Talking to Computers: What is 'Real' Conversation?
Weizenbaum & Delusional Conversation
When Joseph Weizenbaum saw that people responded to his "artificial intelligence" program ELIZA as if it were human, he was shocked and dismayed. The program was, for him, a caricature of human conversation that if anything underlined how crude and inhuman the computer still was. And yet Weizenbaum describes how people became intensely emotionally invested in their exchanges with it, with even psychiatrists suggesting that the program might one day replace face-to-face consultations (Weizenbaum, 1976 p.6). What had happened to the world, he wondered, that we could even conceive of a computer replacing a real, live person? His distaste is typical of what Sherry Turkle later described as the period of "romantic reaction to the notion of the almost-alive machine", during which "the seeming animation of the machine challenged the boundaries between persons and things" (Turkle, 1997 p.84). Weizenbaum asked,
"What must a psychiatrist who makes such a suggestion think he is doing while treating a patient that he can view the simplest mechanical parody of a single interviewing technique as having captured anything of the essence of a human encounter? ...What can the psychiatrist's image of his patient be when he sees himself, as therapist, not as an engaged human being acting as a healer, but as an information processor...?" (Weizenbaum p.6).
I'll return to this unanswered question later. For Weizenbaum, the answer seemed clear enough. That "people were conversing with the computer as if it were a person who could be appropriately and usefully addressed in intimate terms" (ibid p.7) was a symptom of a grave cultural shift, during which the human being had come to be conceived of as a machine - and who could thus be unproblematically replaced by one. Weizenbaum maintained an emphatic distinction between genuine human contact, and contact with a machine, describing those who enjoyed psuedo-human conversation with ELIZA as "delusional" (ibid p.7) and computer-loving programmers as "pathological" (ibid p.121).
Weizenbaum's insistence on the unsuitability of computers as conversational partners, and the primacy of "those direct experiences that formed the basis for, and indeed constituted, [pre-industrial] reality" (ibid p.25) echoes the sharp division between "reality" and "spectacle" which the Situationists were drawing at around the same time. There are echoes of Situationist sentiment in Weizenbaum's conception of artificial intelligence as a dangerous fake which lures us away from the 'real thing'. A conversation with ELIZA isn't really a conversation, but it looks and feels (superficially at least) the same. Weizenbaum attributed to a reductionist scientific worldview what the Situationists did to capitalism; the result, for both, is a set of "bogus miracles" (Vaneigem,1983 p.26) usurping genuine relationships.
Nichols' Cybernetic Simulations
Writing around 20 years later, Bill Nichols touched on ELIZA in his discussion of the social implications of cybernetic technologies. While sharing Weizenbaum's desire for a more communal and connected society (Nichols, p.629), he offered a more nuanced reflection on the role that our conversations with computers might play. Indeed, he seemed to take almost the opposite stance to Weizenbaum when, discussing Norbert Weiner's concept of the cyborg, he described "the new identity which, instead of seeing humans reduced to automata, sees simulacra which encompass the human elevanted to the organic" (ibid p.635). Weizenbaum's argument against artificial intelligence was that the very idea of it required the theoretical reduction of human thought to a set of computational procedures (Weizenbaum p.67). But in Nichols we have the inverse: "simulacra...elevated to the organic". Maybe Nichols' more hopeful view of cybernetic simulations - the power of which "prompts a redefinition of such fundamental terms as life and reality" (ibid p.635) - could be a more helpful way of trying to conceptualize what it means to become emotionally invested in conversing with our computers.
So what does Nichols mean by "cybernetic simulations"? Best & Kellner (1997), in their book The Postmodern Turn, devoted a chapter to a thoughtful exploration of the theoretical shift from the Situationists' "spectacle" - thoroughly grounded, like Weizenbaum, in a modernist insistence on a pregiven 'real' - to the idea of simulation, as proposed by Baudrillard (ibid p.96). For him, in our postmodern culture "simulation devours the real and… leaves behind nothing but commutating signs, self-referring simulacra that feign a relation to an obsolete real" (ibid p.101). So, our loving dialogue with machines is no longer a delusional imitation of 'real' conversation, but a simulation in its own right. It's interesting to note that although Best & Kellner acknowledged the importance of Baudrillard's contribution, challenging the Situationists' perhaps naive reference to an unblemished 'real' beyond the reach of the spectacle, they rejected the extremity of his conclusions, according to which all difference between signifier and signified collapses. They advocated instead a "qualified use" of his later theory (ibid, p.105), which seems to be shared by Nichols in his ambivalent assessment of the possibilities and pitfalls of cybernetic simulations (Nichols, p.632). Nichols rightly notes that precisely because the two-way communication of cybernetics simulates (rather than merely reproducing) human social exchange, it provokes us into asking what human social exchange actually is, or could be (ibid pp.629-30).
