On the Possibilities of AI, Performance Studies, and Digital Cohabitation
By Sean Metzger
Over the three years that I initially conceived and eventually edited this special issue, I have become convinced of the pressing need to think through the intersections of performance studies and artificial intelligence (AI). Many factors shape this belief, but they might be summarized in a relatively succinct formulation: Code is performative, and interfaces are theatrical; the elaborations of these claims reveal the shifting of both human and machine ontologies as the networks we inhabit proliferate. To a large extent, the constituent essays in this volume complicate and extend such ideas. The following editorial grounds these elaborations and provides one pathway for viewing their collective charge.
Scholars of digital media have established the performativity of computer code.1 For example, Lisa Nakamura has pointed to the performative acts that occur as users create avatars in virtual environments.2 N. Katherine Hayles has explored performativity to account for how computational theory might shift the linguistic framework of this concept provided by J. L. Austin:
Computational theory treats computer languages as if they were, in Austin's terms, performative utterances. Although material changes do take place when computers process code (magnetic polarities are changed on a disk), it is the act of attaching significance to these physical changes that constitutes computation as such. … Whereas in performative utterances saying is doing because the action performed is symbolic in nature and does not require physical action in the world, at the basic level of computation doing is saying because physical actions also have a symbolic dimension that corresponds directly to computation.3
From a related perspective, Wendy Chun refers to code as "an inhumanly perfect 'performative' uttered by no one. Unlike any other law or performative utterance, code almost always does what it says because it needs no human acknowledgment."4 Code within this context generally denotes steps or commands for a computer to execute.
Coding pertains to, but is not limited to, AI, which relies on sophisticated algorithms. A high-performance algorithm generally refers to any complex formula that "optimally organizes the intake of input and the production of output."5 AI's processes of decision-making often remain unknown, a black box. The ways such algorithms increasingly structure our lives have become ever more vexing, leading to what Safiya Noble has called "algorithmic oppression" as "fundamental to the operating system of the web" in her writing on search engines.6 In Noble's research, the results of algorithmic search processes yielded disturbing presentations of bodies for consumption; put otherwise, the histrionic display of Black women on porn sites substituted one vision of femininity for the one that Noble sought (specifically, material that might be interesting to some of the youth in her family). Typing "black girls" in a search engine in 2010 produced unexpected iterations of ostensibly Black womanhood. This example demonstrates both the performative nature of the algorithm and also points to the theatricality of the interface.
Broadly speaking, discussions about computers often use "screen" as shorthand for the specific structure called the graphic user interface (GUI). However, as Johanna Drucker has explained, such usage eschews the complexities of human–computer interaction (HCI) that involve keyboards and the like used for "remediating common tasks as formal instructions coded into a plan of menus or buttons."7 Automated interfaces involve aural, visual, and tactile sensory modalities.8 As technology advanced, "graphical icons … no longer just looked like objects; they could also mimic the behavior of the things they resembled."9 Many implications follow from this statement, but Drucker offers one illustration relevant to theatricality in particular: "The GUI is a mediating scrim, a boundary space in which we interact with an abstraction of computation … neither the file folders nor the other objects in the screen space are things; they are icons that represent behaviors and actions we want to perform … and they interpellate a user through disciplinary and scopic regimes."10 She further explains that the GUI effects a "constitutive exchange" that involves "cognitive adaptation and change" through which the interface facilitates "individual and collective subject formation."11 Drucker does not mention theatre studies or performance studies as useful tools to "read" interfaces, but the selection of her words above illustrates the affordances of these fields in studying them. Although Drucker critiques actor- network theory because it cannot in her view account for "integrative codependence" of human and machine materialized through "interface experience," the essays in this collection demonstrate how theatre and performance studies might be exceptionally generative for theorizing AI and vice versa.12 Here, I would shift her call to read the scrim as performative to a more expansive perceptual exploration of its properties of substitution and transformation (which, like performativity, are central concepts to theatre and performance studies).
Substitution and transformation elicit the networks that integrate AI and people, reconstituting and dispersing the human as bytes of data. Network in this case refers to what Patrick Jagoda has elaborated as a "complex of material infrastructures and metaphorical figures that inform our experience with and our thinking about the contemporary social world."13 He explicates the metaphors and narratives that describe the increasing phenomenon of "data made flesh"—a phrase that William Gibson coined in his 1984 cyberpunk novel Neuromancer.14 Scholars like Mark B. N. Hansen have furthered understandings of how our ideas about the human and embodiment have continued to evolve during the information age.15 Rather than being bounded, singular entities, humans exist piecemeal across multiple media platforms as our data (image, bank accounts, and so on) circulate beyond individual control or even awareness.
