Since their inception in the middle of the twentieth century, digital technologies developed from big machines accessed only by small groups of scientists and corporative experts to “personal” devices providing opportunities for interaction to large masses of users. As a wide range of hardware and software interfaces were introduced, allowing users to employ and engage with digital machines, the field of human–computer interaction (HCI) tackled the pragmatic and theoretical implications of this change. This chapter takes up the theories and practical toolbox developed in this context to ask new kinds of questions about the history of automata in the nineteenth century. As automata were offered to public spectacle and, to some extent, consumption, commentators discussed the reactions of the “users” and observers of these devices, creating a body of theoretical reflections that can be read as HCI ante litteram.¹ By considering cultural texts and artifacts that contributed to debates about “human–automata interaction,” I will mobilize later debates in HCI and artificial intelligence (AI) to reconsider the ways in which Victorians discussed and/or imagined how people reacted to automata exhibiting the appearance of intelligent behavior.

There is a wealth of research bringing into dialogue the history of nineteenth-century automata with histories of computing and AI. Historians of AI routinely point to automata as a precedent for the development of this field and its aspiration of creating a “thinking machine.”² Some have drawn parallels between clockwork writing automata and natural language-processing software that produce journalistic or literary texts, or between humanlike automata and robots.³ Amazon’s reclamation of the Mechanical Turk to name their microwork platform has also attracted attention with regard to the politics of labor of contemporary capitalism and the Western-centric orientation of both Victorian automata and contemporary AI.⁴ While these works provide a useful reference point for this chapter, their main focus has been to reconstruct the longer trajectory of the dreams of AI and artificial life. Less attention, however, has been given to the fact that automata also anticipated dynamics of interaction between humans and computers. At the same time, existing histories of automata have rarely focused on the specific problem of the interaction between automata and users. Rather than locating a prehistory of HCI within histories of automata, or reading the past with the lens of the present, my hope is to unveil how some issues that today fall under the label of HCI were envisioned, albeit through different concepts, frames, and contexts, before the invention of computers.

The difference between automata and digital computers is undisputable. Automata mechanized and reproduced movement, thereby providing the appearance of human- or animal-like behaviors. Computers instead produce numerical calculations, which are mobilized to execute complex tasks, including logical operations and other activities that are considered evidence of intelligence when carried out by humans. Automata are analog while modern computing is (predominantly) digital. Yet there is at least one crucial element of continuity between the two. Across at least two centuries, first automata and then computers stimulated and supported the dream of creating machines that imitate or reproduce human characteristics and skills. While automata and computers, in this sense, differ markedly in terms of the technologies used and their functioning, they are more similar in terms of how people have perceived, interpreted, and conceptualized them. The connection between automata and computers is to be found, in other words, in the eye of the beholder and not in the internal character of these machines. For this reason, considering the element of interaction, which shapes how users and audiences make sense of a technology, is especially important to interrogate and understand the cultural trajectory of automata.

The chapter is organized in three parts, each considering an issue that is relevant to the history of both automata and HCI. The first section interrogates the key concept of interaction and how it can be reframed for the case of automata (and for some threads in the history of computing, too) in terms of responsiveness. The second section explores the role of games: Just as digital games were instrumental in sparking innovation in interface design and HCI, human–automata interactions were significantly shaped by the use of games as environments where humans entered in interaction with automata. Finally, the third section tackles the issue of spectacle, interrogating the implications of the fact that automata were often conceived and created with the purpose of being offered to audiences in theaters and exhibition rooms.

