Article Text

Download PDFPDF

Black, white or green: ‘race’, gender and avatars within the therapeutic space
  1. Mark A Graber1,2,
  2. Abraham D Graber3
  1. 1Department of Emergency Medicine, University of Iowa Carver College of Medicine, Iowa, USA
  2. 2Department of Family Medicine, University of Iowa Carver College of Medicine, Iowa, USA
  3. 3Department of Philosophy, University of Iowa, Iowa, USA
  1. Correspondence to Mark A Graber, Department of Emergency Medicine, University of Iowa Hospitals and Clinics, 200 Hawkins Drive, 1008 RCP, Iowa City, IA 52242, USA; mark-graber{at}


Personal identity is critical to provider–patient interactions. Patients and doctors tend to self-select, ideally forming therapeutic units that maximise the patients' benefit. Recently, however, ‘reality’ has changed. The internet and virtual worlds such as Second Life ( allow models of identity and provider–patient interactions that go beyond the limits of mainstream personal identity. In this paper some of the ethical implications of virtual patient–provider interactions, especially those that have to do with personal identity, are explored.

  • Medical ethics/bioethics
  • philosophy of medicine/healthcare

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Personal identity is critical to provider–patient interactions. Patients and doctors tend to self-select, ideally forming therapeutic units that maximise the patients' benefit. Part of this decision is based on each party's ‘race’ and gender.1 2 Though one cannot make a blanket statement encompassing all practitioners, female and male doctors tend to have different practice styles.3 Moreover, patient satisfaction is increased when the patient and practitioner are of the same ‘race’.4 5 While this might not be ideal, it is reality. (Note: we are choosing to put the word ‘race’ in quotation marks as an acknowledgement that it is a controversial term that remains ill defined, yet useful.)

The internet and virtual worlds such as Second Life ( allow for provider–patient interactions that go beyond the limits of mainstream personal identity. During these interactions, providers and patients are each represented by an ‘avatar’, a virtual construct named after the worldly incarnations of the Hindu gods. Avatars may or may not share physical similarities with the person they represent and indeed may be purposeful misrepresentations.6 7

The purpose of this paper is to examine some of the issues related to personal identity and virtual therapeutic relationships. We will then draw conclusions about the ethical permissiveness of a provider choosing a ‘race’ or gender of avatar that is discordant with their appearance in the non-virtual world. This is not just idle navel gazing. Online therapeutic relationships currently exist and will likely continue to grow with the maturation of the technology. Some examples of current online therapeutic relationships include virtual psychotherapy and counselling.8–11

Definition of terms

We will use the term ‘provider’ to encompass anyone who may have a professional virtual therapeutic relationship with a patient. When we use the term ‘avatar’, we are talking about the visual representation of person in the virtual world. Clearly the visual aspect of a person is only one facet of personal identity. We will use the term ‘non-virtual world’ to refer to our corporeal existence instead of the term ‘real world’.12 The term ‘real world’ has built into it the connotation that other worlds are not ‘real’. This creates a psychological bias minimising the ‘realness’ and impact of virtual worlds. When we use the term ‘discordant avatar’ we mean an avatar that has phenotypic characteristics that are intentionally different from those possessed by the individual being represented.

Defining virtual telemedicine

In large part, definitions of telemedicine are broad enough to cover the scope of this paper13 14: for example ‘any system of care in which the doctor and his [sic] patient are at different locations’.15

For the purposes of this paper, we suggest a bifurcation in telemedicine between ‘classical telemedicine’ and ‘virtual telemedicine’. The fundamental characteristic of ‘classical telemedicine’ is not the technology being used; it is the nature of the interaction. ‘Classical telemedicine’ is characterised by the substitution of a telephone or live video communication for face-to-face interaction.16 If the internet is used to eliminate face-to-face interactions or to replace face-to-face interactions with video, the interaction continues to be classified as classical telemedicine.

