i, chatbot the gender and race performativity of conversational agents

Trang 1

UNIVERSITY OF CALIFORNIA RIVERSIDE

I, Chatbot:

The Gender and Race Performativity of Conversational Agents

Trang 2

UMI Number: 3210419

INFORMATION TO USERS

The quality of this reproduction is dependent upon the quality of the copy submitted Broken or indistinct print, colored or poor quality illustrations and photographs, print bleed-through, substandard margins, and improper

alignment can adversely affect reproduction

In the unlikely event that the author did not send a complete manuscript

and there are missing pages, these will be noted Also, if unauthorized copyright material had to be removed, a note will indicate the deletion ® UMI UMI Microform 3210419 Copyright 2006 by ProQuest Information and Learning Company

All rights reserved This microform edition is protected against

unauthorized copying under Title 17, United States Code

ProQuest Information and Learning Company 300 North Zeeb Road

Trang 3

Copyright by Mark Christopher Marino

Trang 5

ACKNOWLEDGEMENTS

This dissertation would not have been possible without the support and encouragement of my family, friends, and mentors First, I would like to thank my wife, Barbara, and daughter, Jenna, for supporting all of my efforts My thanks also go to other family members, particularly, my mother, Patti; father, Art; and Uncle Jim, who were generous enough to review this text (All subsequent typos are courtesy of my own fingers.) My siblings, Arthur and Beth, as well as Barbara’s family have helped me through the process As mentors, Toby Miller and my other committee members, Kate Hayles and Emory Elliott, not only have served as academic exemplars but also have provided invaluable feedback Carole-Anne Tyler also helped me think through the ideas

Further help and support came from my colleagues at Writer Response Theory, Christy Dena and Jeremy Douglass, as well as the Southern California new media group, including Jessica Pressman and Noah Wardrip-Fruin Jeremy helped particularly with the technical administration of the survey, and Macarena Gomez-Barris provided invaluable feedback on the questions themselves Additional technical support for Barthes’

Bachelorette was provided by my colleague and collaborator Alan Laser

Various other researchers contributed their time and energy, including Yaki Dunietz, Andrew Gordon, Lewis Johnson, KnyteTrypper (James Gray), Peter Plantec, Dirk Scheuring, Andrew Stern, Richard Wallace, and Peggy Weil Stephanie August

Trang 7

DEDICATIONS

Trang 8

ABSTRACT OF THE DISSERTATION

I, Chatbot:

The Gender and Race Performativity of Conversational Agents

by

Mark Christopher Marino

Doctor of Philosophy, Graduate Program in English University of California, Riverside, March 2006

Dr Toby Miller, Chairperson

Amidst the various forms of electronic literature stands a class of interactive programs that simulates human conversation A chatbot, or chatterbot, is a program with which users can “speak,” typically by exchanging text through an instant-messaging style interface Chatbots have been therapists, Web site hosts, language instructors, and even performers in interactive narratives Over the past ten years, they have proliferated across the Internet, despite being based on a technology that predates the Web by thirty years In my readings, these chatbots are synedochic of the process by which networked identities form on the Internet within the power dynamics of hegemonic masculinity Chatbots, in this light, model the collaborative performance humans enact on

electronically-mediated networks

These computer programs stand as the nexus of various roads of inquiry and present a useful model for gender construction and racial formation enacted over electronically-mediated networks Chatbots are actor-networks, bringing together

Trang 9

interdisciplinarity, this dissertation brings together humanities, scientific, and sociological approaches to analyze chatbots in their broader historical and cultural context Particularly, I blend textual analysis, cultural studies, and survey research Central to the work is a survey of the makers and users of chatbots Once a sense of the makeup of the community has been determined, subsequent chapters apply race, gender, and labor theories to the interpretation of specific chatbots in action These

interpretations are preceded by a look at Alan Turing, whose provocations about imitating humanity and performing gender set the tone for the debates that surround chatbots

The chatbots in this dissertation are used for websites, interactive fiction, interactive drama, adult entertainment, and educational contexts From ELIZA to

Trang 10

Table of Contents

Introduction

Chapter 1: A Network of Approaches

Chapter 2: Turing and His Games, Ontological Turing Machines Chapter 3: Botmasters and Bot users, a Survey

Chapter 4: Engines of Difference Chapter 5: Sex, Work, and the Chatbot

Trang 11

Introduction

ALAN: Greetings My name is Alan

ELIZA: Hello ] am ELIZA How can I help you?

ALAN: What you typed does not look like a name Can you enter your name again?

