Toward a Theory of Critical Computing:

The Case of Social Identity Representation in Digital Media Applications

1. Introduction

"Why should the computer scientist read African-American poets? What does information science have to do with race-critical or feminist methods and metaphysics? The collective wisdom in those domains is one of the richest places from which to understand these core problems in information systems design: how to preserve the integrity of information without a priori standardization and its often attendant violence. In turn, if those lessons can be taken seriously within the emerging cyberworld, there may yet be a chance to strengthen its democratic ethical aspects." [1]

In everyday life our identities constantly shift. Identities are formed through enactment: developing innovative algorithmic solutions makes one a Computer Scientist. Speaking to a confidant makes one a Friend, while speaking to a co-worker makes one a Colleague. Identities are formed through social infrastructure: checking a certain box on an application renders one an Asian/Pacific-Islander, checking another makes one a Female. Identities are formed through imagination: the Los Angeles Times newspaper has described immigrants metaphorically as "a sea of brown faces," [2] and in 1952 Ralph Ellison conceived of the racial condition of the African-American metaphorically as that of the Invisible Man. [3] And now, identities are formed through computation: in video games and social networking systems users navigate through virtual environments as avatars, and users are reduced to information profiles as their preferences are recorded, unnoticed, by Google or Amazon.

These identities become convoluted in practice. Wrong words can change a Friend to an Adversary. In South Africa, people of Chinese descent have been reclassified as black so that that they receive economic benefits of post-Apartheid reforms. [4] The United States Constitution imaginatively counted enslaved people as three-fifths of a Person. Popular current games such as Elder Scrolls IV: Oblivion and World of Warcraft computationally implement and amplify many disempowering social identity constructions; in Oblivion females of some races are more intelligent than their male counterparts, and individuals of the ostensibly French "race" (Bretons) are twenty points more intelligent than their ostensibly Norwegian (Nords) counterparts, regardless of gender. [5]

Everyday computing activities require users to construct computational identities. Computational identities are not mediated only by social interaction, but by the particular implementations used to instantiate them. This means that a great deal is at stake when these technologies are developed. If they are developed without mindfulness of the limitations of folk or everyday models of identity, then prejudice, stereotyping, static classifications, and inequitable access to resources will be perpetuated in computational applications and environments. Yet, how many computer scientists developing social networking, e-commerce, gaming, or educational applications consider the latest theories of social classification from sociology or theories of conceptual stereotyping from cognitive science? The quotation by sociologists Geoffrey Bowker and Susan Leigh Star beginning this section suggests that, looking outside of the field of computer science, there exists rich information and identity construction strategies developed by individuals who have had to navigate the shifts, convolutions, and problems of social identity. The reliance of computer scientists on intuitive understandings of identity, rather than nuanced theories that view identity as enacted, infrastructural, and imaginative, has resulted in software that at best ignores opportunities for empowerment, and at worst results in perpetuating longstanding social ills of discrimination and disenfranchisement. We can do better.

Improving upon this situation is not a technical issue alone. Using the case of computational identity representation as an in-depth motivating example, because it is laden with social and technical complexities, as elaborated above, this paper is meant to demonstrate a developing type of practice called critical computing argued for in my article "Toward a Theory of Phantasmal Media." [6] The next subsection describes the meaning of the critical computing concept and explains how the concerns about social identity as mediated by computing technology raised above are an excellent case with which to exemplify the idea. Section 2 of this paper describes a theoretical framework for analysis and design of computational identity systems. Section 3 uses this theoretical framework as a basis for analyzing computational identity in several gaming applications. Section 4 describes ongoing research under the rubric of the Advanced Identity Representation (AIR) project that I have pursued. Section 5 provides concluding remarks.

1.1 Critical Computing

Critical computing refers to the potential for using algorithmic processing and data structuring as a basis for expressing commentary about, and making impactful change upon, the real world -- especially to engage disempowering hegemonic norms and socio-technical conditions. It relates, in part, to critical technical practices, a concept introduced by Philip Agre that represents a union between technical research and development and critical theory. [7] His work has great alignment with other philosophical and critical commentaries upon AI [8], however with the difference that, rather than abandoning the field, Agre seeks to reform it. His approach also has a far-reaching impact in technological fields beyond AI -- for example, Phoebe Sengers's research applies Agre's ideas to highly original approaches to Human-Computer Interaction (HCI) designs using computing within everyday experiences and environments. [9] In other related work, Brian K. Smith led the "Explanation Architecture" research group at the MIT Media Lab, an endeavor that used the term "critical computing" earlier, though in a different way as Smith's focus was on designing "tools that help people engage in critical reflection around their attitudes, beliefs, and behaviors." [10] Recently, theorists and artists including Arthur and Marilouise Kroker and many others have coalesced into an international community pursuing cultural readings of, and interventions through, computational technologies under the banner of Critical Digital Studies. [11] Critical awareness in my conception in this paper is directed both externally toward the world and internally toward technology itself, rather than focusing only on self-critical approaches to computing and information sciences.

The critical computing concept helps technologists to move beyond issues of utilitarian and productivity-oriented applications and to think about issues such as social identity, power relationships, and political configurations. In areas such as user-interface design, computer-supported cooperative work, and social networking, issues of users' knowledge and experience bases, social groupings, and facility for social interaction, are all intrinsic to the technology. Though occasionally relying upon sociological or anthropological methods, rich areas of social critique and insight from marginalized peoples and humanistic discourses are often overlooked in computer science.

One initial area in which a critical computing approach would enable significant advancement over the current state-of-the-art is that of social identity computing technologies, in light of the issues raised in the introductory paragraphs above. Social ills such as racial discrimination, gender-based inequity, and cultural stereotyping exist in computational media in ways that both perpetuate old media models and construct new problems of social disempowerment. In the face of this challenge, this paper asserts that a powerful new possibility exists to develop transformative and ethically and critically engaged models of social identity (including race, gender, ability, profession, relationships, style, etc.) using techniques from computer science and cognitive science. Crucially, however, self-representation in gaming and social computing technologies involves not only computational and cognitive issues, even as bolstered by humanities-based concerns. It also involves art, aesthetics, and style. The aesthetics of transformative and empowering social identity is that of polymorphic poetics -- an expressive set of possibilities centered upon the contingency and imaginative nature of even the most entrenched social identities, such as those of gender and race.

1.2 The Advanced Identity Representation (AIR) Project

As an example of a critical computing project with concerns for polymorphic poetics at its core, the AIR Project is a five-year National Science Foundation supported endeavor that develops new models and technology to enable empowering user self-representations for deployment in social networking, computer supported cooperative work, computational media arts and games, recommendation/preference systems, and educational technologies. These user-representations may be dynamic, integrated, sensitive to social context, empowering for diverse and underrepresented groups, and ethically engaged. Broad applicability of the models and technologies is enabled by (1) methods that invoke the shared cognitive mechanisms for categorization that undergird human construal of socio-cultural identity classifications, and (2) computer science approaches to multimedia semantics that address algorithmic and data-structural reciprocities between the two dominant models of computational self-representation: avatars (mediating proxies for action in computational environments) and user profiles (informational surrogates in computational applications).