‘Virtual telemedicine’ is characterised by the replacement of face-to-face interactions with a facsimile that is not intended to be an accurate recreation of a face-to-face interaction. A patient–provider interaction mediated by avatars is the paradigm of ‘virtual telemedicine’. Virtual telemedicine raises novel considerations not found in classical telemedicine. The purpose of this paper is to explore one such consideration.

Motivating the question

The ethical question we will address is relevant and important to contemporary medical practice. Telemedicine can be a powerful therapeutic tool for a number of reasons. Consider the ‘online disinhibition effect’ (ODE), specifically benign disinhibition, in which individuals are more willing to share personal information online towards the end of an emotional catharsis.17 The benign ODE could be strikingly useful within a therapeutic context. While the ODE has therapeutic promise, telemedicine poses well known difficulties for the provider–patient relationship. In particular, ‘providers may not be able to establish rapport with remote clients because of the impersonal nature of the service’.18 This is, in part, caused by technological limitations that mask non-verbal communication. In this regard, video conferencing may be preferable. Although video conferencing can help to close the interpersonal gap inherent in all telemedicine, it also has the potential to undermine the ODE. Suler writes that ‘anonymity is one of the principle factors that creates the disinhibition effect’.17 Moreover, ‘even with everyone's identity known, the opportunity to be physically invisible amplifies the disinhibition effect’.17 Video conferencing, for obvious reasons, undermines anonymity and destroys physical invisibility.

It seems likely that the use of avatars would allow one to, in part, overcome the difficulties posed for the provider–patient relationship by telemedicine. As demonstrated above, patients feel more or less comfortable with their provider depending on various physical traits of the provider. There is at least some reason to believe that patients will react similarly to these traits when present in an avatar as they would react to these traits present in a non-virtual person.19 By manipulating these traits, one may be able to gain some of what is lost in the patient–provider relationship while not undermining the therapeutic benefits of the ODE.

We are not automatically endorsing this role for avatars. An endorsement would require empirical evidence of benefit that we do not currently possess. However, for those sceptical of the premise of this paper, we hope to have provided at least one reason to find the ethical question at hand a pressing one.

Virtual provider–patient relationships

This discussion is moot if there is not a virtual telemedicine provider–patient relationship that has the same the same rights and responsibilities of that in the non-virtual world. Despite the differences between classical telemedicine and regular medicine, there is no question that a provider–patient relationship can be established.20 21 For the remainder of this paper let us stipulate that the provider–patient relationship in question is as close to a paradigm instance of a provider–patient relationship as possible. The provider and the patient both acknowledge that there is a relationship; there is a tacit contractual agreement and the provider is reimbursed.

Clarifying the project

It is essential that we be clear regarding our goal. When considering the moral permissibility of some action, one can consider its prima facie permissibility or its ultima facie permissibility. Something that is prima facie permissible may be ultima facie impermissible once there is additional information such as knowledge of an outcome or conflicting obligations. We intend to argue only for the prima facie permissibility of the use of discordant avatars. It may well be that, in the light of as-yet ungathered empirical evidence, it turns out that the use of discordant avatars is ultima facie impermissible. This is not to undermine the importance of our project. The opening salvo in the debate regarding the ultima facie permissibility of the use of discordant avatars is to establish that their use is prima facie permissible. If the use of discordant avatars is not even prima facie permissible, there is no need for further debate.

In the remainder of the paper we will argue that the use of discordant avatars in treatment is prima facie morally permissible. First, we will present two arguments supporting the use of discordant avatars and deal with objections to these arguments. We will then address what we believe to be the most troubling objections to the use of discordant avatars.