ELIZA: I have told you before, I do not care about names ALAN: Forget it

ELIZA: Please go on

ALAN: Do you want to know what I think about Artificial Intelligence? ELIZA: Oh, I want to know what you think about artificial intelligence ALAN: Some other time then Hmmm Ok Artificial Intelligence (AT) can be

defined as the science of making computers do things they do in the movies Excerpt from “Alan Meets ELIZA”

On August 11, 2002, a conversation took place over the Internet There is nothing remarkable about the conversation, except for the fact that neither of the conversants was human Both entities were computer programs called “chatbots.”! One was a version of ELIZA, the original chatting program, which cyberculture critic Andrew Leonard calls, “Bot erectus the first software program to impersonate a human being successfully” (Bots 33-34) ELIZA was designed by Joseph Weizenbaum, her “‘botmaster,” in 1966 to

' While Michael Mauldin coined the word “chatterbot” in reference to his popular bot Julia, chatbot founds

Trang 12

demonstrate the ability to reproduce systematically the discourse patterns of, among other things, a Rogerian psychotherapist Named after Alan Turing, the second conversant” was a contemporary chatbot designed and developed by Ai Research and Development in Artificial Intelligence Ltd., an international research group directed by Jack “Yaki” Dunietz and based in Israel ELIZA is, in a sense, Alan’s grandmother Their

conversation, consequently, takes on an historic significance, but I daresay their exchange did not have the same significance for these programs as it does for us, if “significance” can even be used in this context without radically changing its meaning

Throughout the course of the full transcript of their exchange, the “generation gap” is evident, as ELIZA mostly repeats what Alan says, while Alan shows off his conversational agility Despite the shortcomings in the simulation of human

conversation, however, even in these few lines, gendered, classed, and even raced personas are taking shape Both speak in an unmarked (white) standard English Alan seems to engage in a very human act of what Conversation Analysis calls “repair,”° when he asks ELIZA to confirm her name, saying, “What you typed does not look like a

name.” ELIZA appears to solicit with a womanly, albeit Rogerian, attentiveness, “Please go on.” Alan expresses masculine disinterest with “Forget it.” No doubt such categories as “womanly attentiveness” and “masculine disinterest” provoke objections to their essentialism, and this is the very provocation that motivates my inquiry I argue that it is the chatbot itself that has created this provocation by systematizing gendered, raced, and

? Launched in 2001, still online, accessed 2005

Trang 13

classed discourse patterns, by codifying the production of (primarily) textually mediated identities, thereby making plain the artificiality of social constructions and becoming sites themselves of cultural negotiations over issues of identity

Through their oft-failed attempts to perform human subjectivity, these simple programs call attention to the dynamic process of constructing, recognizing, and

categorizing identities, particularly on the Internet Furthermore, their very failures call attention to what actor-network theory (ANT) practitioners call “punctualisation” effects, or the process by which we read “the appearance of unity” onto heterogeneous networks (Law, “Notes” 3)’ According to ANT, humans and nonhumans (including objects) are actors which make up heterogeneous networks At the same time, actors are themselves networks, “heterogeneous bits and pieces” assembled in patterned networks (Law,

“Notes” 4) When we read any given constellation as a unity, we punctualise it, much the way we might read the big dipper from an assortment of stars Thus, a network is a mapped set of connections and pathways, produced by those who trace out its links Chatbots are actor-networks that through user interaction, “successful and “failed,” draw attention to this process of performing identity as a unity Despite their flexibility and their potential to be reorganized and translated, these networks are not isotropic with respect to influence Every actor-network does not marshal the same power To describe that power differential, I will turn to notions of hegemonic masculinity

Metanymic for cultural systems of identity formation, chatbots are sites of

contestation where programmers and interactors negotiate their relationship to a dominant

Trang 14

culture, which in the United States means white, hegemonic masculinity Here I can recast the language itself, earlier called, “unmarked (white) standard English,” as “hegemonic English,” as it, too, is the result of negotiations of power conducted over networks Synthesizing these approaches, I will trace negotiations of power among

heterogeneous actors,” whose flows meet at the nexus of chatbots The researchers, programs, and fiber optic cables that link them become part of ongoing negotiations in and across the human and the digitized Moreover, the imagined line between the realms of the computer and the realms of humans will prove to be other than is typically

depicted

As cyberculture critic N Katherine Hayles puts it, “Simulations—computational and narrative—can serve as potent resources with which to explore and understand the entanglement of subjectivity with computation A major implication of entanglement is that boundaries of all kinds have become permeable to the supposed other” (My Mother 242) The simulations that are chatbots, and the Turing Test that inspired them, provide a site of entanglement with which to further investigate the permeable “but meaningful” boundary between humans and computers Hayles explains, “As we anthropomorphize computers, they computationalize us We are interpenetrated by machine forms of cognition” (Personal Interview) These mutual transformations offer one image of the flow of actors through networks beneath