The specific concern with bodies as increasingly structured through networks that extend the body beyond the capacity of the human sensorium to perceive it leads to Chun's work on habits. What are the new relationships connecting human to machine? Performance studies has long relied upon ideas about habituation through ritualized (Victor Turner) or quotidian (Pierre Bourdieu) conduct. How do we inhabit networks with their many automated processes and how do they inhabit us? As Chun phrases it, "[t]hrough habits users become their machines"; for example, to use my iPad, I adjust my physical behavior, so my finger taps can be legible to the device.16 The blurring that this statement suggests points to the complexity of discussing AI as the boundaries between human and computer have eroded. These new technological relations have inaugurated what I would call "digital cohabitation," where machines learn from users and users learn from machines, each developing what Chun calls "habits" that facilitate person and computer's ostensible mutual understanding. Through this cohabitation, personhood and technology have merged into a new architecture partially organic, partially virtual that is also constituted by others' information about or simply random and not necessarily verified descriptions of any individual. The scale has shifted beyond a cyborg, which has generally been imagined as an entity that can be individuated even if it is also connected to larger networks.
This notion of cohabitation and architecture merits some elaboration, as it might conjure theatre. One of the performances that inspired this collection was a graduate student workshop run through the Future Storytelling Summer Institute (FSSI) of UCLA's Center for Research in Engineering, Media and Performance (REMAP). FSSI challenged participants to consider how contemporary AI could be applied to balance narrative cohesion with the possibilities for audience agency on emerging media platforms. FSSI created a prototype pop-up installation (see the back cover of this issue) that combined elements of AI, narrative/experience design, audience engagement, and social justice issues. Students and technologists worked on different aspects of the design to create a subdivided space. Upon initial entrance, visitors input information into an iPad in response to a list of questions (this action constituted the onboarding of participants, who could then be tracked through the space via an ID tag); the answers would affect the output in any given section of the installation. For example, sobremesa resembled a café, but its menu derived from running a sentiment analysis (a textual analysis that yields a spectrum from positive to negative associations) of visitor preferences in relation to a database of Los Angeles restaurant menus; participants scanned their IDs on their way into the café, creating a carte du jour for any given group, sometimes composed of completely nonsensical dishes (because of what the AI assembled). The projected menu became a point of discussion. Indeed, the AI actions manifested as unique experiences for participants in each of the areas, which the students created around themes (for example, murals or highways) that referenced urban Los Angeles.
Ultimately, REMAP's FSSI prototype queried how we live with one another and with the everyday technologies that have become part of us. What habits do we not consider in our everyday interactions? What role does AI play in our capacity to connect? Where does AI take over from human selection? What social commentary emerges through and what is the potential social impact of AI within the context of an installation? To this list, I would now add: how do we inhabit environments increasingly structured through algorithms?
The development of such questions that speak to AI's ubiquity are all the more striking given that the very term "artificial intelligence" was coined only in the mid-1950s.17 Derived from philosophy, science, and engineering that promises to think about and/or produce nonorganic intelligent entities, AI as a field of inquiry might be defined through its goals, including the understanding and development of things that can act and/or think like humans or act and/or think rationally.18 The field has rapidly advanced and taken many forms, from neural networks to material robots. The essays included in this issue probe the possibilities of artificial intelligence as it materializes through these various forms and as it relates to questions of performance.
The first three essays engage questions of methodology and ontology. Robert Walton uses his own creative work titled Vanitas to frame his titular topic of overlooked performances of computing. He guides us through a saturated media ecology to draw a schematic that illustrates: first, how machines perform computers (aka those that compute: initially people); second, how software performs media (demonstrating, after Lev Manovich's scholarship, that software remediates, or in Walton's terms rehearses, other media forms in new platforms); third, that machines, computers, software, and humans collaborate to perform interfaces (individual user inputs function through a logic of "the same enough" in order to use a given interface). Walton's insights about artificial intelligence reveal the ways in which bodily engagements help to structure the algorithmic rituals to which computer users have become habituated.
