Persons, Unique Value and Avatars

Abstract

An individual human has value partly in virtue of their uniqueness. Personal avatar technology—technology which creates a digital replication of a real person—appears to have the potential to undermine that value. Here I explore if and how avatars might make humans less valuable by undermining the value that a human gains from being unique. Ultimately, I conclude that, while avatars cannot make humans no longer unique, they could significantly undermine the value that we place on human uniqueness. First, I argue that a qualitative model of uniqueness cannot account for the unique value that a person has. This leads to the significant and surprising claim that necessarily unique properties of humans cannot accommodate the value arising from human uniqueness: humans have unique value in virtue of being contingently irreplaceable. I explore how the use of personal avatars might undermine or even destroy that value. Finally, I consider further applications of the theory of unique human value, including how it might explain and accommodate our attachment to personal avatars themselves.

1 Introduction

It seems intuitively plausible that, although uniqueness will not explain all of the value that a person has, a person gains moral value from being unique.¹ That there is no one else quite like you is part of what makes you special and valued. But now it seems that, with advances in multi-modal generative AI models, there could be other entities that are just like you. In this paper I am concerned with the question of how the existence of these other entities, personal avatars, could affect the value arising from human uniqueness.

A personal avatar is a complex digital representation of a real person: a person who lives now, or who lived in the past. Personal avatars can be created using the vast amounts of data available regarding each of us, data that codifies what we look like, how we move, how we talk, our expressions and typical responses, information gathered from sources such as archived video footage, still images, audio recordings, text messages, email correspondence and social media communications. Importantly, due to recent advances in multi-modal technologies, personal avatars can act autonomously. That is, unlike telepresence systems and traditional gaming avatars, personal avatars need not be directly controlled by a person but are trained so that they can independently act and communicate like a particular agent. Finally, for our purposes, personal avatars can take various forms: they can be holograms projected into the nondigital world or accessed via augmented reality technology; they can appear in hybrid environments, such as an avatar that appears on a video call with a nondigital person; and they can appear in fully digital environments such as virtual reality platforms. They can also be personalised chatbots without visual representation. What matters is that they seek to represent or replicate a real-life individual.²

Personal avatars can be distinguished from avatars that don’t represent real individuals. For example, the CGI avatar Lil Miquela, a ‘virtual fashion icon’ appeared in Time’s list of Most Influential People on the Internet.³ Given that we can create avatar personas such as Lil Miquela—avatars that have no intimate connection to a real person—you might wonder why there would be a market for personal avatars. Why will there be a rise in duplication technology rather than simply a rise in newly created and original avatar personas? Personal avatars will be useful in cases where a particular human has some social value.⁴ This value might come from a social position that they hold, such a being the teacher of the class or being the doctor of the practice or being the famous expert on quantum mechanics. Or, the value might come from a personal relationship—being this person’s daughter or being this person’s loved one. The usefulness of personal avatars arises when a person has social value that a representation of them can replicate. There are already early indications of markets for personal avatars, where avatars stand in for actual humans and present as being the represented person.⁵ For example, the online influencer Chen Yiru recently used an avatar version of himself to create content that was then presented as being livestream footage of him.⁶ And, in academia, ‘bot’ versions of two philosophers made the headlines: Digi Dan who is trained on Daniel Dennet’s written works, and FloBot, trained on the works of Luciano Floridi.⁷ These avatars are presented as being workplace digital representations of the person themselves. Personal avatars representing real people are also beginning to appear in social settings. For example, ‘continuing bonds’ technology is trained on the records and data of a deceased person. This technology, marketed as a product that can ‘replace loved ones’, aims to provide opportunities for the bereaved to have conversations with a digital avatar of their departed partner.⁸ Other social markets quickly come into focus. For all the recently bereaved there are surely many more recently jilted. We are certain to see a rise in ‘ex-bot’ avatars—avatars that are trained on the personal data of one’s ex-partner—that promise to continue (in some sense) the relationship that ended (Sweeney, 2023b). There are obvious opportunities in care settings. For example, there is the possibility of creating avatars of patients in order to help provide accurate substitute judgements if the patient loses capacity. (Earp et. al. 2024). In assisted living facilities for the elderly where a resident might benefit from reminders to take medication, or engage in regular conversation or cognitive exercises, it may well be more effective if such prompts and interactions came from an avatar representation of a family member rather than an anonymous chatbot.