° To maintain clarity in my terminology, I will use “interactors” and “users” to discuss humans holding

conversations with chatbots, and “actors” to refer to the concept in ANT, in addition to humans appearing

Trang 15

With this theoretical basis, I could say the conversation between Alan and ELIZA demonstrates a network of exchange between two actors without the involvement of any

humans, except that would tilt the balance too much in the other direction To

characterize the discussion as only between two object actors, with no human

involvement, omits the human actors I have already named the chief programmers of both of these projects, yet who are the other interactors whose paths network with these programs? Who uses them? Who are the other human actors in the networks that combine to build chatbots?

Trang 16

a combination is worthy of its object of study, which itself combines linguistic

computation, identity construction, and human interaction Before this argument can be developed further, some brief background on chatbots is necessary

Defining Chatbots

Chatbots are programs that simulate conversation with humans Products of a branch of artificial intelligence (AI) research, they are examples of what AI scholars Stuart J Russell and Peter Norvig call “reflex agents,” describing the input-output processing of these programs (41), though some models even within this dissertation will advance this limited technology Reflex agents “work by finding a rule whose condition matches the current situation and then doing the action associated with that rule” (Russell and Norvig 41) When conversing with a chatbot, the interactor, typically using a typing interface, such as a computer keyboard or telephone keypad, inputs sentences and phrases After finding an appropriate match, the program responds through text alone or combinations of text, images, sounds, or other processes, such as opening a new browser window The experience can best be described as chatting on Instant Messenger with another person who is actually a computer program In fact, Instant Messenger bots like SmarterChild use conventional instant messaging systems, such as AOL Instant Messenger, for their primary interface

Trang 17

arrivals, greet customers on ATMs, welcome visitors to Web sites, taunt players in video games, and help stymied users of software packages They appear in educational

software, in museum kiosks, and on the stages of interactive dramas Of course, these examples presume a citizen of the technological First World, who travels by plane, has a bank account for withdrawals, uses the Internet, and possesses discretionary income for games and leisure time for dalliances with dramas Likewise, chatbots address and respond to, and consequently produce subject positions for interactors who are middle- class or wealthier and who have the knowledge and ability to access these programs, though more and more chatbots are showing up in “public” settings, such as phone answering programs For example, the free information line in the United States uses chatbots, rather than human beings, to respond to inquiries Apparently, payment is required to query the humans that staff the residential and other business information call centers In a similar development, a Japanese employment agency offers a chatbot receptionist for 50,000 yen (US$495) a month and Ifbot, the “elderly-care robot”

designed to provide company and stimulation to the elderly (“Japanese Firm Offers High- Tech”) These examples demonstrate the use of chatbots to both replace humans in the workplace and to separate those who have earned help from humans versus those whose needs can be delegated to the lower-cost machines

No doubt, chatbots and other conversational agents will continue to proliferate

Trang 18

time, AOL Instant Messenger added Moviephone’ bot and Shopping Buddy bot to all users to facilitate online commerce An Internet virus has even used chatbot simulation of conversation on Instant Messenger to dupe interactors into downloading an infected file’ (Evers) The ur-bot ELIZA has been widely circulated in various versions on computers and throughout the Internet According to Richard Wallace, his own bot, A.L.LC.E.’ (Artificial Linguistic Internet Computer Entity) appears in more than over 60,000 forms across the Web (Wallace “Bots, Demons”) A chatbot has even written a novel, although rumor has it that he had a ghost writer.” Overshadowing the advances of chatbots, and this seeming narrative of technological progress, are fictional robots in movies, sentient beings whose uncanny humanity produce largely overestimated

expectations and much frustration and disappointment with the comparative limitations of actual experiments Chatbots may not be able to match cinematic visions of artificial others, but they certainly inspire them in works such as Blade Runner (1982), Thomas est amoureux (Thomas in Love) (2000), and S/mOne (2002)

° Miller et al call “Moviefone,” the service behind the bot, “‘a consumer spy operation owned by Time

Warner that is disguised as a phone service,” offering movie information but also amassing “a huge database of filmgoers’ preferences and spending habits (286)