Two arguments for the (prima facie) moral permissibility of the use of discordant avatars

First argument

Having a discordant avatar is analogous to how we currently change our identity within the provider–patient relationship in the non-virtual world. For a convenient example, consider a doctor who at home wears black, is a ‘headbanger’ and wears multiple piercings. (Please don't send correspondence about our choice of an example. We realise that headbangers are often nice people. Some of our best friends are headbangers.) While this persona is perfectly appropriate at home, it may not work so well in the office. One makes some aspects of one's personality prominent when at work and others more prominent when at home. Moreover, we change who we ‘are’ from patient to patient visually and with the content of our conversations. We are more formal and may wear our white coat when seeing some patients. We are more casual with other patients; perhaps even seeing some patients wearing a t-shirt and shorts. We may be relatively guiding and paternalistic with some patients while we practice absolute shared decision making with others. We rely (ideally) on patient feedback to determine if our current ‘self’ is the appropriate one. The only difference in the virtual world is the degree to which we can choose the ‘person’ we project. In a virtual world we can change clothing and adornment as well as phenotypic ‘race’ and gender too. We are free to choose any representation we desire. Given that the moral permissibility of the selective presentation of oneself within the provider–patient relationship is largely unquestioned, it follows that the use of a discordant avatar would also be allowed. The burden is on the opponent of the use of discordant avatars to demonstrate the morally relevant difference between the two cases.

In response, one may claim that phenotypic ‘race’ and gender are so fundamental to our identity that they are sacrosanct and inviolable. This seems incorrect for several reasons. First, defining someone's ‘race’ and gender as sacrosanct characteristics automatically lessens the role of higher cortical functioning in defining the individual. Second, ‘race’ and gender are social constructs (to prove that ‘race’ and gender are social constructs is beyond the scope of this paper and we refer the reader to the references).22–24 To demand that ‘race’ and gender are sacrosanct and inviolable to one's identity is to undermine an individual's autonomy by placing social constraints on choices fundamental to living one's life as one desires. Consider, for example, a transgender individual. For such an individual, living a fulfilling life may require that s/he identify as a gender distinct from the gender assigned to him/her by social convention. One can easily imagine a similar situation in which the self-identification of ‘race’ may evolve with time.

Finally, consider the following thought experiment. Picture an opaque screen. The patient is on one side of the screen and the therapist on the other. They can speak freely but cannot see each other. One may be able to guess the gender of the other based on his or her voice. Now add a voice processor to the therapist that allows for natural speech but does not give away the therapist's gender. Nothing has really changed in the interaction the individuals can have. Now change the ‘race’ of the therapist on the other side of the screen. Again, nothing of significance has changed in the interactions they can have. However, once one changes the higher cortical functions of the therapist (eg, less skilled with a lack of empathy) the entire nature of the interaction changes.

A more powerful objection holds that shared gender or ‘race’ entails a background of shared experience that allows for greater depth of understanding and empathy. While one may be able to simulate phenotypic variance via avatar, one will likely never be able to simulate shared experience and semantic patterns. We have several responses to this objection. First, patients see practitioners of a different ‘race’ and gender all of the time. The assumption is that any well trained practitioner of any ‘race’ and gender is at least competent to care for any patient. It is up to the patient to decide if the interaction is valuable. If not, he or she can seek care elsewhere. This does not change in virtual telemedicine. Second, not all provider–patient interactions are optimised in the non-virtual world and there is no reason to believe that they should or would be in virtual telemedicine. Third, the expectation that an avatar may not be a facsimile of the user will likely mitigate some of the potentially negative consequences. Online users are used to the employment of discordant avatars. Most users will likely have encountered this situation multiple times before it comes up in a therapeutic relationship. Giving that the use of discordant avatars is often understood to be a part of the online experience, it may well be a mistake to expect patients to be surprised at their practitioner's use of a discordant avatar.

We do not deny the importance of this question. It may well be that, the inability of a provider to appropriately emulate the speech patterns and content of the ‘race’ or gender associated with the discordant avatar is such an overwhelming problem that the use of discordant avatars is ultima facie impermissible. But the very need to collect this data argues for the prima facie allowability of discordant avatars.