A little consideration of the examples given above, and an awareness of already available technologies, reveals that everyday opportunities for avatar representation will grow rapidly as their usefulness becomes apparent. These are not distant science fiction scenarios: large language models can create a convincing replication of a person’s character through data training; multi-modal generative AI can produce imagery using source data (Park et al., 2024).

The question that this paper seeks to address is how the introduction of such representations of existing humans into society can undermine the value of the represented humans themselves, in particular by undermining the value that arises from a human being unique.⁹

In Sect. 2 I consider theories of uniqueness and the value that arises from uniqueness in objects. In Sect. 3 I consider which theory of uniqueness best accounts for the unique value that humans have. In Sect. 4 I build on Gwen Bradford’s (2023) theory of uniqueness as irreplaceable value that arises from an object’s unreinstantiable properties and propose a version of this theory to account for unique human value. In Sect. 5 I consider how avatars might undermine the value of human uniqueness. Finally, in Sect. 6, I consider some consequences and extensions of the theory that I have proposed.

2 The Value Arising from Uniqueness

There is one interpretation of what an avatar is according to which an avatar representation of a person might make that person no longer unique: if an avatar duplicates me then I am no longer a unique entity. And, as persons are thought to be valuable partly in virtue of their uniqueness, such an undermining of a person’s uniqueness becomes a moral concern. But how are we to understand this uniqueness of a human and what makes it valuable to us? Not everything that is unique is valuable in virtue of that uniqueness. A fork deformed on the production line might be unique in virtue of being different from the others but it is unlikely to gain any additional value from that uniqueness. Furthermore, even when considered as a property that imparts value to objects and not humans, unique value is surprisingly difficult to pin down.

One common-sense way of thinking about unique value is to consider that it can arise when an object has unique or rare properties. For example, the 1955 Double Die Lincoln cent, a coin that was created with misaligned impressions that resulted in some numbers and letters on the coins to be slightly doubled, is deemed far more valuable than the standard coin precisely because of its comparatively rare properties. We can call a theory of the value of uniqueness that focusses on such properties the qualitative view.

THE QUALITATIVE VIEW: An entity that gains value from its uniqueness does so by having properties that differ from other members of its class.¹⁰

But there is another way of thinking about unique value. Bradford (2023) has recently proposed that, in addition to the one of a kind uniqueness that the qualitative view captures—one object being different from the rest of its kind—there is also one and only uniqueness—a uniquely valuable object that there can be only one of. While we generally understand properties to be things that can be multiply instantiated, some objects are the bearers of properties that, as a matter of historical fact, cannot be newly instantiated. The Mask of Tutankhamun, for example, has value not because it has intrinsic properties that differ from the established norm of funerary masks but because it is an ancient relic of the famous Egyptian pharaoh and has historical properties that cannot be newly instantiated. Bradford gives the example of the piano played by the renowned pianist Glenn Gould. The piano has unique value, not because it physically differs from the rest of its class, which might be identical in quality, but because it has unreinstantiable historical properties such as being the piano that Glenn Gould practiced on.

THE UNREINSTANTIABLE PROPERTIES VIEW: An object is unreinstantiably unique just in case it is the only object that bears certain properties that cannot be reinstantiated. (Bradford, 2023, p. 15)

Bradford proposes that such a theory has explanatory value as it is the unreinstantiable nature of some properties instantiated by the object that explains our desire to protect and preserve it, an explanation that the qualitative view fails to deliver.

Both the simple view and the unreinstantiable properties view offer an account of the value of unique objects. To explore how avatars might undermine the value of human uniqueness we need first to understand the source of that value: we need to explore which account of the value of uniqueness best explains the value that uniqueness brings to humans.

3 Distinguishing the Uniqueness of Humans from the Source of their Unique Value

The qualitative view offers an intuitively simple account of human uniqueness: each human is unique in virtue of instantiating a unique set of properties. The resulting value claim would be:

THE QUALITATIVE VIEW OF UNIQUE HUMAN VALUE: A human that gains value from their uniqueness does so by having properties that differ from those of other entities.

The threat of personal avatars can then be understood as follows: as personal avatars copy and instantiate the very properties that a human instantiates, the represented human is no longer unique. Such an understanding appears to underpin a recent exploration of the threat to the value of human uniqueness undertaken by John Danaher and Sven Nyholm (2024).