The new AIR Project models also apply to artificial intelligence (AI) identity representations such as software agents and characters. The following section provides a theoretical framework that has been developed to inform the AIR Project specifically and that more generally exemplifies one modest and developing critical computing approach that is grounded in several influential cognitive science, computer science, and humanities/social-sciences approaches and results.

2. A Critical Computing Theoretical Framework for Identity

Accounting for computational identity representations requires new approaches. In recent years there has been an explosion of research converging from a range of disciplines addressing the problem of identity in virtual worlds, games, social networking, and related technologies, as I have outlined in my article "Computational and Cognitive Infrastructures of Stigma." [12] All of this work points to the need for infrastructure for identity representation suited to the dynamic contingency of real life experience. Addressing this need, the below presents the AIR theoretical framework. This framework consists of two pillars: one grounded in computing and the other in cognitive science. A summary of insights arising from this body of work and the AIR Project aims to address them is presented in Figure 1.

Figure 1: AIR Project Research Aims

2.1 Shared Technical Underpinnings of Computational Identity Applications

Various computational identity applications, such as social networking sites, avatars/characters in computational environments, and online accounts for services such as e-commerce, are implemented using a limited and often overlapping set of components. There are two important motivations for describing these components: (1) to identify an appropriate level of abstraction for analyzing the technical side of computational representations comparatively across different types of applications, and (2) to identify components that can be analyzed both in terms of how they appear visually and how they are implemented algorithmically and data-structurally. For example, numerical statistics appear in both accounts for online auction sellers and characters in computer role-playing games -- both are clearly represented as interface components (often framed in a subwindow on the screen) and in terms of back-end data. Figure 2 describes the six components that comprise the majority of widely-used computational identity technologies [13].

Figure 2: Computational elements of a Sociodata Ecology:
Shared Technical Underpinnings of Computational Identity Applications

It is important to point out that these technical underpinnings exist in what could be called a sociodata ecology, in which technical infrastructure, specific data-structures and algorithms, and specific code is looked at as it relates to issues such as embodied experiences, subjective interpretation, power relationships, and cultural values. The goal is not to reduce socio-cultural issues to technical ones, but rather to be able to clearly describe how specific technical features bear upon human experience. I highlight computational components because the many brilliant insights of cultural theorists and critics often fail to address the specificities of how systems are engineered; in short, I am attempting to fill a gap. This is important for those of us who want to build, and empower diverse groups to build, new technologies customized to their epistemologies and values. This type of research compliments study of the many practices that users engage in that nonetheless treat computing applications as landscapes upon which to operate rather than as technologies that users can produce and alter. Manipulating these computational aspects, especially the relationship between formal semantic annotation and modular graphical models, can enable dynamic and context-sensitive models of social identity in computational environments. The following uses examples from computer gaming, online communication, and social networking to explicate this model. The examples are presented to simultaneously reveal not only the influences and affordances of their technical bases, but also to highlight the complicated ways that they interact with social practices in a sociodata ecology.

Avatars, in their simplest forms, are often nothing more than graphical images, such as jpeg files, that are presented next to user names. For example, many online forums allow users to select images to display next to their postings, such as the comic book images in Figure 3.

Figure 3: Static image avatars on a forum, here the signature images also comprise static image avatars [14]

These avatars, though rather simplistic technically, afford a range of self-representations. The images in the context of the forum above integrate variations of common idioms, normative data such as gender and location, remediated pop-cultural characters, and sophisticated montage. Furthermore, they are deployed in communities of exchange -- individuals create such images for one another, and communal evaluation -- there is a discussion thread in which people rank the signature image of the previous poster on a scale of one to ten. There are technical limitations regarding file size, resolution, image size, image placement, and more, yet creative practices abound in ways that slightly more complex technologies such as modular avatar creation systems (e.g., may not allow.

Most avatar systems feature modular graphical models rendered with variable skins that offer the illusion of acting as transparent proxies for our real selves to act within games and virtual worlds.

Figure 4: A user-customized avatar in Second Life

However, users are limited to the set of multimodal signs and communication mechanisms (e.g. chat windows, emote commands) developed by the system's designer. Our natural discourse production and performances are then filtered through a constrained set of possibilities. This is a fascinating duality -- one achieves the ability to create representations that may be far removed graphically from real worlds identities, yet one loses everyday communicative abilities. Furthermore, as we shall see later, users cannot avoid mapping real world social identity concepts onto even the most abstract avatar.

The affordances provided by our characters to complete diegetic (storyworld) actions are often implemented via sets of statistical/numerical values. For example, Figure 5 depicts the initial attribute values for characters in the computer game Elder Scrolls IV: Oblivion.

Figure 5: Initial character attribute values in Elder Scrolls IV: Oblivion; columns indicate race and gender while rows indicate attributes [15]

Note that several interesting initial identity characteristics are hard-coded. Female Orcs and Argonians are ten points more intelligent than their male counterparts, and the human Imperials are twenty points more intelligent than human Redguards of either gender. Hence, choice of race and gender results in essential characteristics reified within the gameworld (a phenomenon usually reserved for human/non-human distinctions).

Figure 6: A gaming website's chart of the racial attributes of the "Imperial" race [16]

Furthermore, higher attribute values, along with skill bonuses, have the effect enhancing performance of particular actions, e.g. bonuses to mercantile and speechcraft skills granted to the Imperial race result in commerce-related advantages. The Imperial race also features provocatively titled special abilities such as "Star of the West" and "Voice of the Emperor." [17]

A hallmark of an older gaming form of role-playing, the text-only Multi-User-Dungeon (MUD), is the player's ability to describe her or his player character in text files. These files are considered "flat" because they are unstructured and do not influence game mechanics; they must be enacted by the player as projected identities. For example, the MUD Armageddon features a race called "Muls," described as "sterile crossbreeds of dwarves and men, bred almost solely by Templar slavers and nobles for combat in the Arenas of Allanak as well as those of Tuluk." [18]

Players provide their own descriptions such as:

the heavy, beady-eyed female half-giant
This enormous woman is extremely pudgy, rolls hanging in her skin. She has obese, round legs that jiggle slightly with any movement. Her enormous stomach is rounded and protrudes from the rest of her body by a noticeable amount. She has two thick, corpulent arms that hang almost lifelessly at her sides at most times, and large hands adorned with chubby fingers. Above her adipose neck is a huge double chin, the false one nearly the size of the real one. She has enormous lips, a small amount of spit always gathering at the corners. Around her stubby nose are her fleshy cheeks, slightly rosy in color against her lightly tanned skin. As a result of her build, she has extremely deep eye-sockets, in which can be seen two beady brown eyes. She has greasy black hair, which falls freely down the back of her neck. [19]

The main feature of such flat files is that, while they allow players complete textual freedom within length constraints, their effects are not regulated algorithmically. Hence, MUD informational sites often provide written descriptions of how to perform identity categories such as race or sexual orientation; for example, Muls are provided the following personality guidance for appropriate role-playing: "being sterile, and thus outside of the typical chain of reproduction, muls often suffer from a sense of meaninglessness." [20] This suggests that ancillary media should be investigated to see how rules and social practices emerge in systematized ways in conjunction with the underlying data representations -- humans mitigate identities when presenting themselves and understanding each other, whether through code or communal practices.