1 Of Interaction and Responsiveness

Although HCI emerged as a specific area in computer science only in the 1980s, its subject – the interface or points of contact between human users and computer systems – encompasses the entire history of computing. The first electronic digital computers that emerged during and after World War II provided relatively scarce opportunities for interaction. They were colossal machines owned by public institutions or big private companies, and one needed to manipulate them at the hardware level of wiring and circuitry in order to “program” them to complete an operation.⁵ From the late 1950s and 1960s, more concerted efforts were made to develop hardware and software interfaces that allowed for more engagement on the part of users.⁶ Much of these efforts were framed in a paradigm that pointed to the metaphor of symbiosis as the objective for successful interaction: The coupling of humans and computers under the auspices of new interface systems, so the argument went, would open possibilities that exceeded the abilities of humans and computers in isolation.⁷ The strategies and means that were developed to achieve this symbiosis are manifold: from devices such as the keyboard, the joystick or the mouse to the graphic interfaces of operating systems like Windows or iOS, or the maps and menus used to navigate the fictional environment of a videogame.⁸

In literature about nineteenth-century automata, the concept of interaction is rarely employed. Jessica Riskin privileges the word “responsiveness” to underline the reactive and restless character of early automata.⁹ Yet commentators in the nineteenth century intuited that the possibility of interacting with the machine was to be a decisive characteristic for automata. Mechanical engineer Robert Willis, for instance, proposed a distinction between three different classes of automata based on their pattern of engagement with human observers and users.¹⁰ The first class comprised those mechanisms whose movements result from the mechanism alone, without any feedback from the external world. These included, for instance, mechanisms attached to clocks whereby the movement was activated by the passage of time alone, with no possibility of interaction.¹¹ The second class comprised those automata that were also moved by mechanisms, but were “enabled to change the regular order and succession of their movements, according to existing circumstances; and hence, in some measure, to assume the character of living beings.”¹² Finally, the third class included those automata that appeared to function exclusively through mechanisms, but were in fact wholly or partly controlled by a human agent. Also, in this case an element of interaction was present, since specific tools were included to control the movement, but this was concealed from observers in order to exaggerate the performativity of the machine itself.

The typology divided and organized automata on the ground of their dynamics of interaction with human users. Yet the primary focus of the typology proposed by Willis was not the outcome of this interaction. Emphasis was, instead, put on the appearance of life that characterized automata and the effect of wonder that this appearance stimulated in observers. The responsiveness of automata was meant to recall the ability of animals and humans to react to the environment and to others’ actions, thereby instilling wonder in the audience.¹³ This is evident also in numerous reports reflecting on the reactions of audiences to the exhibition of automata. A writing automaton brought to the stage by magicians John Nevil Maskelyne and George Alfred Cooke in the late nineteenth century, for instance, was praised for its capacity to react to prompts from the audiences, which stimulated more admiration (as well as an amount of skepticism regarding the authenticity of the feat) if compared to other automata that wrote a fixed number of texts.¹⁴ Similarly the Magician, an automaton built in London at the beginning of the same century by Swiss mechanician Henri Maillardet, invited members of the public to select one of the available questions, which were inscribed in oval medallions, to which he would provide an appropriate answer; the effect that the mechanism produced on its audience was enhanced by the fact that it reacted to their prompts, even though the answers had been prepared in advance.¹⁵

Interestingly, for all the difference in the vocabulary used, the “responsiveness” of automata involved some of the dynamics through which interaction is defined in the HCI milieu.¹⁶ Computer scientists and programmers who worked on the development of interactive systems began very early on to reflect on how people respond to observing or using machines that display the appearance of intelligent behavior, much like nineteenth-century automata stimulated commentators to reflect on how people reacted to these machines.¹⁷ Responsiveness, they noted, was often mistaken for intelligence by observers and users who did not understand the deep functioning of computers, to the point that, as AI pioneer Marvin Minsky ironically observed in 1962, “intelligence can be measured only with respect to the ignorance or lack of understanding of the observer.”¹⁸