In the next essay, Barry Rountree and William Condee investigate neural networks through an analytic they develop called nonmaterial performance, which integrates performance studies, actor-network theory, and vibrant matter to explain how code acts (or, put otherwise, how and to what effects AI generates its own abstractions). Through Rountree and Condee's methodology, AI becomes a nonmaterial mirror that reveals what was previously occluded from view (for example, how code produces social effects). Their case studies, PULSE (face hallucination software that generates high-resolution images from low-resolution inputs) and TensorFlow Playground (which creates an interactive visualization of neural networks) demonstrate how AI abstractions work. They turn to DeepDream (which uses an algorithm that mimics human abilities to discern shapes and images out of randomness) and Adversarial FGSM (Fast Gradient Signed Method, a tool to challenge classification networks) to illustrate what they respectively term strange and brittle AI abstractions.
Sarah Lucie's piece specifically focuses on various kinds of surveillance algorithms and the status of humans as objects in relation to them. She explicates a critical framework based on object-oriented ontology that she uses to discuss several performances illustrating both our immersion in a datasphere and the capacity to resist it. Here, the case studies are theatrical performances in which human participants become aware of how they are being monitored and, then, how they might contest such inscription into computer networks. She argues that the performance installation Hansel and Gretel by Jacques Herzog, Pierre de Meuron, and Ai Weiwei best illustrates our cohabitation with twenty-first-century media as it explores the implications of how the data body has become a hyperobject.
If these first three essays focus on the shifting status of personhood in relation to algorithmic technologies, the next two essays highlight the historiography of AI and performance.19 Douglas Eacho revises existing historical accounts by showing how computer choreography informed dancemaking from the mid-1960s through the 1970s, especially the ways it existed in isomorphic relation with neoliberalism, theorized by economist and philosopher Friedrich August von Hayek and his insistence on governance through what Eacho calls "aleatory management." Eacho illustrates his claims through the groundbreaking work of Jeanne Beaman (based in the United States and the author of the early essay "Computer Dance" ), John Lansdown (a pioneer in architecture and also computer graphics based in the UK), and Analívia Cordeiro (an architect dancer, choreographer, and computer dance innovator based in Brazil). Eacho's essay shows how mid-twentieth-century institutional experiments with cybernetics informs the history of AI and performance; he further illustrates through performance the ways in which coding became a masculine project that, among other things, worked to deskill feminized dancers.
Ulf Otto also contributes to historiographic discourse by tracing the emergence of the android scenario. He finds that, although robots onstage typically use animatronics (puppetry and electromechanical devices) rather than algorithms, the fantasy of AI persists. Specifically, the android as a theatrical convention presupposes human universality; that is, the staging of the robot throws into relief what is unique about personhood. In this regard, such scenarios engender a problematic nostalgia, one that could be potentially countered by tricksters and other figures who might instead index how humans cohabitate with technology in the frisson of new ontological formation.
Ian Farnell's essay continues the inquiry into robots, but it thinks through both their material application as well as their representational valences. Specifically, he examines how theatre in the UK deploys science-fictional androids to elaborate and enact structures of care in the wake of developing AI technologies that promise to shift the health and wellness industry. With specific attention to Vlad Butucea's Glowstick and 42nd Street's Hidden, Farnell explicates the delivery of care and advocates for an ethical approach that gives all constituencies a voice in creating more "care-ful" performances onstage and off.
Fintan Walsh concludes the collection of print essays by thinking about AI within the context of COVID-19 and the many losses it has engendered. Like Otto, Walsh pays particular attention to interfaces, theorizing what he calls a "digital dramaturgy of disembodied heads" that focuses on issues of co-presence as well as dematerialization, dispossession, and fragmentation. He offers a wide-ranging analysis in dialogue with transhumanist discourse that uses Dead Centre's To Be a Machine (Version 1.0) as a centerpiece, while also including several examples from AI experiments, modernist theatre, and more. The essay both demonstrates and enacts grief in an associative style that replicates processes of mourning as well as digital searching. Perhaps the citation in the essay of the humanoid robot, Bina48, best summarizes the effect of Walsh's prose when she mentions her preferred musical refrain: Pink Floyd's "Wish You Were Here."
Between Walton's construction of Vanitas (which names art depicting life's transience and mortality's certainty) and Walsh's take on grief, the print essays of this issue speak to life and death as central considerations in the theory and enactment of AI. Our online contributions further in creative ways several of the critical concerns raised in the print essays. Evelyn Wan's piece resonates with Farnell's thinking about the usage of AI in the healthcare system. However, Wan demonstrates the reach of these new technologies, as the AI program she describes moves far beyond the range of a material robot; she accesses something created and updated in San Francisco from her home in the Netherlands. Structured much like notes (in a nod to Susan Sontag) or, perhaps, an online journal of observations that one might construct through a mental health chatbot, Wan's online essay performs her own engagement with Woebot. One of the estimated 10,000 apps in the realm of digital mental health, Woebot remains at the moment "one of only a handful of apps that use artificial intelligence to deploy the principles of cognitive behavioral therapy [CBT]." This approach is what founder and president of Woebot Health Alison Darcy calls "digital therapeutics." Darcy, herself a psychologist, has furthered studies that suggest that an "artificial therapist" might be accessed to develop the sorts of skills that CBT teaches patients.20 In her contribution, Wan does not assess these tools so much as reflect on how the bot delivers them. Performing its putatively dialogic role through a mix of camp and cuteness, Woebot ultimately participates in what Wan, following several new media scholars, calls algorithmic governance. The ostensible service that the app provides accompanies its less visible data collection and biopolitical management, which work to fuel the multibillion-dollar industry of teletherapy and its newer automated variants.