In conjunction with numerical statistics, other information about virtual identity is formalized and represented using abstract data structures identified with the character/avatar/profile. For example, the modular graphical elements are linked to strings that indicate which files should be displayed currently. Such representations are often tokens to be evaluated during execution to trigger particular effects, for example the dynamic depiction of blood color as it varies by creature type or race.

Figure 7: Data types for player characters in Neverwinter Nights [21]

Finally, some aspects of character performance of actions can occur algorithmically, without input by the user, or only with high-level guidance by the user. For example, many games include physics-based models that determine effects in animation, such as collision or fluid motion, or even behavioral models that determine and/or constrain character actions. For example, a simple form of behavior was implemented for player characters in Sega's Sonic the Hedgehog series. When the character remained idle to long the sprite would "grow impatient" and reveal a series of images to construct an animation including foot-tapping or growing tired.

Figure 8: A sprite of the character "Tails" growing tired in response to user inaction [22]

These six components that commonly form the basis for users' character construction have been used in nascently expressive ways to implement characters with identities that transform in response to user actions, context, or just the passage of time. Such examples are revealing in the technical facility for transformative identity they implement, the conventional metaphors they enact, and the constrained set of computational approaches upon which they are based. The AIR project suggests that we can also use computational models to represent critical perspectives on complex models of social identities themselves. Though each of the six components used to implement computational identities are not present in every application, an observation underlying the AIR Project is that parallels in these components, particularly at the level of formal annotation, can allow diverse applications to be integrated. Furthermore, within individual applications we can favor dynamic data-structures that are informed by conceptual models of cognitive science. Conceptual models that define social relationships, preferences, metaphors, and values could be used to add salience to computational experiences. The following section describes how this project utilizes cognitive linguistics insights as a scientific basis for how computational modeling can enable dynamic, contingent, and multivalent models of social identity in computational environments.

2.2 Cognitively Grounded Model of Computational Identity

The notion of an identity-related sociodata ecology, as illustrated by the AIR cognitively grounded model of computational identity, is summarized in Figure 9. [23] Since cognitive science now addresses the embodied and situated nature of cognition, such a model need not be reductive or technologically deterministic. Human imaginative cognitive processes for creating categories are fundamental, and cognitive science results are more easily reconciled with cultural theoretic insights because recent cognitive science engages issues of artifact use, embodiment, social situations, and communication. My AIR model emphasizes how technical artifacts (ranging from specific code deployments in role-playing games to broad infrastructures for classification like national census data structures) and performed/constructed social identities are based on imaginative cognition -- an idea commensurate with many constructive approaches to identity. As well as imaginative cognition process (specified below), we look at how the human ability to map concepts onto artifacts and interactions enables diverse computational identity phenomena.

Figure 9: A Sociodata Ecology:
The AIR Cognitively Grounded View of Computational Identity

The AIR Project approach begins with the basic cognitive building blocks of identity (discussed in Section 2.2.1), upon which all social identity categories are built. Cognitive scientists have proposed that human conceptual categories form "idealized cognitive models" (ICMs) upon which categories of objects in the world are built. [24] These ICMs are akin to what are known as ontologies in database and artificial intelligence (AI) research. Yet, most user categorization is not done using AI knowledge representation techniques. Social networking sites group users into categories called "friends," while games may group users into categories called elves or half-orcs. Technical infrastructures implement and reify (often incorrect) stigmatizing identity classification models [25]; indeed, some games feature data-structures instantiated with values in which some races/genders are less intelligent than others. [26] Cognitive science theory is presented in Section 2.2.1 below to provide more robust models. These models can explain how users project their identities into their computational surrogates/proxies. [27] This model is new in that it integrates computational, cognitive, and sociocultural analyses, and in that it applies recent cognitive science ideas to areas that they have not been used to explicate previously. Toward this end, a brief account of these underlying ideas follows.

2.2.1 Cognitive Categorization

The AIR approach to identity is influenced by the prototype theory of the psychologist Eleanor Rosch, and especially by work in categorization by the cognitive scientist George Lakoff. [28] Lakoff's work in this area over two decades ago is well known and influential, yet it is a thread that has been seldom applied to issues of social identity (an exciting exception being the work of the linguist Otto Santa Ana on conceptual metaphoric based bias in Brown Tide Rising [29]). Furthermore, this approach has not been applied elsewhere to computational identity.

Traditional theories or "folk" views of categorization define categories by the common objective properties of their members. In contrast to folk theories of categorization, Lakoff asserts that categorization is a matter of both human experience and imagination.[30] Lakoff cites research from the fields of psychology, computer science, neuroscience, anthropology, and other disciplines to reveal a convergence of evidence disputing the traditional theory. Important observations from this convergence of evidence include arguments that:

There is not a single, well defined collection of common properties shared by all objects in some categories

Important for the purposes here, Lakoff describes a metonymy/metaphor-based account of how imaginative extensions of "prototype effects" result in several phenomena of social identity categorization that have proven useful for the AIR Project:

Since the AIR Project technology involves techniques to formalize and implement ICMs as computational data-structures, identity phenomena become amenable to algorithmic manipulation and experimentation. The AIR Project entails computationally modeling phenomena that define normative expectations and stigma (stereotypes, ideals, salient examples, etc.) within computational identity applications and enabling more nuanced and expressive representations for social analysis and user empowerment.

2.2.2 Sociology of Classification Infrastructures

The AIR Project is influenced by accounts of classification from sociology. In Sorting Things Out, Geoffrey Bowker and Susan Leigh Star argue that classification systems are necessary for information exchange and communication. [32] Bowker and Star call attention to the concepts of membership and naturalization. Membership is the experience of encountering objects and interactions native to particular communities and increasingly engaging in naturalized relationships with them. Naturalization refers to deepening familiarity with use and enactment involving such objects and interactions. The problem with enforced naturalization is that it always creates problems of marginalization. "Marginal persons" are those who either exist outside of communities or are less "prototypical" members of communities. Marginalization can occur through exclusion or through multiple memberships in communities where an individual must switch frequently between interaction and object use protocols within each community, often with varying degrees of success. Typically, when discussing marginalization it refers to exclusion or difference from normative behaviors (stigma) and/or dominant, privileged, and/or hegemonic communities. The concept of category markedness indicates that unlike normative categories, marginal categories are linguistically demarcated. Identity torque describes a situation in which self-classification of individuals differs from how broader society classifies them.

2.2.3 Sociology of Stigma

An important early theory of identity construction can be found in Erving Goffman's work. [33] In Stigma, Goffman grossly describes three types of stigma. These are differences of (1) the physical body, (2) individual character, or (3) "tribal" classes of "race, nation, and religion." Goffman describes each of these categories as deviance from "those who do not depart negatively from the particular expectations at issue" that he calls the "normals." [34] The postulatation here is that the experience of stigma largely rests in the human cognitive ability to map characteristics of the second type of stigma, that of character, moral value, will, belief, and passion, onto physical characteristics and attendant categorization into socially recognized races, nationalities, and other so-called tribal classes (which may be described using theories of conceptual metaphor and blending). Also, these mappings are reinforced and reified in social classification infrastructures, including computational infrastructures.