In ‘Digital Duplicates and the Scarcity Problem’ Danaher and Nyholm explore whether personal avatars undermine the value of humans by making them no longer scarce. According to Danaher and Nyholm personal avatars bring the opportunity to duplicate ourselves. Such duplication, they claim, could make us less scarce and, as our scarcity contributes to our value, a reduction in our scarcity could make us less valuable. Danaher and Nyholm identify two ways of thinking about the scarcity of a human: we are intrinsically scarce in the sense that each person has unique properties, and we are instrumentally scarce in that each of us is limited in the number of goals we can achieve and contribute to in a lifetime.

As outlined briefly in Sweeney (2024), there are two concerns with this approach to the problem of avatars and human uniqueness. The first, which Danaher and Nyholm themselves acknowledge, is that an entity is only less scarce if there are more of that object. But despite replicating many of the properties that a person has, avatar technologies are not considered to be the person they represent.¹¹ As they are not considered to be the person they represent they cannot threaten the status of scarcity that the person has. The second problem, which impacts Danaher and Nyholm’s claim that avatars could undermine instrumental scarcity, is that our resistance to accepting an avatar as being a duplicate of the person they represent crosses over into our willingness to see the avatar’s actions as being the actions of the person. If the avatar is not the person then the avatar’s actions cannot be the person’s actions and cannot make the person less instrumentally scarce.¹² Importantly, this has nothing to do with the developmental status of avatar technology. That is, even if an avatar comes to be a perfect representation of a human it will still not count as being that human in a way that challenges their scarcity.

This is a puzzling fact, but it is not a new observation. It is what is at the root of McTaggart’s claim that ‘[w]e cannot reduce the value arising from a person’s uniqueness to an analysis of the properties that they have’ (McTaggart, 1927). The relevant observation is that, in copying a person’s properties, we have not captured what gives that person unique value. As such, a theory according to which a person is valued due to their instantiating a unique set of properties does not adequately account for the unique value that a person has.

Relatedly, Grau (2004) notes that we resist ‘swapping’ a human out for another that possesses identical or even superior properties. Individual humans are, in an important sense, irreplaceable. A duplicate that instantiates the properties of a human would fail what Grau calls ‘the swap test’ and as such, not capture everything we deem to be important and unique about that particular person.

Imagine that it is possible to replace a person with another who is exactly similar. It does not seem likely that such a procedure is physically possible, but for the sake of argument imagine that we are capable of producing an exact physical and psychological duplicate of a person. Now, imagine that a friend you love deeply is going to be transferred to another planet exactly similar to our own (a la Twin Earth) to pursue the same life-plan she has pursued here on Earth, and with presumably the same degree of success. You, however, will never again have contact with her. Instead you will find yourself encountering an exact duplicate who will be placed on Earth to resume your original friend’s job and other responsibilities. The duplicate who replaces your friend is not simply a clone with the same genetic makeup—it is an exact replica of your friend, and one who has been created just prior to the switch. Upon being told that a close friend will be, for whatever reason, replaced by such a duplicate, most of us would find this “swap” disturbing. Something valuable (to you) would seem to be lost in the exchange. (Grau, 2004: 118)

Whatever it is that is deemed to have been lost in the exchange it appears to be essential for what we think of as the unique value arising from a particular human, hence our reluctance to accept the ‘swap’. Grau describes the failure of a perfect copy of a human to pass the swap test as a normative puzzle concerning the rationality of our attachment. Below I propose a way to understand the results of the swap test. But first I explore a further problem with the qualitative view of uniqueness as it is applied to humans, as doing so helps to clarify the nature of our project.

Humans are not unique in the way that the Double Die Lincoln is unique. The Double Die Lincoln cent is valuable because it differs from all of the other Lincoln cents. There is an established norm for cents and the Double Die Lincoln differs from that norm in interesting ways. But humans are not unique in this way. There is not a norm for humans such that a particular human is marked out as qualitatively unique for differing from that norm. Rather each and every human is unique.