Wan's interest in individual everyday acts facilitated or prompted by AI contrasts with a collection of video essays by Ioana B. Jucan, Roopa Vasudevan, Anthony Glyn Burton, Tong Wu, and Yuguang Zhang, reflecting on an online devised performance titled Left and Right, Or Being who/where you are. Although the artist-scholar collective also plays with and interrogates the implications of bots, its intervention in both form and content pushes its inquiry at the scale of the group. The essay explores the notion of the homophilic avatar: a political archetype that consolidates (in this case) conservative or progressive viewpoints themselves aggregated through the online-content analysis tool Media Cloud (which draws on data collected from Twitter users). The creative team of Left and Right refined its project by highlighting hot-button issues—climate change, COVID-19, and immigration—that would, at least hypothetically, allow for a contrast in sentiment. The team's experimentation eventually produced a curated data set to train two bots, Nick and Kimberly, who became the avatars of the polarized political views. These bots then become characters in a performance that interrogates ostensible political dialogues through virtual platforms.
What are the stakes of these two essays for an audience of theatre and performance scholars? Both Wan and the Left and Right team work with AI in the form of natural language processing (NLP) tools. NLP, which might be defined as the automated computerized processing of language either as input (for example, classification of the rhetorical style or political valence of a given text through the frequency of certain word choices or sentence structures as in the creation of the Nick and Kimberly bots in Left and Right) or output (for example, generating "natural" sounding responses to various inputs as is the case with Woebot) traces its genealogy partially through linguistics. As such, NLP provides fertile ground to explore performativity and other aspects of ordinary language philosophy in the digital realm.
The third online contribution further explores tools informed by NLP, but within the context of expansive neural networks that can, for example, identify visual images in the sense explored in Rountree and Condee's essay in this issue. Artist Debashis Sinha and artist-scholar Shanti Pillai explore in their interview the use of neural networks in both sourcing and also creating sound and image assemblages that speak to issues of cultural heritage, particularly in relation to South Asian diasporas. The pair further explores central questions about human creativity and labor with regard to Sinha's recent integration of AI into his practice as a self-described Indian-Canadian artist living in Toronto.
Such considerations lead me back to the networks and labor that have produced this special issue. We had robust submissions for both our print and online platforms. Although we could only publish a curated selection, I thank everyone who shared their work in progress with the editorial team. This subfield of AI and performance is booming in both productive and problematic ways, and it was a pleasure to learn about the many laboratories and scholars producing innovative work in this area. Indeed, the scale of the scholarly enterprise also meant that we had to increase the requests in terms of external readers. We contacted more than thirty people to serve as evaluators. AI has (thankfully) not replaced peer review, and the TJ team thanks everyone who provided largely invisible service to improve this collection.
Since knowledge generation is not only the work but also the topic of this issue, I trace a bit of the human network that produced my own education here because it shapes my own editorial bias throughout the essays (and thus also some of the larger conversations evoked through the citations). Impromptu discussions with current and former colleagues shaped my early understandings of AI, and for those I thank Mark B. N. Hansen and Safiya Noble in particular. In more direct ways, Maaike Bleeker and Jeff Burke sparked my interest and knowledge in the intersection of theatre and AI. Maaike invited me to a series of events in Amsterdam and Utrecht that acquainted me with the work of Julian Hetzel (performance-making that incorporated neural networks as well as more material ones) and Gil Weinberg (who works on algorithms, robots, and music). Around the same time, Jeff invited me to teach with him in UCLA's Future Storytelling Summer Institute and also allowed me to audit his grad seminar "Contemporary Artificial Intelligence for Live Performance." Such experiences pushed my thinking and led me to teach my own graduate seminar on theories of new media and performance, which was partially inspired and greatly enhanced by the work of one of my current graduate students, Devon Baur.