2.3 Blended and Double-Scope Identities: Elaborating the Virtual and the Real Identity Nexus

Learning scientist James Gee's concepts of the real, virtual, and projective identities provide a useful starting point for thinking about how embodied identity experiences and values in the real world intersect with the affordances and semiotic values of computational representations. [35] Gee's descriptions of player identities as mediated through characters in games in this triumvirate of personae, which, while eliding the complex mappings we invoke to cognitively construct virtual senses of self, highlight that discussing only "real life" or only technology is not enough. For Gee, player-representations as projective identities, the most novel construct in his triad, manifest the ways that real player values are reconciled with values understood as being associated with avatars. I propose going further and considering how avatar values are expressed and mediated through the in-game affordances such as graphical models and skins, along with the cognitive processes that accompany deploying them.

Gee's constructs raise significant unanswered questions. Is the real identity an individual psychological construct, a social construct, an objective and material reality, or some combination of those? We can ask the same questions of the virtual identity. For example, Gee discussed the virtual identity Bead-Bead, a half-elf character in the game Arcanum. [36] At first glance, it is easy to say that this is a fictitious character in a fictionalized gameworld. However, this does not go far enough because the gameworld does not exist separately from an individual's cognitive understanding of that world. The individual's understanding is mediated by social context, historical experience, the artifact implementing the world, and more. One person's understanding of "half-elf" may recall a part-human (default everyday category) and part otherworldly aesthete being. Another person might see a half-elf as cynically referring to an outmoded concept of miscegenation. Another person might see a half-elf in purely game-mechanical terms: it entails possessing a 30% resistance to sleep, charm spells, and 60' of infravision (seeing in the dark).

However, game studies researcher Zach Waggoner discusses Gee's constructs as describing the virtual identity as "the avatar that exists in the fictionalized world." [37] So, the actual implementation for half-elves in a game might not support all of those interpretations equally, in fact it may exclude some. The virtual identity, then, is a merger of a computational representation with a meaningful interpretation of that representation. In that sense, both the virtual and the real identity are comprised cognitively (which for us includes context and embodiment) and materially (through code and hardware). It is impossible to look at virtual identity as a purely conceptual construct that is not somehow embodied. The concept of an identity sociodata ecology is meant to emphasize computational infrastructure at many levels -- material, conceptual, social -- and to aid in precise theorization of concepts like real and virtual identities.

The AIR Project approach emphasizes projective (here "projected") identity. Jill Walker Rettberg has discussed this type of identity as ontological fusion. [38] Using terminology from the cognitive science theory above, this can be seen as metaphorically mapping ICMs (mental spaces) that humans have of themselves onto characters, or to use terminology from Gillles Fauconnier and Mark Turner's conceptual blending theory [39], as selectively projecting aspects from conceptualizations of both a real identity and a virtual identity into a blended identity. Since this entails integrating concepts from quite different, even clashing, conceptual frames, this would be called a double-scope identity. [40] Notice that blending a real and virtual self means something quite dynamic. The projected, or blended, self at one moment may be different from the next. One person's projected identity will be different from another's.

Conceptual blending theory is not currently a predictive theory that can forecast exactly how humans combine ideas. [41] Rather, it describes constraints on the process of combining concepts -- what makes one way of blending concepts more optimal than others -- and provides a systematic way to talk about integrating concepts. One person might project a relation such as (written logically) "morally-righteous (self)" onto the avatar (let's say a vampire character) and another might not. This would result in the former person finding a clash of values in the projective identity, while the latter does not find the clash to be the case. The blended identity is quite fluid. In practice, we always uses systems at the level of the blended identity -- a more precise account would describe precisely how this blend is constructed, what its constituents are, and how it changes dynamically.

Because of the focus on stigma and prejudice here, it is important to distinguish identity theory from identity politics (narrowly construed). In the end all social identities are political because we cannot escape them and they affect our everyday lives and potentials. [42] What should be emphasized is that studying perspectives on identity is not a matter of issues such as determining which group is most marginalized, which has been most oppressed, etc. Rather, everyone has an identity (or identities), everyone constructs categories, and our identities are already integrated with digital media through communication, commercial and financial databases (at the very least), and our identities' impacts on our environments, as discussed by Eric Kabisch. [43] It is possible, however, to look at how institutions, infrastructures, technologies, and more relate to issues of social change, power, justice, equity, etc., and assess what relationships exist.

The novel theoretical framework presented in Section 2, synthesized in the concept of a sociodata ecology applied to the case of identity in the AIR Project, ranging from accounts of computational identity representations to a theory of stigma from seminal work in sociology, assists in providing insight into user identity self-representations across media as diverse as forum profiles, virtual worlds, and computer games.

3. Critical Computing: The Case of Computational Identity Representation

Current user representations in computational media are inadequate for capturing complex phenomena and subjective experience of social identity. The case of racial representations in computer gaming can illustrate this phenomenon. In light of the theory presented above, we can continue examining popular current games such as Elder Scrolls IV: Oblivion and World of Warcraft, which implement existing social identity constructions. For example, the ostensibly African-featured "Redguard" characters in Oblivion are often described in terms of the essentialist stereotype of the black athlete: they are "the most naturally talented warriors in Tamriel? born to battle, though their pride and fierce independence of spirit makes them more suitable as scouts or skirmishers?. In addition to their cultural affinities for many weapon and armor styles, Redguards are also physically blessed with hardy constitutions and quickness of foot." [44] The bull-like Tauren race in World of Warcraft live in teepee-like structures replete with totem-poles where, "despite their enormous size and brute strength, the remarkably peaceful Tauren cultivate a quiet, tribal society," thus embodying an exploitive cultural stereotype of the innocent savage and exemplifying Goffman's third type of stigma. [45] Since Tauren are a part of the Horde category in the game, comprised primarily of monsters with negative moral implications, as discussed in Gee's What Video Games Have To Teach Us About Learning and Literacy [46], Tauren represent a good example of mapping Goffman's second type of stigma onto his third within a computational system.

Figure 10: Racial stereotypes in popular computer games

Such games, we have seen, hardcode racial stereotypes into their infrastructures and fail to ask even obvious questions, such as: What is the effect of racial profiling upon exploration in the storyworld? How can race be represented as more than a set of numerical variables? And how can we examine the subjective experience of social oppression or privilege as opposed to cosmetic changes to avatar appearance?