Second argument

Patients have the right to expect the maximal therapeutic benefit for their money and we have a fiduciary responsibility to provide it. We have already noted that the gender and ‘race’ of a provider enters into the calculus of patient provider choice and satisfaction. As providers we should do everything (within reason) we can to maximise the therapeutic effectiveness of an interaction (maximal beneficence). We would argue that providers have a duty to be represented by an avatar that will maximise the patients' outcome. How to choose the representation of one's avatar to maximise therapeutic benefit is difficult to know. One way to do this would be to allow the patient to choose the representation of the provider. This allows the patient to tailor his or her experience (at least visually) to meet his or her definition of a comfortable interaction. Or, the patient could just leave the selection up to the provider. Interestingly, this could lead to each person in a virtual treatment group having a ‘different’ therapist/group leader. As noted above, we are not commenting on the effectiveness of such a system, but rather on its permissibility.

Arguments against the use of discordant avatars

The first, and strongest, objection to using a discordant avatar is that it is possible that one is engaged in deception/lying. If the choice of both avatars is left to the patient, no deception occurs. There is no issue. If the choice of the therapist's avatar is left to the therapist, the question then becomes one of whether deception is ever allowed within a therapeutic relationship even if it is acknowledged to be probable by both parties. In what follows we will give two distinct responses to this argument. First, we will argue that, if the use of a discordant avatar constitutes a genuine case of deception, it is not a case of deception that is morally impermissible. Second, we will argue that, in fact, the use of a discordant avatar does not in fact constitute a genuine case of deception.

There are two powerful reasons to find lying on the part of a provider problematic: lying undermines a patient's autonomy and lying on the part of the provider threatens to undermine the provider–patient relationship. There are important differences in these objections. Autonomy is a deontological notion; that is, it derives from a duty or obligation on the part of the provider. One cannot respond to objections regarding autonomy by showing that lying may have positive consequences. The only way to show that lying is permissible within a provider–patient relationship is to demonstrate that it does not importantly undermine autonomy. The second objection, that lying threatens to undermine the provider–patient relationship, is consequentalist in nature. If one can show that the benefits offered by lying will overwhelm the probable negative consequences, the second objection can be defused.

Lying undermines autonomy by restricting a patient's access to information. Deprived of information (in this case on the nature of the provider) the patient has a more difficult time making informed decisions, this in turn limits the patient's choices. The lying constituted by a provider's dissimilarity to her avatar is not unmotivated. The goal is to facilitate treatment. In particular, the claim is that by choosing a discordant avatar that more closely resembles the patient, the provider can overcome potential biases on the part of the patient, making the patient more open to the provider. The expected outcome includes a patient more easily capable of hearing information and more easily capable of integrating that information into his decision making. The important point is that, somewhat paradoxically, lying in this instance may, in fact, increase autonomy by increasing patient understanding and openness to ideas. It can be argued that this smacks of manipulation and paternalism with the provider making decisions for the patient, to whit the nature of the provider's avatar. However, as noted above, the patient has the right to choose the provider's avatar or leave the decision up to the provider.

One may point out that, in principle, lying involves decreased autonomy and that this case cannot be any different. The important questions to be asked, then, are: (1) the extent to which autonomy is lost and, (2) the type of autonomy loss. The autonomy loss (if any) is minimal; few choices on the part of the patient will be dependent on the physical characteristics of the provider. The trade off then seems to be positive quantitatively and qualitatively. Additionally, even were there a net loss of autonomy, there is no proscription against a patient limiting his autonomy voluntarily. When using a discordant avatar, both parties understand the nature of the relationship and the possibility that there may be some restriction of autonomy as a result of the choice of each party's avatar. In accepting a virtual telemedicine provider–patient relationship, both sides are acceding to these conditions.