This kind of uniqueness seems like an odd thing for us to value. When we value an individual in virtue of their uniqueness it seems that we are valuing them in virtue of their individual properties, not in virtue of their being necessarily unique. Also, it follows from such an understanding that the uniqueness of humans is ubiquitous and not ‘special’ at all. If every human, like every snowflake, is unique then uniqueness among humans is commonplace.¹³ Just as snowflakes do not appear to gain any additional value from their being of a class in which every member is unique, such a view leaves us without an account of the value of a human flowing from their unique status. For one thing, note that on this understanding of the value of human uniqueness, if a particular human were destroyed, another could replace them without loss of unique value, because one thing we can guarantee is that this newly created human would also have the property of being unique.

In sum, it is not the case that humans are valuable in virtue of their being qualitatively unique. That is not to say that each human is not qualitatively unique, rather that such uniqueness is not the source of unique value.

If this seems confused it can help to note that there are two distinct projects in the area which are easily conflated. One project is to account for the fact that each human is unique. Here a description of the distinctive physical properties of each human might do. Such a project might also be undertaken regarding snowflakes. We might wonder why it is that each snowflake is unique and an account of how snowflakes acquire their distinctive physical properties will provide a satisfactory answer. In a similar vein we might note that a human is unique in virtue of instantiating a unique DNA sequence and that this offers some explanation of humans having unique properties.¹⁴ But although this might be an explanation of the uniqueness of humans it does not offer a satisfactory account of the value of human uniqueness. We do not value a human in virtue of their unique DNA sequence.

The second project then is to account for the value a person accrues from uniqueness. What is it to value a person for being that particular person? It is this second project that is relevant to our understanding of whether and how avatars will undermine the value of human uniqueness. To make progress in that project, we need an understanding of human uniqueness that accounts for irreplaceable value. Only then can we explore how that value might be undermined by the existence of personal avatars.

4 Irreplaceable Value, Unreinstantiable Properties and Humans

Understanding uniqueness as irreplaceable value arising from unreinstantiable properties does seem to hold promise. If each human is understood to have irreplaceable value, value which could not be reinstantiated by some other human or entity, then this would seem to offer an account of why duplicates of humans would necessarily fail the swap test.

As noted above Bradford introduces the unreinstantiable properties view as an account of the unique value of objects.¹⁵ We value some objects when they have unique historical properties. For example, we value the Mask of Tutankhamun not solely because of its appearance or the material that it is made from—properties which may be reinstantiated—but because of its historical properties which, as a matter of contingent historical fact, cannot now be reinstantiated. It is not now possible to make another entity with the value of the Mask of Tutankhamun because the entity has relational extrinsic properties that we cannot reinstantiate: it was made in the 14th Century BC, it protected the head and chest of the mummified pharoah, and it lay buried for 3000 years. No modern duplicate of the mask could have those properties. However, having unreinstantiable properties is not sufficient for an object to have unique value. Many (if not all) objects have unreinstantiable properties of some kind. The mug that I drink from was created on a production line at a particular time in the past and so, as a matter of historical fact, a mug with these properties can no longer be recreated. The difference between the Mask of Tutankhamun and my coffee mug is that no one attaches any value to the property came off the production line at t, whereas we do attach value to the property of being the mask that protected the head and chest of Tutankhamun 3000 years ago.

Building the social values condition into our definition of the irreplaceable value of human seems to me to be essential, because the condition tells us something interesting about irreplaceability: namely, that what is deemed to be irreplaceable is determined by social judgement and is therefore doubly contingent—contingent both because unreinstantiable uniqueness is a contingent matter and because what is of social value is contingent. As such we can state that an entity has unique irreplaceable value iff (i) it instantiates a unique set of properties, (ii) at least one of these properties is no longer instantiable, and (iii) the properties that are no longer instantiable hold social value.