Computational identity representations have currently been used in only modest ways to implement characters with identities that transform in response to character actions, context, or just the passage of time. The most common type of transformation implemented in popular computer gaming represents a moral binary in which the "Life is a Journey" basic metaphor is mapped onto a quest involving the corruptibility of the protagonist. [47] Occasionally, this has been used to subtle and striking effect; for example, in the game Shadow of the Colossus, the player is forced into a projected identity in which the standard "save the female victim" narrative (in this case she is a glowing phantasmal white being that has already seemingly died) requires the player to slay colossal, monstrous, but innocent, beings. [48] Despite its innovative technique of increasing affective expressivity by constraining player agency at the level of the virtual identity -- a fateful melancholy arises as the player's agency is restricted to the local level of spatial exploration and battle while global moral decision-making is not possible -- the virtual identity's transformation is relatively straightforward. The protagonist's face and clothing become streaked and stained in a clear parable for moral degradation.

Figure 11: Character transformation over time in Shadow of the Colossus [49]

The Microsoft Xbox game Fable is notorious for a similar, if less elegantly executed, effect. Rather than pitting the projected and virtual identities at odds with one another, and thereby manipulating the player's sense of agency as an expressive resource, Fable transparently reflects the game's moral order through the transformation of the virtual identity. [50] Consumption of alcohol or other "immoral" actions result in conventional demonic attributes such as red eyes and horns, while a "noble" path results in the virtual identity representation's transformation to an august being who is showered in praise from bystanders and beloved by children non-player characters.

Figure 12: "Noble" and "Corrupt" eventualities for the player character in Fable [51]

The independent art game Passage utilizes an abstracted retro-graphics style (it was developed for a contest, under the constraint that all games be displayed in 256 pixels or less) in order to effect identity transformation at the graphical level to express the conventional "Life is a Journey" metaphor mentioned earlier. [52] In this game, the player's virtual identity representation transitions from being statically located on the left of the narrow screen to the right while exploring an abstracted and pixilated space through a blond, male virtual identity. He accumulates meaningless treasures, and perhaps a life partner, and eventually goes bald, grows old, loses his partner, and dies.

Figure 13: Three screenshots taken at different stages of life in Passage

Such examples are revealing in the technical facility for transformative identity they suggest, the conventional metaphors they enact, and the constrained set of computational approaches upon which they are based. These examples display the technical possibility for evocative transformation within gaming. They are based on conventional metaphors such as "Time as Space" [53] or on moral identity based again in binary opposition and discrete sets of attributes arranged hierarchically, and are implemented based upon the characteristics described above (modular graphical systems, numerically and formally encoding subjective qualities such as "goodness" or "evil"), and algorithmic representations of metaphorical phenomena such as moral descent. We can also use computational or mathematical models to represent critical perspectives on complex models of social identity themselves. Mathematical modeling of identity need not reproduce disempowering identity representations as in Oblivion, nor need they represent simplistic ethical choices. Questioning the limitations of current identity representations in games does not require the production of didactic games that eschew the fantasy inherent in the medium; the call here is not for diversity training games. Indeed, Shadow of the Colossus and Passage in particular represent poignant affective possibilities for character transformation. Yet, toward the aim of critical reflection on identity, we can go further.

4. Examples of Critical Computing Systems Addressing Computational Identity

We are modeling attitudes, objects, and various aspects of a relation between them; to do this job, we are using, among a large group of things and relations, various of those things and relations to stand for the objects, attitudes, and relations we wish to model.

Ordinary, informal, nonrigorous language overcomes all these problems, however, with a bravura, panache and elegance that leave the formal logician panting and applauding. [54]

The epigraphs above hint at the relative powers of both mathematical logic and imaginative language for capturing subjective, nuanced, and contextually contingent concepts. My own work comprises an interdisciplinary computational approach that can attempt to leverage human subjective expression with the polymorphic possibilities of computational data-structuring and algorithmic manipulation. One outcome of my research has been an expressive form called polymorphic poetry or polypoems, which combine prose poetry, dynamically reconfigurable data-structures, and forms culturally grounded in African diasporic oral traditions of signification, with the use of algebraic techniques to construct imaginative metaphors on the fly. [55]

As an example of applying such a system, "The Girl with Skin of Haints and Seraphs" is a polypoem implemented in the GRIOT system for developing computational narratives. [56] It works by establishing a set of stereotyped theme domains such as skin, angels, demons, old Europe, and old Africa, composed of sets of axioms. During the execution, each time the user enters a term it is scanned for relevance to the domains and a response is produced as output to the screen. The system constructs conceptual spaces, using the algebraic semiotic framework, and blends these to construct metaphors using a conceptual blending algorithm. These are then combined with narrative templates, in the case of "The Girl with Skin of Haints and Seraphs," these are based on a model from socio-linguistics research, a formalization of William Labov's structure of narratives of personal experience. This narrative template is integrated with metaphors generated on-the-fly.

This polypoem implemented with the GRIOT system actualizes the winding together of the separate cords of computational and mathematical modeling, social identity, and imaginative metaphor in fiction in a cultural artifact. The output could be said to construct a precursor to a polymorphic persona that is composed of text-based stereotyped binary opposition that can be thrown into new conjunctions each time -- oppositions that are thematically related but ultimately incompatible and untenable: at one moment the protagonist is "raising imperialist, cherub children" and at the next she "is [a] melaninated and impoverished-elder." Example output can be found in my article "Algebra of Identity." [57]

The GRIOT system has also been used for composing graphical images to create a type of visual poetry that also addressed identity. This technology enables the GRIOT system to be used to construct graphical and textual character representations for games, social networking sites, and educational software. In this manner, we have also experimented with changing iconic human representations based upon context in computational environments, as illustrated in an image from a system called Generative Visual Renku in Figure 14. [58]

Figure 14: A looping walking (or riding) animated human figure in each square shaped tile is represented differently depending upon its context based upon artifacts appropriate to the location

Kenny Chow and I have defined a set of combinatorial, recursive, and procedural rules to allow such transformations to take place. This is, however, just an early step toward the types of applications envisioned in the AIR Project.

4.1. Pilot Work: AIR Project Case Study Systems

The following are examples of systems that demonstrate polymorphic poetics via mutable character or profile representations. The systems have been described elsewhere; however, the following descriptions elaborate the particular ways in which their development engages with and was driven by the theoretical framework outlined above in Section 2.

Figure 15: A screenshot of Loss, Undersea
Figure 16: Screenshots of Chamelonia: Shadow Play

Figure 17: Three screenshots from Avatar Breeder
Figure 18: A screenshot of the IdentityShare interface
Figure 19: A screenshot of DefineMe -- Chimera (left), illustration of potential avatars (right)

Loss, Undersea is an AI-based interactive narrative in which an avatar is generated and transformed dynamically based upon the emotional tone associated with users' actions. [59] Loss, Undersea models naturalization by generating new narrative discourse and avatar configurations according to the user's action and the current location in the narrative (e.g., waking up in bed, in the shower, at work, etc.).

Chameleonia: Shadow Play is a prototype critical identity politics game in which an avatar and its shadow (performed and socially-constructed selves) dynamically transform, along with the cinematic presentation of the scene, based on player-selected gestures and the current location. Chameleonia implements naturalization phenomena as well. One of the major ways in which humans naturalize within communities is by displaying contextually appropriate gestures. It also implements the concept of torqued identity. The player character represents the user's view of herself, while the shadow represents the socially constructed self.

