Do Digital Agents Do Dada?

Do digital agents do Dadaism? To answer this question, we review a series of theatrical experiments involving human improvisers and AI-controlled Cyborgs in front of audiences. We describe these experiments and discuss the use of conversational digital agents (DA) on the stage. We identify two basic strategies of staging machines: the “immersive approach”, and the “Dadaistic approach”. We draw on Dadaistic conceptions of embracing modernity specifically in the Dadaists’ obsession with androids and cyborgs. Through analysis of several stage experiments we contend that digital agents, while attempting to build believable characters, do Dada, only if we do Dada too. Motivation: Improvisation and Digital Agents Improvised theatre, improv, is an art form modelled on natural human interaction which demands constant adaptation to evolving contexts. Previous research has paralleled improv theatre and jazz music and formulated both as “realtime dynamical problem solving” (Bruce and others 2000; Johnson-Laird 2002; Magerko and others 2009). The first problem is collectively creating stories by impersonating believable characters and incorporating narrative elements suggested by the audience. The second problem is how to be truthful and in the moment of the scene, while accepting the offers made by the other improvisers, the audience, or their own cultural background (Johnstone 1979; Merritt and Hines 2019). The third problem is finding the limits of an audience’s expectations. Logic in Improv and the Circle of Expectation Improvisers follow the rules of logic while establishing a specific universe in which an improvised scene takes place. In this way, the performers and the audience can follow the story, and can predict how the scene continues (Merritt and Hines 2019). Practitioners introduced the Circle of Expectation as a concept to qualify the difference between adding obvious narrative elements that make the story more specific versus those that violate the expectations of the audience (Johnstone 2014). The circle contains the assumptions and associations that define the dramatic world (Mathewson and others 2020). Improvisors make offers by modelleing Figure 1: TL: Improv with a robotic digital agent controlled by an AI chatbot. TR: Improv where a human actor reads lines from the AI chatbot displayed on a screen. BL: Psychotherapist chatbot ELIZA appearing as a projection. BR: Improv (downstage) with one actor receiving lines from an AI chatbot controlled by a computer operator (upstage). Credits at improbotics.org obvious next-steps from the mind of the audience. Improvisers first establish the who, what, where, when, and the relationship between the characters, and then “do the most obvious thing” to move the story forward (Johnstone 1979). Digital Agents in Improvisational Theatre For digital agents (DA) based on randomness, adhering to logic in improv, while staying within the Circle of Expectation, seems an impossible task. Nevertheless, the field has adopted technological trends continues to innovate toward this goal. Recent work builds upon computational improvisation in music and dance performance (Fiebrink 2011; Hoffman and Weinberg 2011; Long and others 2020), and collaborative storytelling (Perlin and Goldberg 1996; Hayes-Roth and Van Gent 1996; Riedl and Stern 2006; Magerko and others 2011). There have also been attempts at generating high-level narrative consistency for improvised theatre. DAs have acted as directing narrators grounding the performances of human improvisers with generated plot points (Eger and Mathewson 2018). Other DAs have quantified and shaped Proceedings of the 11th International Conference on Computational Creativity (ICCC’20) ISBN: 978-989-54160-2-8 488 the narrative arc of generated text in interactive humanmachine dialogue, in order to reveal or conceal information according to the Circle of Expectation (Mathewson and others 2020). DAs have been used elsewhere in improv theatre (Bruce and others 2000; Baumer and Magerko 2010; O’Neill and others 2011; Knight and others 2011; Jacob 2019) and Simone’s Bot Party: Improv Comedy with Robots. Progress in machine learning for natural language processing, specifically neural networks for text generation (Vinyals and Le 2015; Radford and others 2019), encouraged practitioners to build DAs for improv by focusing on the conversational and storytelling aspects. Mathewson and Mirowski innovated in this space with their collaborative human-robot improv group HumanMachine.1 They developed A.L.Ex., an advanced conversational chatbot trained on film dialogue from OpenSubtitles (Tiedemann 2009) which interfaced with speech recognition, synthetic text-to-speech, and controlled a humanoid robot stage partner (Mathewson and Mirowski 2017a; 2017b). Advancing Digital Agents for the Stage This study focuses on AI improv, where the robot is replaced by a human, who performs lines provided by an AI chatbot. Examples of such shows include