At this point Nichols goes on to discuss, in psychological terms, what actually goes on in the minds of those engaged with cybernetic simulations. He describes the way in which our high-tech situation "poses the simulation as an imaginary Other which serves as the measure of our own identity". It becomes clear that new technologies are not the only thing with which we engage in this way; indeed, this metaphor and its language are lifted directly from psychoanalysis' framing of the infant's relationship to its mother (Nichols p. 630). The experienced difference between a 'real' encounter and an imagined one is problematised by psychoanalysis and its observation, rephrased here by Nichols, that "our cognitive apparatus treats the real as though it consisted of those properties exhibited by simulacra. The real becomes simulation" (Nichols p. 635).
Discussing Cold War projects such as Star Wars, Nichols notes that "These simulacra of war... are fought with an imaginary enemy, in the Lacanian sense" (Nichols p.635). Yet, he emphasises that bullets fired by such systems will kill real people. The complexity of this situation is not accounted for by Weizenbaum and others' blunt insistence on a distinction between the real and the imaginary. Real people become blurred in our minds with simluations (Star Wars); nonexistent people are fabricated, becoming real in our imaginations (ELIZA). Sherry Turkle's research on internet chat confirms that in text-only dialogue, it can be difficult to distinguish real people from machines:
"I myself have made this kind of mistake several times, assuming that a person was a program when a character's responses seemed too automatic, too machine-line. And sometimes bots are mistaken for people. I have made this mistake too, fooled by a bot that flattered me by remembering my name or our last interaction. ...I found myself confronted with a double that could be a person or a program." (Turkle, p.16).
Conversation, Imagination, Projection
Weizenbaum asked derisively what a psychiatrist's "image of his patient" must be to think that a computer could replace her therapist. A plausible answer would be that she is a subjective being, capable of working out her subjectivity upon whichever projective surfaces are available. An answer, in fact, which could be quoted from Weizenbaum himself:
"It is within the intellectual and social world he himself creates that the individual prehearses and rehearses countless dramatic enactments of how the world might have been and what it might become. That world is the repository of his subjectivity. Therefore it is the stimluator of his consciousness and finally the constructor of the material world itself. It is this self-constructed world that the individual encounters as an apparently external force. But he contains it within himself; what confronts him is his own model of a universe, and, since he is part of it, his model of himself" (Weizenbaum, p.18).
Weizenbaum in fact had a nuanced, and rather beautiful, understanding of the complex psychological relationship between humans and the tools they create. But could this quote not apply equally to the way we relate, in psychoanalytic terms, with other humans? While reserving a priviledged place for unmediated human encounter, Weizenbaum perhaps overlooked the importance and possiblities of our dialogues with imaginary Others, which Nichols emphasised have very real manifestations in the physical world. And he perhaps overlooked, too, the role that fantasy and simulation play in all human relationships. In fact, as Turkle points out, psychotherapy is traditionally enabled in part by deliberately obscuring the therapist's subjectivity, in a style which resembles the natural conditions of computer chat:
"[online chatrooms] encourage projection and the development of transferences for some of the same reasons that a classical Freudian analytic situation does. Analysts sit behind their patients so they can become disembodied voices. Patients are given time to project onto the analyst thoughts and feelings from the past" (Turkle, p.207).
It is hard to argue that the patient's relation to her therapist in this setting is any less imaginary, or indeed "delusional", than it would be with ELIZA. Weizenbaum's caricature of the therapist imagining himself as an "information processor" discounts another explanation for the pleasure of simluated dialogue, which focusses instead on the client's experience. For Turkle, the simulation of dialogue with machines means that "people are confronted with the degree to which they construct relationships in their own minds" (ibid p.207). Thus, we learn as much about human subjectivity as we do about computer science from these 'imaginary' conversations.
Perhaps what computers help us to see, then, is that 'real' dialogue is a wider field than that allowed for by Weizenbaum. They illustrate, in their ubiquitous mediation of our encounters with each other and in their prominence in our imaginary worlds, the extent to which even 'genuine' human relationships are made up of imaginary projection, anticipation, memory, and other effectively isolated mental activities. Perhaps rather than replacing genuine dialogue with simulation, computers bring to our attention the simulation inherent in all of our relationships in the first place.
- Best, S. & Kellner, D. (1997) 'From the Society of the Spectacle to the Realm of Simulation: Debord, Baudrillard, and Postmodernity' in Best, S. & Kellner, D. The Postmodern Turn (New York/London: Guilford Press) pp.79-123.
- Nichols, B. (2003) 'The Work of Culture in the Age of Cybernetic Systems' in The New Media Reader by Wardrip-Fruin, N. & Montfort, N. (eds.) (Cambridge, Mass & London: MIT Press).
- Turkle, S. (1997) Life on the Screen: Identity in the Age of the Internet (London: Phoenix).
- Vaneigem, R. (1983) The Book of Pleasures (London: Pending Press).
- Weizenbaum, J. (1976). Computer Power and Human Reason: From Judgement to Calculation (New York: W.H. Freeman and Company).