Avatar Breeder is a generative satirical artwork constructed by Daniel Upton and Jisun An in a graduate level course I taught. It serves to undermine the essentialist nature of normative categories of identity encountered on bureaucratic forms. Avatar Breeder allows users to breed avatars together to create new ethnic categories, then labeled by users. A user is provided with an initial pool of avatars labeled with ethnicities from a Georgia Tech form. The user selects two parents, which genetically combine to create four potential children. Subsequent generations can be created by selecting a child and a random mate to "breed." Users can continue to genetically combine avatars, generating a family tree. The system implements metonymic categories including stereotyping and salient examples. As user-generated content grows, the normative categories are eventually marginalized.

IdentityShare, a social networking site for "non-friends," and Daniel Upton's MS thesis project in Digital Media, was also developed under the umbrella of the AIR project. [60] The system provides users with facilities to construct profiles, follow web-browsing trails of others, and comment upon other users, exploring both similar and different profiles. IdentityShare offers a dynamic means of self-representation based on open-ended categories and tags. Standard profile models that include normative categories such as name, age, gender, location, and race are bolstered by a customizable database of user-created fields, with design features to ensure database consistency. Users can select which categories are most important to them using checkboxes. The system also implements category centrality. [26] A user's profile, as a collection of categories that define a user, is no longer viewed as just a set of static characteristics that are true about this user, but rather as a set of characteristics in which some may be more definitional to the user's self-conception. A future implementation could offer a ranking system for each category, not only providing centrality, but centrality gradience since "members (or sub-categories) which are within the category boundaries may be more or less central." [61]

DefineMe: Chimera is a social networking application (like Facebook) in which users define metaphorical profiles and avatars for each other rather than solely specifying their own representations. DefineMe implements metonymic ICMs for categorization to allow users to communally co-construct other peoples' avatars. [62] The DefineMe database, the same as in IdentityShare, relies on tags to create additional descriptors for each member. For instance, one user could describe a friend as a "lion" because she "is" "strong" (the tag). Another user could add an additional tag, stating that she is a "lion" because she "tends to be" "carnivorous." These tags can comprise vertical parent-child links (e.g., a "lion" is an "animal") or horizontal implicit links (e.g., in another user's profile a "lion" is an "Ethiopian symbol," yet the system may still create a category linked by the concept ?lion'). The initial content domain consists of animal types (constructing chimeras) because they are potent and entrenched metaphors for human personality [63], however the model extends to more everyday social categories such as scenes or fashions. The system implements identity torque when the avatars differ from users' self-conceptions.

4.2 Polymorphic Poetics

The pilot work discussed above is intended to have transformative effects in part through expressive use of media, in other words, the domain of poetics. Now performing feminine displays of muscle-bound preening, now daughter of colonialist demonized separatists, now imbibing the nectar of negritude to spew lyrically, rhythmically onto the page, now checking the census box for multiple ethnic memberships -- our identities are in flux, constructed cognitively through our abilities to map from one concept to another as conceptual metaphors, and to blend concepts to form new ones. Accounting for ways in which the structure of computational media enables expression of dynamic content comprises a type of poetics. Recall Roman Jakobson's description of poetics within linguistics:

Poetics deals with problems of verbal structure, just as the analysis of painting is concerned with pictorial structure. Since linguistics is the global science of verbal structure, poetics may be regarded as an integral part of linguistics. [64]

Contemporary cognitive linguistics sees issues of concept generation, metaphor, and narrative as pre-verbal phenomena, hence cognitive poetics must account for meaning-making at large, including in computational media. The notion of polymorphic poetics holds at its center a concern for the dynamic nature of computational media representations as they are acted upon by algorithmic procedures and given computationally amenable form by abstract data structures. Social identity representations are a clear case through which to discuss polymorphic poetics.

With most current computational social identity technologies (such as computer games), our self-representations inherit none of the fluidity of our personae in our embodied, situated, and imaginative lives. Even in computer games that allow for the greatest degree of character customization and flux, we find that real world characteristics are reduced to a restricted range of representations, implementing player characters' attributes as numerical variable sets, modular computer graphics models, and related techniques.

In contrast to approaches that have inevitably served to duplicate the status-quo, in which identities are comprised of discrete sets of characteristics resulting in classifications (often in hierarchical binary oppositions, as criticized in my "Algebra of Identity" [65]), technologies imbued with the aesthetics of the polymorphic exploit the mutability of computational data-structures as a mechanism for critiquing our real life social identities. These are self-representations with the following characteristics:

  1. generativity (dynamically constructed and elaborated)
  2. semantics-based interaction (built upon formalized subjective representations of social knowledge)
  3. reconfigurable discourse structure (expressed differently based upon user interaction)
  4. strong socio-cognitive grounding (acknowledging the limitations of computational/mathematical representations to capture our embodied, situated, and technologically and socially distributed selves).

These characteristics emerged from previous research developing theory and technology for interactive and generative narratives and poetry based upon an interdisciplinary theoretical framework of computer science, cognitive science, and cross-cultural discourse models [66]. As seen in the pilot work and other projects I have worked on, outcomes based on this framework have included polymorphic poetry (texts that are generated to vary in semantic content, such as metaphorical exposition and affective tone, in response to user-input on each iteration) and polymorphic personae (graphical characters bolstered by semantic back-end data to enable them to visually transform depending on context and user actions). The notion of polymorphic poetics generalizes these results in order to articulate a prescriptive and analytical account of computational media aesthetics applicable to other domains, including the realm of social identity, as discussed here. However, this generalization does not arise simply, since clearly text-based works of poetry are quite different from avatars or social networking profiles, and future work must be done.

4.3 AIR Models and Future Work

Computational models of cognitive categorization and social classification drove the development of the systems above, however development of future systems based on this framework can be pursued in a more systematic way. It would even be desirable for diverse types of systems to be developed using a shared programming toolkit or application programming interface (API), a proposed outcome of the AIR Project. Extensions to the models to be developed as a part of the AIR Project consist of developing and refining formalizations and techniques to implement a small subset of cognitive and social identity phenomena in software, initially addressing torque, metonymic category models, marginalization, markedness, naturalization, and category gradience. These are summarized in Figure 20.

Figure 20: AIR Models for Use Across Integrated Applications

Such AIR project technologies would enable subjective meanings to be incorporated into user representations along with graphical images. This means that in addition to graphical models, semantic information about users will be formalized and represented in data structures representing characters. Such technologies could enable possibilities such as: a user seeing other avatars differently than the users' controlling those avatars do based upon cognitive models of stereotyping, virtual artifacts such as clothing or characteristics such as identity-in-context (a "gangster" in a rich area) can affect an avatar's possible actions and environmental responses to those actions, or both avatars' physical appearances and possibilities for social interaction could change depending on a history of discourse interactions such as typed conversation or gesturing. For a simple example of racial profiling in a computer game, if one wanted to cause all non-player characters of a particular type to respond suspiciously to a user, the ontology (knowledge specification structure) should entail that metadata indicating characteristics that are prototypical for a given race are negatively attributed. Similarly, such knowledge structures can inform how a given user-representation should be presented, or can even enable a user to have multiple representations (e.g., in the case of torque).