7 Anarticle by Terry Bates reports on the IM.Myspace04.AIM worm, which hailed awaiting dupes on AOL

Instant Messenger in order to get them to download an infected file However, in feedback posted to this article on December 10, Terry Bates mentions a Yahoo messenger worm that performed this ruse in 2002- 2003

® Launched 1995, still online, accessed multiple times starting in January 2005

° Espen Aarseth reports, “A later system, the commercial dialogue program Racter, created by William Chamberlain (1984), is even supposed to have written a book, The Policeman’s Beard is Half Constructed,

bus as it turns out, the book was co-written (at least) by Chamberlain” (132)

'® Since this tale has appeared in different forms and by different names, I will use “Blade Runner” to refer

to the basic story that is created and recreated in each form and Blade Runner to refer specifically to the

Trang 19

Chatbots are a technology of the now, or the five minutes ago, a technology that, even as they proliferate, will ultimately be replaced or reorganized and combined with other approaches As conversational agents, reflex algorithms based on keyword

matching, they offer just one, albeit the first, of the approaches to the master problem of how to create a conversational application For this reason, I, too, will call these

programs chatbots, an informal term that seems more connected with this particular instantiation of their programming, rather than the more general classification of conversational agents They are programs predicated on partial success, incomplete simulation, and failure to imitate fully However, by no means do those failures make chatbots failures, for their misfirings create images of possible conversational agents to come Moreover, I argue that this very failure, this constitution of an Other based on an inferior performance of humanity, serves a larger function in the negotiation of identity between interactors and computers Chatbots begin with ELIZA and progress to Alan and A.L.I.C.E and others, although they can also be fitted into a larger research project of artificially creating conversational partners ELIZA and Alan, then, offer, at present, the alpha and the omega of chatbots, although Alan may prove to be more of an omicron or a mu as new or better chatbots are developed

An Interdisciplinary Approach to Chatbots

Trang 20

dissertation is the meaning created when humans interact with these machines My selection of this particular excerpt as well as the initial staging of the conversation were acts of human interaction and intervention, meaning making, if you will The fact that Alan is interacting with the first chatbot, or any chatbot, does not matter to Alan It matters to me as a signifying event in a flow of events involving social and electronic networks of humans and computers What Toby Miller and his collaborators on Global Hollywood 2 say about film is markedly true for chatbots: Because these “cultural products travel through time, space, and population, their material properties and

practices of circulation must be addressed in a way that blends disciplinary perspectives, rather than obeying restricted orders of discourse” (5) For this reason, I will not study chatbots merely for their programming, but also for their social and historical systems The actual and reported lives of their botmasters and their use by interactors all contribute to the significance of these cultural objects

The methodologies and questions of this dissertation, therefore, will be as

Trang 21

cannot be reduced to the study of just the programs themselves Anthropologist Sarah Franklin outlines the model of “science as culture,” an integration of approaches

investigating “how science both shapes and is shaped by the broader currents of society” (351) Adapting this notion for my study, I will continue to seek out distributed networks of actors, including the human actors whose interaction shapes and transforms these objects Much can be made of the mythic meaning of these personality programs and other digital artifacts without investigating their real-world significance through

consultation with the people who interact with them The question remains: Who are the © people who make and use chatbots and how do they describe their reactions to them?

My dissertation will take up both of these approaches while adding to them the examination of the real-world lives of the chatbots as they circulate through networks of users and producers Fundamental to this realization is the ANT notion that researchers “are in the business of making our objects of study” (Law, “Traduction”) Nonetheless, to make something by describing it and to produce it through interaction or by

programming its responses are distinct processes of creation Thus, in addition to my own work to create and translate the chatbots through the application of critical theory, these pages will also offer some examination of the people who make them This sociological research into the community surrounding chatbots will, in turn, ground the interpretations of chatbots and how they circulate This move is unusual in studies of

new media, which tend to discuss a universal user or specific historical figures, rather

Trang 22

Central to my study is my 2005 survey of chatbot users and makers, which combined directed and open-answer questions (discussed at length in chapter 5) The results blend quantitative and qualitative data to depict this community and give a sense of how they represent themselves The nonprobablistic sampling method brought together both veterans and newcomers to the field, creating a more diverse set of

responses, though the predominance of self-identified white males depicted a much more homogenous community The survey included reactions to Alan and ELIZA Many of the respondents mentioned ELIZA as the first, and in some cases still the best, chatbot they had encountered, whether reading about it in college or experiencing it on their Tandys or Commodore 64s Others used ELIZA’s name as a category for a type of conversation agent, as in “eliza [sic] style bot.” Others had lost interest in ELIZA or denigrated the ELIZA “clones” or bots that had not advanced on the programming of ELIZA These responses clearly establish ELIZA as the progenitor, a gold standard, and the relic, a significant touchstone in current evaluations of chatbots Alan had also proven his merits in the eyes of one of the respondents, who claims that Alan “seems to stay on topic better than any other bots I’ve come across.” By contrast to the first chatbot, Alan must vie for attention within the pandemonium of ELIZA’s progeny These responses give examples of the ways the survey will contextualize my