Through the issue of responsiveness, moreover, one can envisage a filiation between the cultural project of Victorian automata and that of cybernetics. An interdisciplinary field that emerged in the middle of the twentieth century, cybernetics was based on the persuasion that biological life is regulated by the principle of feedback, by which observed actions are taken as inputs to activate further actions. Proponents of cybernetics believed that such feedback, which could be reproduced mechanically, could be mobilized to explain and reproduce the deep functioning of both living organisms and machines. In a similar fashion, as Riskin points out, responsive automata “implied that sensation, the way in which living creatures respond to the world, could be understood in mechanical terms.”¹⁹ In the second half of the twentieth century, some of the pioneers of the emerging computing technology used automata to demonstrate at a practical level the widely held assumption that there was a substantial similarity between living beings and artifacts programmed to respond to stimulations and the environment. Claude Shannon, for instance, created a range of simple devices that demonstrated that simple mechanical systems could navigate complex environments by responding to external prompts, while the Biological Computer Laboratory of Heinz von Foerster explored the boundaries between mechanical responsiveness and life.²⁰ These clever mechanisms remind one of philosophical toys that, throughout the nineteenth century, aimed at entertaining and educating their users; these included optical devices, like the phenakistiscope that created an illusion of movement, but also the small automata or “mechanical toys” that were widely commercialized for domestic consumption.²¹

The concepts of interaction and responsiveness, in summary, demonstrate that a connection or, to use the vocabulary of cybernetics, a feedback loop can be traced between the concept of interaction in HCI and the notion of responsiveness in the history of automata. Despite the fact that the notion of interaction was not used explicitly to discuss automata, interaction was in practice incorporated into nineteenth-century automata and its implications were often discussed by those who witnessed the performances of mechanical attractions exhibiting the appearance of life. Similarly, responsiveness did not figure in the key conceptual toolbox that inaugurated the emergence of HCI systems under the auspices of the symbiosis metaphor; yet some of the reactions observed when automata demonstrated responsive behavior characterize early perceptions of computers too, as audiences and users tended to overstate the apparent similarity between digital computers and the human mind.

2 Games and Playfulness

According to a widespread prejudice, games and playful activities are “just for fun”: They do not bring significant technical innovations or societal change and should be deemed distinct from serious scientific inquiry and experiment. Yet throughout history games have frequently been an engine for innovation and change.²² With regard to the history of computing, games encouraged researchers and developers to envision new pathways of interactions between humans and machines. Chess and other table games, for instance, made possible a confrontation between humans and computers that required turn taking, an important element in human communication that was later implemented into the interface design of a range of applications, from early operating systems to chatbots and voice assistants. Later, digital games opened the way for experimenting with hardware and software interface systems involving the use of other human senses, such as sight and touch.²³ After the 1960s, when time-sharing technologies and minicomputers made it possible for individual users to access computer resources, play was one of the first forms of interaction to be implemented. Games, moreover, were among the first programs coded by early “hackers” experimenting with the new machines.²⁴ As Brenda Laurel puts it, the first programmers of digital games realized that the computer’s potential lay not “in its ability to perform calculations but in its capacity to represent action in which humans could participate.”²⁵

Histories of HCI, in summary, show that games are a key gateway for experimenting with human–machine interactions. The rules of a game create the foundation not only for entertainment but also for the establishment of a controlled system where machines and humans can interact and communicate with each other. Game studies scholars thus argue that the division between humans and machines in videogames is “completely artificial – both the machine and the operator work together in a cybernetic relationship to effect the various actions of the video game in its entirety.”²⁶ Within the frame of the game, the computer can be identified by human users as a potential player and, therefore, as an agent within the game’s design. This has consequences on the relationship between human and computer players, which are in principle indistinguishable within the boundaries of that game.²⁷

It can be revealing, in this regard, to consider the role of games in automata from the perspective that games later assumed for HCI. The most famous and influential example of the applications of automata to games was the Mechanical Turk, an automaton that (apparently) performed the extraordinary feat of playing chess against human players in Europe and the United States in the eighteenth and nineteenth centuries. The Turk anticipated a tradition of spectacular confrontations between human and mechanical players that reverberates to the present day.²⁸ Its appeal was not only due to the level of intelligence that playing chess was assumed to involve, but also to the dynamics of exchange between human and mechanical components that the controlled system of chess allowed.