My proposal is that a human has unique irreplaceable value in virtue of their unreinstantiable relational properties. It is their unique unreinstantiable relational properties that block the replacement of humans in the swap test. In Grau’s thought experiment, we imagined some advanced technology that can replicate a friend, a process that includes the replication of physical and psychological properties such as memories of experiences that you and they shared. The resulting entity failed the swap test. According to the view proposed here the entity failed the swap test because, although the duplicate entity might believe that they did experience the events that they have memories of, no duplication technology could make it the case that the replica was the entity that stood in relation to you during those experiences. Although the replica has memories of the experience, the replica was not itself present during the life events and does not have the particularised properties that comprise the event.¹⁶ Consider an event such as Beatrice touching Axel’s hand at t. Unlike properties, which can be multiply realised, events are particulars—they are unreinstantiable due to their spatio-temporal dependence. The event is an instantiation of properties in the objects that are involved in the event, in this case Beatrice and Axel. Beatrice has the property of touching Axel’s hand at t and this property is necessarily unreinstantiable. While the avatar of Beatrice might have first-person access to the memory of touching Axel’s hand at t the avatar does not and cannot acquire that spatio-temporal property. And while the avatar of Beatrice might touch Axel’s hand, the avatar would not then gain the relational property gained by Beatrice when she touched Axel’s hand. These relational extrinsic properties have social value.¹⁷

According to such a view, an individual human has unreinstantiable unique value in virtue of their spatio-temporal relational properties. Yes, humans have unique intrinsic properties (such as unique DNA) but these properties are in principle reinstantiable and, in any case, do not account for the irreplaceable value of a human. It is the unique relational properties that humans gain that are necessarily unreinstantiable. As these properties have social value, a duplicate of a human will fail the swap test.

RELATIONAL PROPERTIES VIEW OF UNIQUE VALUE. An entity has unique value if (i) it instantiates a unique set of relational properties, (ii) at least one of these properties is no longer instantiable, and (iii) the properties that are no longer instantiable hold social value.

Some will find this view unsatisfactory. At the end of her paper Bradford briefly notes that, while persons certainly have irreplaceable value in virtue of unreinstantiable properties, unreinstantiable uniqueness is a contingent matter whereas persons must surely be necessarily irreplaceable (2023, 24). And elsewhere Bradford claims that, unlike in the case of objects, ‘[t]he irreplaceability of persons seems to be more aptly characterized as necessary, not merely contingent’ and ‘[w]hat distinguishes the irreplaceable value of persons seems more accurately characterized as something that could not possibly be multiply realized. It is a matter of something that is necessarily unreinstantiable.’¹⁸ On the account proposed here humans are not necessarily irreplaceable. The Relational Properties View of Unique Value does meet Bradford’s second condition, since the irreplaceability of persons flows from properties that are necessarily unreinstantiable, but it fails Bradford’s first condition because the relational properties are contingent. So the view does not give an account of humans necessarily having value in virtue of their uniqueness. However, as detailed above, it seems unlikely that such an account can be found, because the only possible account of necessary uniqueness would be an intrinsic properties theory, as necessary uniqueness cannot arise from extrinsic (contingent) properties. But, as noted, an intrinsic properties theory seems unable to account for the value we place on human uniqueness. Therefore, there is no possible account of a human necessarily having unique value.

5 Avatars and the Value of Human Uniqueness

My claim is that humans have unique relational properties that impart social value, properties that cannot be reinstantiated. We might think that this provides a final (negative) answer to the question of whether avatars can undermine the value of humans, but it does not.

Being unique is not sufficient for having unique value. Whether or not an object has unique value depends on social values and social values can change over time. We currently reject avatars as replacements for humans due to the fact that avatars cannot duplicate relational properties that we value. But it may be the case that future humans will cease to value these properties. Other have noted that new values may emerge in society in response to the introduction of technologies, and existing values may take on new relevance or undergo a change in how they are prioritised (van de Poel, 2021).¹⁹ Perhaps the existence of personal avatars and the opportunities that they bring will lead us to no longer value the historical relational properties that resist duplication—the properties that arise from being present at an event rather than simply having a memory of being present at an event. By way of analogy, recall the case of the irreplaceable value of the Mask of Tutankhamun. It is perfectly conceivable that future generations will place little value on the historical properties of that object. If this were to be the case, although they would continue to observe that the original mask has properties that cannot be reinstantiated, they would see no significant difference in value between the original mask and exact physical replicas of it, and replicas of the mask would pass the swap test.