To enable dynamic profiles/identity-meaning generation system, procedures need to be constructed that allow underlying representations of characters to change on the fly. A clarifying example to illustrate how this could aid in constructing more critically aware games follows. In games, character actions performed by functions with names such as "cast_prayer" (associated with a priest character in a role-playing game) in "practice_painting" (associated with an artist in a suburban life simulation game) have effects that belie their names. Rather than enabling the actual actions, they merely manage numerical statistics, altering the values of hit (health) points or points on a career trajectory respectively. In contrast, distinctions between ethical or unethical actions could change the character's semantic information about morality. Such meaningful functions are the type that the AIR Project enables.

5. Conclusion

At the beginning of this paper we were motivated by Geoffrey Bowker and Leigh Star's questions: "Why should the computer scientist read African-American poets? What does information science have to do with race-critical or feminist methods and metaphysics?" Toward answering these questions, this paper has raised a broader call for a type of computing practice that engages the best of what cultural theorists, cognitive scientists, and social scientist have to offer to account for social phenomena as encountered in lived human experience. The case of social identity representation in computational media provides an excellent example, showing that social sciences and humanities insights are necessary for identity technologies that enable critically awareness. The preceding has described specific examples of how my current and future work attempts to do this.

Our interlocking perceptions and conceptualizations of the social worlds we inhabit are mitigated through imaginative cognition processes by which our identities are constructed in a manner far more complex than allowed by haunting checkboxes strung in arbitrary conjunctions -- such as, Caucasian or Near-Eastern, Asian or Pacific Islander, Black non-Latin. Even the rich mathematical language of computational data structuring and algorithmic processing can only grossly begin to approximate the complex possibility spaces of projected selves. This article has presented a case for, and example of, an iconoclastic approach to computing. This approach materializes concern for the human condition by using data-structuring and algorithmic manipulation to illustrate real world complexity via subjective metaphor, not to reify it in monolithic infrastructures. The work presented here has embodied an approach to developing theoretical models for analysis, design, and implementation that emphasize an aesthetics of polymorphic poetics and an approach that has been termed critical computing. The overriding hope is that the path hewn, speculating about phantom shapeshifting personae in a world of unstable social play, can critically spook those past poltergeists of essential identities enough to disturb them from their easy resting places in our computational media machines and perhaps cast their odd glows in a more illuminating way.


[1] Geoffrey C. Bowker and Susan Leigh Star, Sorting Things Out: Classification and Its Consequences. (Cambridge, MA: MIT Press, 1999).

[2] Otto Santa Ana, Brown Tide Rising: Metaphors of Latinos in Contemporary American Public Discourse. (Austin: University of Texas Press, 2002).

[3] Ralph Ellison, Invisible Man. (New York City: Random House, 1947).

[4] BBC "S Africa Chinese 'become black'." BBC News (Accessed on July 17, 2008,, 2008.

[5] Bethesda, "Oblivion," 2006. Blizzard, "World of Warcraft," 2005.

[6] D. Fox Harrell, "Toward a Theory of Phantasmal Media: An Imaginative Cognition- and Computation-Based Approach to Digital Media," CTheory, (2009)

[7] Philip E. Agre, Computation and Human Experience. (Cambridge, U.K.: Cambridge University Press, 1997), Philip E. Agre, "Toward a Critical Technical Practice: Lessons Learned in Trying to Reform AI," in Social Science, Technical Systems, and Cooperative Work: Beyond the Great Divide (Mahwah, NJ: Lawrence Erlbaum Associates, 1997), p. 131-158.

[8] Terry Winograd and Fernando Flores, Understanding Computers and Cognition: A New Foundation for Design. (Norwood, NJ: Albex Corporation, 1986), Hubert L. Dreyfus, What Computers Still Can't Do: A Critique of Artificial Reason. (Cambridge, MA: MIT Press, 1992).

[9] Kirsten Boehner, Shay David, Joseph 'Jofish' Kaye, and Phoebe Sengers, "Critical Technical Practice as a Methodology for Values in Design " in CHI 2005 Workshop on Quality, Values, and Choices, 2005.

[10] Brian K. Smith Critical Computing (Accessed on 4/11/2010,, 2002.

[11] Arthur and Marilouise Kroker, Critical Digital Studies: A Reader, Arthur Kroker and Marilouise Kroker, Eds. (Toronto, Canada: University of Toronto Press, 2008)

[12] D. Fox Harrell, "Computational and Cognitive Infrastructures of Stigma: Empowering Identity in Social Computing and Gaming," in Association for Computing Machinery (ACM) Cognition and Creativity Conference, Berkeley, CA, 2009.

[13] Ibid., See Also D. Fox Harrell, "Digital Metaphors for Phantom Selves: Computation, Mathematics, and Identity in Speculative and Fantastic Fiction and Gaming " in The 29th International Conference on the Fantastic in the Arts, Orlando, FL, 2008.

[14] D. Fox Harrell, 'The Gentleman Ghost vs. Shadowcat' (Accessed,, 2006.

[15] UESPWiki Oblivion:Races (Accessed on 1/30/2010,, 2009.

[16] A popular unofficial website dedicated to the game declares: "It is assumed that Imperials are based on the Romans, as it is assumed that Bretons are based on the French." Unknown Oblivion:Breton (Accessed on 4/06/2008,, 2008, Krzysztof Gonciarz The Elder Scrolls IV: Oblivion Game Guide (Accessed,, 2006.

[17] UESPWiki The Unofficial Elder Scroll Pages (Accessed on 1/30/2010,, 2009.

[18] Unknown, Roleplaying Muls on Armageddon (Accessed on 1/30/2010,, 2000.

[19] Ibid.

[20] Ibid.

[21] Bioware, "BioWare Aurora Engine Creature Format," 2003.

[22] Sonic Team, "Sonic the Hedgehog 2," Sega, 1992. Unknown sonic.gif (Accessed on 1/30/2010,, 2010.

[23] Harrell, "Computational and Cognitive Infrastructures of Stigma: Empowering Identity in Social Computing and Gaming," 2009.

[24] George Lakoff, Women, Fire, and Dangerous Things: What Categories Reveal about the Mind. (Chicago, IL: University of Chicago Press, 1987).

[25] Joseph Goguen, "Towards a Social, Ethical Theory of Information," in Social Science Research, Technical Systems and Cooperative Work, L. G. Geoffrey Bowker, Leigh Star, William Turner, Ed. (Mahwah, N.J.: Lawrence Erlbaum Associates, 1997), p. 27-56. See also Bowker and Star, Sorting Things Out, 1999.

[26] James Paul Gee, What Video Games Have To Teach Us About Learning and Literacy. (New York City: Palgrave Macmillan, 2003).

[27] Lakoff, Women, Fire, and Dangerous Things, 1987.

[28] Ibid.

[29] Otto Santa Ana, Brown Tide Rising, 2002.

[30] Lakoff, Women, Fire, and Dangerous Things, 1987.