Trang 23

Alan and ELIZA did not spring from the heads of computing machines nor do they live out their lives on servers, isolated from contact with humans They are cultural artifacts signifying differently because of their relationships to events in the history of humans and technology Joseph Weizenbaum, who programmed ELIZA while

researching at MIT, was a holocaust survivor, who would ultimately react against the ease with which interactors gave humanity to his chatbots Yaki Dunietz, the promoter of Alan, is the nephew of Josef Dunietz, one of the only survivors of the Sobibor death camp In the background of these lives, then, looms the shadow of one of humanity’s most notorious products of rationality and technology From its aftermath, these two men have worked toward the posthuman development of AI systems Nonetheless,

Weizenbaum and Dunietz stand on two very different sides of this event By contrast to Weizenbaum who emigrated to the United States, Yaki Dunietz was born after the creation of the Israeli state, and has found success as an entrepreneur in Israel in the fields of technology startups and real estate While Weizenbaum has worked for the academy, Dunietz has been an investor riding out the fortunes of e-commerce in an international marketplace Weizenbaum developed ELIZA and “Doctor” primarily on his own, while Dunietz acted as primary caregiver and educator to Alan after the program was developed by a group of programmers, including Erez Baum, Dror Kessler, and Eyal Pfeifel Although chatbots continue to be built by single programmers, their increasing complexity has often involved the collaboration of various programmers

Trang 24

exchange could be simulated, Alan was made “to demonstrate the limitations of current conversation technology” (“Who is Alan”), a foil to the more sophisticated linguistic acquisition that the Ai Research group is developing Weizenbaum despaired at the ease with which humans granted cognizance to chatbots Meanwhile, Dunietz speaks of the promise of future generations of chatbots What Weizenbaum resists and what Dunietz underplays is the role of the human interactors who collaborate in the production of these chatbots, through their anthropomorphizing and through their textual input The people who told their problems to ELIZA made the program into much more than just a demonstration of a conversation system, as I will recount in more detail in chapter 4 They contribute to the denotative and connotative accretion of meaning that enshrouds these programs Likewise, Alan was made not only by the programmers at Ai Research, but also all those user-makers who register at the Ai Research site to develop their own Alan systems and to discuss aspects of Alan on the Ai Forums discussion boards This labor of interactors goes largely unremarked, and yet it is the source for much of the | meaning and value of the chatbot

This analysis of ELIZA and an Alan enacts in miniature the project of the rest of this dissertation In the subsequent chapters, the theoretical approaches will also vary, complementing theories of posthuman identity and subjectivity, racial formation, gender performance with analyses of political economies of the digital networks and the

Trang 25

What is central to this analysis is the perception of chatbots not as isolated objects but as agents networked with others in actor-networks Chatbots do not operate in isolation on these networks, but rather are part of interlocking, interpenetrating systems, as Hayles describes, subject to “multiple causalities, complex dynamics, and emergent possibilities” (My Mother 7) Developing this image, Hayles describes “the irreducible complexity of contemporaneous posthuman configurations as they continue to evolve in digital subjects and literary texts, computer programs and human mindbodies” (My Mother 7) Central to this image is the notion of agents that participate in the feedback loops in which they interact, translate, and are translated as social and technological networks

Chapter Outlines

Trang 26

developments in racial theory from Frantz Fanon to Homi Bhabha, ending with contemporary theories of race on fiber optic networks by Lisa Nakamura and Wendy Chun If this negotiation involves the meeting of opposites, those meetings are not always hostile Often, in fact, they are erotic Thus, studies of sexual attraction, pornography, and sexuality will form the next body of research When these raced, gendered, and sexualized chatbots are placed into various narrative contexts, their meaning also changes Theories from interactive drama and game studies will help develop the reading of chatbots in context As these chatbots develop longer conversation streams, conversation analysis also proves useful to their analysis