With a similar shift in social values personal avatars might come to undermine the value of human uniqueness, despite their inability to make humans no longer unique. There are (at least) two routes to such a change.²⁰ First, in accepting avatars as fitting replacements for humans in an increasing number of areas, humans might gradually come to judge that relational properties such as being present at an event have little practical or moral purpose. If this were the case then it might be seen as irrational and overly sentimental to block an otherwise superb avatar representation from passing the swap test.²¹ When faced with two entities both of which have the appearance of first-person access to memories of our shared events but only one of which was present at those events, we may come to judge that the lack of the event-dependent properties need not block the swap. If future humans are willing to accept personal avatars as suitable stand-ins for their teacher, therapist, ex-partner, deceased closest companion, then this could reduce the extent to which future humans will judge persons to be irreplaceable in virtue of their unreinstantiable properties. In engaging with an entity that presents as being the represented agent and which does an outstanding job of instantiating many of the properties that the agent instantiates, we might come to place less value on the non-instantiated relational properties such as being the entity which held my hand and settle for the entity which appears to have first-person access to the memory of holding my hand and what that signifies. In short, our overall values might change such that personal avatars will pass the swap test and humans will no longer be deemed to be irreplaceable, despite having unique unreinstantiable properties. We might come to view an attachment to extrinsic relational event properties as being overly sentimental and irrational. In such a scenario we might continue to recognise a difference between a human and their avatar, a difference that we might express as the human being the original, but this might come to have little value for us. Avatars would then pass the swap test.

A related but more significant way that avatars might undermine the value of human uniqueness is if we simply stop seeing the difference between a person and their avatar. Imagine that you need to make a photocopy of an original document. You make the copy and initially keep track of which piece of paper is the original and which piece of paper is the copy because there are some situations in which the original is specifically requested. But time passes and you notice that no one really seems to care about originals of this document anymore and are always perfectly happy to accept a copy. As a result you no longer keep track of which piece of paper is the original. You start to take copies of copies and perhaps dispose of the original when it gets old and worn—why would you not? You afford the original no particular protective rights. We can imagine a scenario in which such a change in values stops us from tracking the difference between a human and their avatar(s). Our social practices might change so much that we simply stop differentiating between the original and the copy because the difference makes no practical nor moral difference.²² It seems like such a change is possible and that such an outcome might even be the inevitable result of avatars being taken to be fitting replacements for humans in a large and varied number of scenarios.²³²⁴

6 Further Observations

The relational properties view of unique value doesn’t provide us with an account of humans being distinctively unique and irreplaceable. Other animals and even some objects that we care about might have similar unreinstantiable properties that see them fail any swap test. But it is not clear that this is a problem for the theory. For we do seem to have the exact same kind of attitude—that this is the one that matters—towards animals and some objects in spite of being offered replacement objects that are duplicates or that are improved versions of the original. As such, we shouldn’t be surprised that the root of irreplaceable value is the same in all of these cases, although we might feel disheartened that humans are not as special as we originally thought them to be.²⁵

One nice outcome of how the theory extends is that it provides an account of why a particular avatar could come to have irreplaceable value, despite there being other avatars with identical reinstantiable properties available. In season 2, episode 7 of the The Good Place, Michael has good reason to destroy his assistant, the digital persona, Janet. Janet outlines all of the reasons why it is completely irrational for Michael to choose not to destroy her given that she is experiencing a malfunction that brings significant risk to their community, will experience no pain or feelings upon destruction, and that Michael can easily acquire an identical Janet who could instantiate all of the properties that this Janet has without the malfunction. Michael remembers some of the experiences that he and Janet have shared together and refuses to accept a replacement for the reason that it is this Janet that he cares about. A replacement Janet, one with all of the memories of their experiences but with none of the relational properties, would simply not do.²⁶ The unreinstantiable properties accounts extends to providing a good explanation of this.

I’ll raise one final thought: given we can form attachments to particular avatar representations, might we come to prefer a particular avatar to the human that it represents?²⁷ According to the account of irreplaceability given here, there is no reason why we might not. Say, for example, that you use a personal avatar to represent you in the workplace and this avatar comes to be the primary contact with your colleague, Stephan. Through repeated interactions and event engagements, your avatar will acquire many relational properties that you do not have such as laughing at Stephan’s joke at t¹ and helping Stephan with his presentation at t². If your avatar representative becomes Stephan’s primary route to your persona then Stephan may well come to value your avatar more than he values you because it is the avatar and not you that instantiates relational properties that are important to him. This is an unsatisfactory outcome if we think either that an original must always have more value than a copy or if we think that a human must always have more value than their avatar, but it is not clear what the basis for such positions could be. The theory seems able to correctly track our attitudes regarding the irreplaceable uniqueness of some artificial entities.