[31] Ibid.

[32] Bowker and Star, Sorting Things Out, 1999.

[33] Erving Goffman, The Presentation of Self in Everyday Life. (New York: Anchor Books, 1959).

[34] Erving Goffman, Stigma: Notes on the Management of Spoiled Identity. (Englewood Cliffs:NJ: Prentice Hall, 1963).

[35] Gee, What Video Games Have To Teach Us About Learning and Literacy, 2003. Following the terminology of conceptual blending theory, and to further clarify the nature of the construct, the AIR model uses the term "projected identity" rather than "projective" as does Gee.

[36] Ibid.

[37] Zach Waggoner, My Avatar, My Self: Identity in Video Role-Playing Games. (Jefferson, NC: McFarland & Company, 2009).

[38] Jill Walker Rettberg "Digital Culture, Play and Identity: A World of Warcraft Reader, " in Digital Culture, Play and Identity: A World of Warcraft Reader, Hilde Corneliussen and Jill Walker Rettberg, Eds. (Cambridge, MA: MIT Press, 2008).

[39] Gilles Fauconnier and Mark Turner, The Way We Think: Conceptual Blending and the Mind's Hidden Complexities. (New York: Basic Books, 2002).

[40] Mark Turner, "Double-Scope Stories," in Narrative Theory and the Cognitive Sciences, D. Herman, Ed. (Stanford, CA: CSLI Publications, 2003).

[41] Raymond W. Gibbs Jr., "Making good psychology out of blending theory," Cognitive Linguistics, 11, (2000). p. 347-358.

[42] D. Fox Harrell, "Algebra of Identity: Skin of Wind, Skin of Streams, Skin of Shadows, Skin of Vapor," in Critical Digital Studies: A Reader, A. Kroker and M. Kroker, Eds. (Toronto: University of Toronto Press, 2008).

[43] Eric Kabisch, "Datascape: A Synthesis of Digital and Embodied Worlds," Space and Culture, (no.11, 2008) p. 222-238.

[44] Unknown Redguard (Accessed on 2/01/2010,, 2009.

[45] Blizzard WoW -> Info -> Races -> Tauren (Accessed,, 2004.

[46] Gee, What Video Games Have To Teach Us About Learning and Literacy, 2003.

[47] George Lakoff and Mark Johnson, Metaphors We Live By. (Chicago, IL: University of Chicago Press, 1980). George Lakoff and Mark Turner, More than cool reason -- a field guide to poetic metaphor. (Chicago, IL: University of Chicago Press, 1989).

[48] Sony Computer Entertainment, "Shadow of the Colossus," 2005.

[49] Unknown (Accessed on 10/30/2007), 2006.

[50] Big Blue Box Lionhead Studios, Robosoft Technologies, "Fable," Microsoft Game Studios, Feral Interactive, 2004.

[51] Ibid., Stephen Tolito, "Letting Gamers Play God, and Now Themselves," in New York Times (New York: New York Times, 2004).

[52] Lakoff and Johnson Metaphors we Live By, 1980, Lakoff and Turner More Than Cool Reason, 1989. Jason Roher, "Passage,", 2007.

[53] Gilles Fauconnier, "Rethinking Metaphor," in Cambridge Handbook of Metaphor and Thought, R. Gibbs, Ed. (Cambridge, U.K.: Cambridge University Press, 2006).

[54] Samuel R. Delany, Trouble on Triton: An Ambiguous Heterotopia. (Hanover, NH: Wesleyan University Press, 1976).

[55] Henry Louis Gates Jr., The Signifying Monkey: A Theory of African-American Literary Criticism. (New York: Oxford, 1988).

[56] D. Fox Harrell, "GRIOT's Tales of Haints and Seraphs: A Computational Narrative Generation System,," in Second Person: Role-Playing and Story in Games and Playable Media, N. Wardrip-Fruin and P. Harrigan, Eds. (Cambridge, MA: MIT Press, 2007).

[57] Example output can be found in Harrell, "Algebra of Identity: Skin of Wind, Skin of Streams, Skin of Shadows, Skin of Vapor," in Critical Digital Studies: A Reader, 2008.

[58] D. Fox Harrell and Kenny K. N. Chow, "Generative Visual Renku: Linked Poetry Generation with the GRIOT System," in Electronic Literature Organization Conference (Vancouver, WA, 2008).

[59] See D. Fox Harrell, "Toward a Theory of Phantasmal Media: An Imaginative Cognition- and Computation-Based Approach to Digital Media," (CTheory, 2009), D. Fox Harrell, "Theory and Technology for Computational Narrative: An Approach to Generative and Interactive Narrative with Bases in Algebraic Semiotics and Cognitive Linguistics." Ph.D. Dissertation, 2007). D. Fox Harrell, "Walking Blues Changes Undersea: Imaginative Narrative in Interactive Poetry Generation with the GRIOT System," in AAAI 2006 Workshop in Computational Aesthetics: Artificial Intelligence Approaches to Happiness and Beauty, Boston, MA, 2006.

[60] D. Fox Harrell, Daniel Upton, Ben Medler, and Jichen Zhu, "Define Me: A Cognitive and Computational Approach to Critical Digital Identity Representation in Social Networking Applications," in Digital Humanities Conference, Baltimore, MD, 2009.

[61] Lakoff, Women, Fire, and Dangerous Things, 1987.

[62] Ibid.

[63] Mark Turner, The Literary Mind: The Origins of Thought and Language. (New York: Oxford University Press, 1996).

[64] Roman Jakobson, "Closing Statement: Linguistics and Poetics," in Style in Language, T. Sebeok, Ed. (Cambridge, MA: MIT Press, 1960).

[65] Often in hierarchical binary oppositions as criticized in Harrell, "Algebra of Identity: Skin of Wind, Skin of Streams, Skin of Shadows, Skin of Vapor," in Critical Digital Studies: A Reader, 2008.

[66] D. Fox Harrell, "Cultural Roots for Computing: The Case of African Diasporic Orature and Computational Narrative in the GRIOT," in Digital Arts and Culture 2007, Perth, Australia, 2007. Joseph Goguen, "An Introduction to Algebraic Semiotics, with Applications to User Interface Design," in Computation for Metaphors, Analogy, and Agents, Yakamatsu, Japan, 1998.

Dr. D. Fox Harrell is a researcher exploring the relationship between imaginative cognition and computation. He is an Assistant Professor of Digital Media at the Georgia Institute of Technology. He directs the Imagination, Computation, and Expression (ICE) Lab/Studio ( in developing new forms of computational narrative and poetry, gaming, social networking, and related technical-cultural media with bases in computer science, cognitive science, and digital media arts. The National Science Foundation has recognized Harrell with an NSF CAREER Award for his project "Computing for Advanced Identity Representation." Harrell holds a Ph.D. in Computer Science and Cognitive Science from the University of California, San Diego. He also earned a B.F.A. in Art, a B.S. in Logic and Computation, and minor in Computer Science at Carnegie Mellon University, each with highest honors. He has also worked as an interactive television producer and as a game designer.