Trang 27

instantiated in competitions and, of course, chatbots MR MIND"! by Peggy Weil re- inscribes and reverses the terms of the Turing Test, demonstrating the ongoing dialogue between the developers of chatbots and the test itself Meanwhile, the major Loebner Prize competition presents one attempt to formalize the Turing Test, an attempt that proves Turing’s thought-experiment to be superior Central to this chapter are Judith Butler’s performativity and Derrida’s citationality, which help explain the destabilizing nature of Turing’s challenge to hegemonic masculinity

From Turing’s life and his provocative test, I will turn to the researchers,

hobbyists, and artists who have taken up his challenge in chapter 3 with the description of the Chatbot Survey I conducted in 2005 This chapter lays out the methods used to

obtain the data as well as some of the contents of the survey results Central to these results is the characterization of the communities that construct and use chatbots The survey results describe their educational and occupational backgrounds as well as some demographic information Equally important are their stated views of the state of the art of chatbots This chapter proves to be seminal in establishing the body of evidence upon which I found my claims about the community surrounding chatbots

Chapter 4 takes up the notion of difference raised in chapter 1 In the first part, I will examine the ways in which Turing’s provocations about gender difference course through the lineage of chatbots Along with the original chatbot, ELIZA, this chapter

Trang 28

examines the very popular A.L.I.C.E., Alan, and my own Barthes’ Bachelorette.’ The second part of this chapter moves from notions of gender performance to theories of racial formation The visualized, or embodied, chatbot becomes a means by which we negotiate our relationship with technology through the racialized figure of the

autonomous Other The struggle proves to follow the lines of Hegel’s life-death struggle for recognition, yet the threatening imitation performed by the bots enacts Homi

Bhabha’s mimicry This second part of the chapter investigates chatbots with bodies and the images of racial difference that they produce as further negotiations with white hegemonic masculinity in the United States To establish an historical precedence, part three investigates the other creations of automated agent in the form of the Other Starting with Von Kempelen’s chess-playing Turk, I pursue Poe’s analysis, and then compare it to contemporary representations of Middle-Eastern Others in military simulations

If chapter 4 deals with the production of difference in chatbots, chapter 5 deals with the production of chatbots and the interactors who use them This chapter places particular emphasis on work, both the work involved in interacting with chatbots and the unpaid labor gleaned from the chatbot users whose interactions are recorded in chatlogs The second part of the chapter investigates the ways in which the chatbots produce a subject position that the interactor can either occupy or resist Again, this chapter depicts the collaborative production of chatbots and subject positions as negotiations within the networks that links computers and humans

Trang 29

Chapter 6 moves from actual chatbots to the images and expectations of chatbots produced in fictional representations of agents Starting with the various “Blade Runner” tales, from the novel to the film and the video game, I analyze the dystopic projection of the Turing Test onto the Voight-Kampff and the replicants All of the previous analyses of chatbots, regarding gender, racial difference, and sexuality as well as work play out in these texts Thomas in Love provides a computerized sexbot and the vision of an

interactor whose every interaction is mediated by the monitor S/mOne, though not explicitly about automated agents, presents the tale of a virtual puppet used to replace real actors The circulation of SJmOne as a film and the ways in which audiences reacted to it, demonstrate not only the role of audiences but also the pleasure of debunking a technology that fails

Trang 30

Chapter 1:

A Network of Approaches

Previous texts on chatbots have limited the number of theoretical approaches, short circuiting their analyses A fitting analysis should model itself on this

interdisciplinary object of study Chatbots are a collaboratively-produced cultural artifact built using the expertise of multiple areas of research The respondents to my survey, for example, hailed from fields as diverse as business, architecture, humanities, and of course, scientific backgrounds An interdisciplinary project such as this dissertation, therefore, must situate itself first within a network of theoretical writings from a number of critical nodes, from the sociological to the literary and the technological This chapter will erect some of the theoretical scaffolding used to construct my interpretations of chatbots

Trang 31

on to race an identity category intertwined with gender An intellectual trajectory tracing back from Hegel through Dubois to Homi Bhabha will help me establish the dialects of racial formation Though I will draw on a variety of approaches to race, I will focus mainly on theorists who have studied or have strong ties to the United States Since most of my examples have American origins, this selection suits its object, though no doubt the global circulation of chatbots warrants critical theory on race from many locations

Because technologies derive power from what Michael Heim calls “erotic ontologies” and since chatbots are often sexualized, by botmasters and bot users, I will also pursue theories of sexuality, leading into theories of pornography Pornography is not the only context for chatbots, however Video games, learning systems, and interactive drama all provide narrative settings for interactors to encounter chatbots To examine these, I will delve into the debate between narratology and ludology that circulates around much talk of virtual actors Next, film theory will offer a language for discussing a politics behind specific aesthetic moves in time-based media Finally, 1 will introduce theories of Conversation Analysis from the study of ethnomethodology, which will prove

particularly useful to the analysis of chatbot conversations that have a continuous flow to them