Improbotics2 by Mirowski and Mathewson, Yes, Android by Etan Muskat and Almost Human3 by Gunter Loesel. In Improbotics, a DA sends lines of dialogue to a human performer Cyborg. While the Cyborg is free to move and to express non-verbal acting and emotional subtext, they can only say AI-generated lines. Those lines of dialogue are generated in response to context typed by a human Operator who also serves as curator in the case when multiple choices are offered. The AI plays the role of an interactive playwright, giving lines to a specific performer, while challenging the other improvisers to justify the potentially nonsensical lines of dialogue. Experiments have investigated what kinds of theatrical frames emerge through the interaction of humans and machines on the stage, how one can describe and explain these theatrical phenomena, and what value these partly machine-generated dialogues can have in an artistic and aesthetic sense (Martin and others 2016; Mathewson and Mirowski 2017b). Previous work uses artistic lenses to contextualize novel AI technology (Horswill 2012; 2016). We analyze DA improv through the lens of Dadaism, which, we argue, is a suitable artistic frame to understand human-machine cocreativity improvisation. Avant-garde Movements and Dadaism The avant-garde movements at the beginning of the 20th century were the first artistic response to modernity and the industrial age. As the art historian Matthew Biro points out, the Dada movement was quite obsessed with the figure of the cyborg (Biro 2009)—though the word “cyborg” did not exist at that time. Dada artists, especially in the Berlin Dada 1 https://humanmachine.live 2 https://improbotics.org/ 3 https://www.stupidlovers.de/de/show-termin/almost-human group, explored and involved machines in various ways to express their ambiguous relationship towards modern life. They were furious activists against the political and military “machinery” that led to the first world war. For example, Hausmann’s drawing “Der eiserne Hindenburg” (1920) displays the German general Hindenburg as half-human, halfmachine indicating the de-humanizing impact of technological war and chauvinistic patriotism. Not only did Dada artists point towards the political aspects of automatization, they also reflected on the human self turning more machine-like. In two famous self-portraits, George Grosz’s “Daum marries her pedantic automaton George” (1920) and Hausmann’s “Selfportrait of the Dadasoph” (1920), the Dadaists displayed themselves as halfmachine and half-human. These works explored a new relationship between human and machine. The humans had no outside perspective to look at the machine as a human. They are already changed and turned into cyborgs and are unable to distance themselves from the machine. This reflection is explored in (Hausmann 1921): “Why can’t we paint pictures today like those of Botticelli, Micheangelo, Leonardo, or Titian? Because human beings have completely changed in terms of their consciousness. This is the case not simply because we have the telephone, the airplane, the electric piano, and the escalator, but rather because these experiences have transformed our entire psychophysical condition.” The naı̈ve perspective of the human as a counterpart of technology was questioned even at this early stage of reflecting on modernity. Consequently, the Dadaists proposed an art not made by humans. Grosz stated this clearly at the First International Dada Fair in 1920, when he declared: “Art is dead. Long live the new machine art of Tatlin” (Broeckmann 2016). Dadaism has a couple of identifiable traits. One of particular interest is that the artists deconstructed meaning, arriving at the smallest units of expression—letters or phonemes—resulting in poems that consist of single letters only (Hausmann, “fmsbw”, 1918). This results in the Dadaistic practice of reducing the world to its basic units. Then putting those pieces together again in a nonsensical way, thereby uncovering the meaninglessness under the smooth veneer of sense-making. Optophonetic poetry (i.e. visible speech sounds) is an early example of the ambiguity of cyborg art. While subverting the meaning of language, it was performed in a expressive way, as can be experienced through recordings with Hausmann reading his own poetry.4 Dadaists did not embrace the machine age. They squeezed it so hard that modernity stopped being something you could look at from the outside. It became a feature internalized by modern human beings. Artistic Strategies for Cyborg Improv In appearance, trying to build a DA that acts like a human seems superfluous artistically. Even if it were possible, which at the moment it is not, such an endeavour would only mimick human-to-human interaction. Quoting (Turing 1951): “But, I certainly hope and believe that no great efforts 4 https://youtu.be/2lVqiCURmFQ Proceedings of the 11th International Conference on Computational Creativity (ICCC’20) ISBN: 978-989-54160-2-8 489 will be p