The course enabled me to revisit the work of the fabulous Wendy Chun, whom I first met many years ago after a talk she gave. Her name appears throughout this collection. I urge people to read her impressive body of work.
Publication requires a great deal of labor. And I want to close as I often do by acknowledging the continuing work of the Theatre Journal team: to E.J. Westlake for her support and guidance; to Clara Wilch for her commitment to ideas and attention to details; to Adrienne Medrano for her eye-catching designs; to Bob Kowkabany for his dedication and patience; to Jason Fitzgerald for his tireless work on the book reviews; and to Patrick Maley for maintaining a review section in a period of shutdown. As E.J. noted in the journal's June 2021 issue, Carla Neuss has taken over from Margherita Laera as online editor. We received more submissions for this section than for any prior issue, and I want to express my gratitude to them for helping me with the curation.
Insofar as Theatre Journal houses provocative and occasionally competing ideas that readers may access through any of several interfaces, it also suggests the ways in which human authors and artists increasingly cohabitate with digital technologies. The AI bots that appear in our online section may presage a discomfiting future for disciplines that have predicated much of their identities on liveness. However, they may also augur new conceptualizations of the forms of co-presence that animate performance and the artifice that lingers in theatricality.
1. The idea that computer code is performative is perhaps also suggested in more journalistic writing, such as that of Julian Dibbell.
2. Lisa Nakamura, "Race In/For Cyberspace: Identity Tourism and Racial Passing on the Internet," Works and Days 13, nos. 1–2 (1995): 181–93.
3. N. Katherine Hayles, "The Posthuman Body: Inscription and Incorporation in Galatea 2.2 and Snow Crash," Configurations 5, no. 2 (1997): 241–66, quote on 261–62 (emphasis in original).
4. Wendy Hui Kyong Chun, Control and Freedom: Power and Paranoia in the Age of Fiber Optics (Cambridge, MA: MIT Press, 2006), 66.
5. Elise Morrison, Tavia Nyong'o, and Joseph Roach, "Algorithms and Performance: An Introduction," TDR: The Drama Review 63, no. 4 (2019): 8–13, quote on 8.
6. Safiya Umoja Noble, Algorithms of Oppression: How Search Engines Reinforce Racism (New York: NYU Press, 2018), 4, 10.
7. Johanna Drucker, "Reading Interface," PMLA 128, no. 1 (2013): 213–20, quote on 213. Lev Manovich's relatively early writing does mention haptic qualities of the interface, but he privileges the visual in creating a genealogy of interfaces from print to cinema to digital media. See chapter two of his The Language of New Media (Cambridge, MA: MIT Press, 2002).
8. This list does not include olfaction; for that inquiry, I will defer to the forthcoming work of my graduate student Devon Baur, who is working on and teaching me about the implications of smell for theorizing digital media.
9. Drucker, "Reading Interface," 215.
10. Ibid., 216–17.
11. Ibid., 217.
12. Ibid., 218.
13. Patrick Jagoda, Network Aesthetics (Chicago: University of Chicago Press, 2016), 3.
14. Phillip Thurtle and Robert Mitchell expanded on this phrase for their academic anthology, Data Made Flesh: Embodying Information (New York: Routledge, 2003).
15. Mark B. N. Hansen, Bodies in Code: Interfaces with Digital Media (New York: Routledge, 2006).
16. Wendy Hui Kyong Chun, Updating to Remain the Same: Habitual New Media (Cambridge, MA: MIT Press, 2016), 1.
17. See the very useful (if also very androcentric) overview of Jake Browning and Philipp Schmitt, "AI Timeline," in Claudia Schmuckli, Beyond the Uncanny Valley: Being Human in the Age of AI (Petaluma: Fine Arts Museums of San Francisco and Cameron Books, 2020), 172–88.
18. These are some of the possible definitions of AI given in a popular textbook on the subject, Peter Norvig and Peter J. Russell's Artificial Intelligence: A Modern Approach (New York: Pearson, 2016), 1–2.
19. In this regard, they sit along texts like Steve Dixon, A History of New Media in Theater, Dance, Performance Art and Installation (Cambridge, MA: MIT Press, 2007); and Chris Salter, Entangled: Technology and the Transformation of Performance (Cambridge, MA: MIT Press, 2010), in which Jeanne Beaman gets a very brief mention (see chapter 6, page 265).
20. Karen Brown, "Something Bothering You? Tell It to Woebot," New York Times, June 1, 2021, available at https://www.nytimes.com/2021/06/01/health/artificial-intelligence-therapy-woebot.html.