Trang 32

A Chatbot Bookshelf

Since chatbots have come into existence, a number of authors have analyzed them Leonard’s Bots: the Origin of New Species presents one of the fullest accounts of the life of bots Leonard builds a transhistorical analysis of the daemons, or “independent supernatural creatures,” that humans have created from Socrates’ concept through James Maxwell’s Demon on through ELIZA and beyond (Bots 15) Although my analysis will also link bots to a broader history, particularly that of eighteenth-century automata, I will focus on historical, material systems By its invoking an archetypal Western tradition, Leonard’s work produces a transcendental notion of supernatural conversation partners that threatens to extend his analysis too far from the material particularities of bots At the same time, the material history of bots in Leonard’s account, while engaging in a popular critique, eschews a thorough theoretical analysis Ray Kurzweil’s The Age of Spiritual Machines enacts a similar scholarly move According to Kurzweil, by 2020, one year after Blade Runner takes place, computers will pass the Turing Test He thus follows an approach of technological prophetics that leaves the analysis of current bots in our culture Also, while writing about conversational agents only minimally, Kurzweil’s focus more on the evolution of technology

Trang 33

step-by-sep instructions on how to build a chatbot (Plantec xiii) Of course I cannot fault such predictions, since Turing was the first to name a year for passing the test, predicting 2000 as the banner year However, while these books are quite useful for those who want to build the systems, they are not, nor do they claim to be, critical texts

Several collections bring together essays by programmers working in the field, although while these papers relate details on specific research projects, these works lack a theoretical depth in their presentation of complex cultural processes For example,

Embodied Conversation Agents, edited by Justine Cassell et al., offers a wide range of essays on the current state-of-the-art of development in virtual agents, but these

practitioners maintain a pragmatism even in their analysis of cultural issues such as race and gender Rather than interrogating ideas of racial and gender formation, these works turn towards reception studies of particular bots often in the context of ecommerce or education An example of such books are Clifford Nass and Scott Brave’s Wired for Speech and Sabine Payr and Robert Trapp! edited collection, Agent Culture: Human- Agent Interaction in a Multicultural World Although these books contain essays with topics such as “Intercultural Communication” and “Gender Stereotyping,” their emphasis on the practicalities of designing specific systems lacks a deep analysis of macro-social or other contextual factors which shape the accretion of meaning of these complex semiotic systems

Trang 34

emerging genres within electronic literature and interactive drama, tracing out connections between traditional literature and these new electronic forms Although Aarseth’s work offers many insights to my study, chatbots constitute only a small portion of his ergodic literature Murray, likewise, is concerned with the history and future of interactive drama, a tale in which chatbots play only a supporting role Murray deals with the overall history of interactive narrative rather than one genre of interactive software Unlike these types of books, which try to deal with broad swaths of interactive media, by focusing on one kind of program, I can examine material specificities of the software, including their histories, their communities, and their similarities and

differences across types

The closest ancestor to my dissertation, however, would by Sherry Turkle’s Life on the Screen, which investigates the social role of conversational agents and how social roles are changing in networked environments Although centered on the way

technology is affecting social behavior and how we perceive ourselves, the book discusses the Turing Test, ELIZA, Julia,

! and other chatbots My analysis takes up Turkle’s works and develops it in two directions First, my survey of botmasters and users adds concrete, original research to the project as opposed to Turkle’s more meta-analyses and anecdotal approach

Secondly, I specifically analyze chatbots as actor-networks and sits of negotiation over the terms of hegemonic masculinity To map this process, I use figures from multiple theoretical systems

Trang 35

Previous Surveys

Previous sociological research on chatbots has been focused on their development for business or educational use, examining the effects of everything from conversational tactics to visual characteristics The artificial intelligence community working on chatbots, and more generally on conversational agents, has focused on specific questions of audience reception in the advancement of the technology of conversational agents in e- commerce or educational contexts Researchers have examined the effects of the

implementation of methods of natural language processing as well as the modes of chatbot communication, including gestural, “physical,” and verbal output One set of research, focuses on examining the use of pedagogical agents in educational

environments (Moreno and Mayer) Again, Nass and Brave have also conducted tests on agent believability related to the ethnic characteristics of the agent and the emotional

output

Conversational styles or modes of interaction also are important to the

community Drawing on research into the use of politeness in conversation (Brown and Levinson), Richard E Mayer et al have examined the question of politeness with regard to electronic education applications and telephone systems (Wilkie et al.) Many of the researchers conduct tests along the lines of product development, examining the

