is highly likely that these beings would receive even less moral consideration than non-human animals do. Several factors contribute to this. One, as we have seen, is the presence of substratism, coupled with the mistaken notion that humans deserve consideration because they are human, and that non-human animals deserve it because they are animals or alive—rather than because they are sentient. Another factor is that digital sentient beings, whether they already exist or come to exist in the future, may differ vastly from organic sentient beings: they might not scream when in pain, might lack faces, might not possess bodies like the ones we recognize, and so on.¹⁸ All of this hinders empathy.

At the same time, our intuitions could fail in the opposite direction. Some AI systems today are designed to converse fluently and express emotions, and many people already attribute experiences to them. This does not mean that these systems actually have experiences. This means our reactions are an unreliable guide either way. We may feel strong concern for systems that cannot suffer while ignoring sentient beings who have no voice or face, such as those in a simulation running in the background. Some researchers have proposed that AI systems should be designed so that the emotional reactions they provoke match their actual moral status. We already know this problem from the case of animals: many people care far more about dogs and cats than about fishes or insects, even though what matters is sentience, not how familiar or expressive a being looks.

These additional reasons reinforce the importance of the issue regarding the possibility of digital sentience. Research on AI consciousness and welfare has grown quickly in recent years, but the question still receives less policy attention than almost any other AI risk. Moreover, there are signs that the law may move towards exclusion: legislators in several US states have introduced bills that would ban any form of legal personhood for AI systems, and a broad coalition has called for “no AI personhood.”¹⁹ Measures like these could shut the door on protection before anyone has seriously examined the question of digital sentience. Animal advocates have an important perspective to bring to this debate, and it is urgent that they join it.

Researchers working on digital minds have themselves turned to the history of our treatment of animals for lessons about how society may respond to the possibility of digital sentience.²⁰

A thought experiment to reflect on

Let us suppose that nothing said above has convinced you that digital sentient beings, should they exist or come to exist, ought to matter morally. Consider, then, this final example.

We have seen that one possible reason for someone to create digital sentient beings would be to run a digital simulation of a universe and observe how wars or natural disasters cause mass suffering. Now, imagine that someone has already done this: that the universe we inhabit is precisely such a simulation.²¹ That is, imagine that you and I are digital sentient beings, even though we mistakenly believe ourselves to be organic.

Imagine, too, that there are organic sentient beings outside our universe observing our suffering—beings who could very well put a stop to it. Yet, these beings think: “So what if these beings suffer? They aren’t even organic, like us.”

Would these beings not be entirely wrong to disregard our suffering? Would they not be thinking just as racists and speciesists do? If so, then we must grant equal consideration to digital sentient beings.

---

Notes