It has also been argued (successfully) that lying within the therapeutic relationship undermines faith in the entire medical system destroying trust, preventing the seeking of needed care and thus causing harm (violating non-malfeasance as well as autonomy). Importantly, lying constitutes a betrayal of trust. However, changing one's avatar to fit the situation is accepted virtual behaviour and as such does not constitute a betrayal of trust. Indeed, since changing one's avatar is the norm and does not represent a betrayal of trust, we would argue that it does not constitute lying at all. In a previous paper we argue that the avatar is conditionally part of the self if imbued with intentionality by the user.12 In fact, this self is fluid and not immutable and can change from one online situation to another. There is an expectation (or at least acceptance) that individuals in virtual worlds may not appear as they do in the non-virtual world. The paradigm for personal identity in virtual worlds has changed. You can be one ‘person’ Monday and another ‘person’ Thursday. Your avatar is fluid and optimised to the situation in which you find yourself. For example, your avatar in Fallout 3, a post-apocalyptic role-playing computer game, is likely to be different than your avatar on an online dating site. As the function changes, the avatar changes. Changing one's avatar in Fallout 3 is no different than changing one's avatar to maximise therapeutic benefit. Both changes fit the ‘self’ to maximise outcome and benefit. One can argue that what is accepted behaviour is not necessarily ethical (witness Nazi Germany). However we have demonstrated that this change in behaviour does not inherently adversely impact on others. This situation (fluid virtual identity) can be abused by unscrupulous individuals, as can any situation. It is also important to reinforce that regardless of the avatar, the provider–patient relationship must still meet the other requirements of medical ethics including autonomy, beneficence, non-malfeasance and justice. A provider cannot choose a discordant avatar to gain some personal advantage in the relationship. It must be selected solely for the (perceived) benefit of the patient.

Finally, the patient must retain the right to ask about a therapists ‘real’ ‘race’ and gender. As soon as the patient demands an end to the situation it should happen. We would not argue the converse, however. The provider does not have an equivalent right to know the patient's gender or ‘race’. While ‘race’ and gender can be fundamental to providing appropriate services (recommending cancer screening, for example) the patient has the right under autonomy to divulge only the information she deems appropriate. And, the patient is bound to the same rules as the provider. The patient cannot use the fluidity of the avatar in an attempt to deceive the provider (into prescribing hormones, for example).

Social impact and appropriateness

Virtual worlds have the ability to be the ultimate ‘post racial’ ‘societies’. Yet, paradoxically, one may recognise the similarity of changing one's virtual race to the donning of blackface in the non-virtual world, a practice that is abhorrent. However, it is the history, social context, connotations and implications that make blackface impossible to condone in the non-virtual world. The very common use of ‘skin lightening’ products in Asia does not provoke the same basic visceral revulsion as does blackface. Nor would one be changing one's avatar in an attempt to demean or ridicule someone of another ‘race’. In fact, the motivation would be positive with the choice of the provider's avatar likely being left up to the patient.

Other issues and further research

Other issues raised by virtual telemedicine include: (1) the dissonance felt by the provider should the patient choose an avatar for the practitioner that is not the same gender/‘race’/age as the provider; (2) the question of trust and the impact of fluid personal identity on the effectiveness of the virtual provider–patient interaction; (3) the use of an avatar in purposeful deception; (4) the impact of avatar choice on the patient–provider relationship in the non-virtual world should a non-virtual world meeting occur or in the virtual world if a meeting occurs in another forum with the patient and provider represented by different avatars (and only one party recognises the other, for example); (5) the disorientation and confusion that might result when the patient realises the gender and ‘race’ of the therapist is not as he thought. Additionally, there are new spheres of autonomy provided for in virtual patient–provider relationships that should be explored including: (1) the patient's right to choose the avatar of the provider, (2) the patient's right to limit the information about personal identity (gender, ‘race’, age, etc.) shared, (3) the patient's right to avoid physical contact in the broad sense of actual personal presence, and (4) what happens when the patient meets the therapist in another forum. The relationship of game theory to avatar choice is also an avenue of exploration. Finally, the definition of phenotype and genotype in a virtual environment is a question ripe for discussion.

In conclusion, we have demonstrated that personal identity may be fluid in ‘virtual telemedicine’ and that there is no inherent ethical proscription against choosing a physically discordant avatar to maximise therapeutic benefit. The wisdom and advisability of changing the ‘race’ or gender of one's avatar within a therapeutic relationship can certainly be questioned and future experience will dictate whether or not this becomes an acceptable practice.



  • Competing interests Neither of the authors have any conflicts of interest or stand to gain financially by the publication of this paper.

  • Provenance and peer review Not commissioned; externally peer reviewed.