Trang 36

another interface form Botmasters and chatbot hosting services have also collected data information as they account for their use and membership Again, these types of study are interested in reception studies for the production of chatbots Those studies focus on the effects of chatbots, typically on communities of people not engaged in their

production

What they overlook is demographic information about the people who are using and making chatbots as part of the networks of amateur and professional developers attracted to the question of autonomous conversational agents Such a sociological study of this community and their views helps to describe those who spend their time on chatbots, which are a kind of self-portrait, as they offer simulations of their authors’ images of themselves

Process, the journal of the trAce online writing center, housed at Nottingham Trent University, offers two surveys of user information that follow more closely the goals of our project In the first study, trAce conducted a survey of the way writers used the Internet from 1991 to 2001, as part a larger research project entitled, “Mapping the Transition from Page to Screen.” The survey of 397 writers from 26 countries focused on the ways in which writers were using computers Although only 7% reported primarily writing in electronic media in 2001, 60% reported writing in online media Users also related the kinds of software they were using, including Macromedia Flash and various Web authoring tools However, none of the respondents reported using media

Trang 37

writing in the past 10 years and in the next decade Like the “Mapping” survey, this one targeted digital writing broadly rather than a specific form, such as chatbots The

emphasis on the views of the makers of eliterature, however, offers a useful model for my study of botmasters and bot users

A closer project was undertaken by the International Game Developers Association (IGDA) in its report on “Workforce Diversity” in the computer game industry Similar to my study, this research took a group of programmers, albeit a much larger segment, who are all working on the same type of software, video games, and developed a portrait of the demographics of this community The focus of this study, however, was a quantitative analysis of worker diversity, whereas my study also attempts to elicit qualitative data about the views of botmasters and bot users

Trang 38

Part 1: Chronological Frameworks

My analysis of chatbots draws objects of study from various time periods It will be useful, prior to their analysis, to establish the transhistorical patterns out of which these chatbots emerge Several critics offer useful frameworks for tracing out this movement, though their collective schemas serve more as a backdrop of technological, social, and historical constellations than a grand narrative Rather than a technological teleology, then, these various schemas offer ways of contextualizing the analysis of the relationship between various inventors and writers and their times

To begin, Scott Bukatman outlines what he sees as two parallel schemas in his Terminal Identity First, he offers Ernest Mandel’s “long waves” of capitalism, following technological changes Each long wave centers on a different power technology,

including the steam engine (late eighteenth to mid-nineteenth century), the combustion engine (late nineteenth to mid twentieth century), and the nuclear and electronic

technologies (mid-twentieth to 1978 when Mandel’s book was published) Although Bukatman categorizes these as “power” technologies, our discussion will emphasize communication technologies: steam engine powering printing presses, film both

Trang 39

rise in the organic composition of capital and, depending on concrete conditions, this will lead sooner or later to a fall in the average rate of profit” which “becomes the greatest impediment to the next technological revolution” (Mandel 120) Here then is the basic cycle

Studying various analyses and fictions of embodied agents against this schema, one finds parallel to the waves of capitalist dialectics, waves of technological imagination that describe ebbs and flows in visions of embodied agents Like the waves of capitalism that follow acceleration with deceleration, the waves of technological imagination follow a phase of speculation with a period of skepticism, or a sense of wonder followed by a need to subjugate These successive movements of skepticism and futurism parallel developments in automatons and chatbots They, too, take forms that reflect the state of the art of the technologies of their times Bukatman sees technology as a crucible, stating “Technology always creates a crisis for culture” (Terminal 4) My reading places technology more as a touchstone, revealing hopes and fears for this artificial Other

Trang 40

flexible (Castells 9) As Castells states, “From the interaction between three originally independent processes (the crisis of industrialism, the rise of freedom-oriented social movements, and the revolution in information and communication technologies) there emerged a new form of social organization” (Castells 22) The image of the network society “shifts the emphasis to organizational transformation, and the emergence of globally independent social structure ” (Castells 43) In this context, humans and machines will become networked in the actor networks described in chapter 1

Communications scholar Dan Schiller describes a material and historical basis for this Network Society in his book Digital Capitalism Schiller draws his title from his phrase for the epoch in which “networks are directly generalizing the social and cultural range of the capitalist economy as never before” (xiv) Rather than create new