Although the debate surrounding the Turk focused primarily on the issue of deception, commentators took into account the interactive dimension as well. Much of Edgar Allan Poe’s famous exposé of the Turk as a trick, for instance, pointed to issues that relate to the circumstances of interaction that the exhibition of the chess automaton entailed. Poe reflects, for instance, on how a machine and a human would differ in their approach to a game, pointing out that it would be impossible for a machine to adapt to the uncertainty of the moves of human players: “Now even granting (what should not be granted) that the movements of the Automaton Chess-Player were in themselves determinate, they would be necessarily interrupted and disarranged by the indeterminate will of his antagonist.”²⁹ Following this same logic, others reflected on the necessity for rules that ensured a fairer confrontation between the machine and the human player.³⁰ Such observations echo early discussions within computer sciences on how the machine needs to adapt to the behavior of humans in order to facilitate interactions. The principle of every computer interface, in fact, is to build a common ground where the interaction between users and computers can happen, and such ground can be created only by bringing machines closer to the functioning of human sensoria, body, and intellect.³¹

Poe gave particular attention to the timing through which the chess automaton proceeded to make its moves. He noted that moves were not made at regular intervals of time but accommodated themselves to the moves of the antagonist.³² This, according to Poe, proved that the Turk was not a “pure” machine, but a human player masquerading as automaton. This observation links to an issue that has always been central to HCI: the relationships and differences between human understandings of time and computer time. As media archeologist Wolfgang Ernst points out, the time of computers proceeds on completely different principles than the time of humans. It is based on the clock cycle, a unit of time that signals how long the computer takes to complete the smallest unit of its operations, which in modern computing equates to extremely tiny fractions of a second – a time that is thus unmanageable and even difficult to imagine for humans.³³ One of the problems that HCI faced since its origins was to create software interfaces that allowed humans to use computers seamlessly, without perceiving the disconnection between their and the device’s time. One of these systems was time sharing, which creates the illusion that the computer is handling different operations at the same time, while in fact it is just alternating between multiple tasks.³⁴

Poe’s point about the Turk adapting to the discontinuity of the human player was meant to provide evidence it was a fraud. A human–automata interaction perspective, however, stimulates us to consider the extent to which the mechanism adapts to the timing of the human, becoming seamless and potentially also invisible to the user. In HCI, this is reflected in computers sometimes being made to respond more slowly than they could, to pass as human or simply not to upset users who are not accustomed to computer-powered speed. Due to the calculating power of modern machines, for example, a state-of-the-art chatbot can reply to a human query in a fraction of second. In theory, answering queries so rapidly should be regarded as evidence of skillfulness and thus (artificial) intelligence. However, because such speed is impossible for humans, chatbots are often programmed to take more time than they would need to provide an answer, so that they can simulate more effectively the conversation of human users.³⁵ In Large Language Models such as ChatGPT, for instance, the system’s responses appear word by word, as they would be if typed by an operator, even if this does not reflect the actual functioning and speed of the software.

Another interesting aspect in the display of the Mechanical Turk was the role of the exhibiter. Moving from the automaton’s board to another board on which the human player played her or his moves, the exhibiter acted as a proxy to enable the interaction between the automaton and the human player.³⁶ Interestingly, something similar happened in the late twentieth-century challenge between the computer program Deep Blue and chess grandmaster Kasparov and in the more recent confrontation between Google’s AlphaGo and Korean Go master Lee Sedol. Also, in those cases, it was necessary for a human to make the moves on behalf of the machine: At the chessboard sat not just the human player but also another human impersonating a machine and moving the pieces on its behalf, as if to remind one that something (a body) was after all missing from the powerful calculators. With regard to this, what complicated the performance of the Turk is the fact that the automaton indeed had a body – or at least, a physical representation of it – in the form of a mechanical puppet, racialized as a “Turk,” that performed actual operations and moves during the game. The exhibiter went from the Turk’s chessboard to its opponent’s, his movements signing not only the specter of fraud (as Poe underlined), but also the specter of incommunicability between the human and the machine, unable to share the same playing field.

The Turk was the most famous, but not, by any means, the only automaton that played with the thin borders between playfulness, technology, and innovation. Alongside Poe’s exposé, other reports underlined the playfulness of automata and their relationship with games and toys.³⁷ As mentioned, moreover, automata made the way into the houses of Victorians in the shape of mechanical toys, reproducing the appearance of human behaviors, including movement and voice.³⁸ As one writer pointed out, however, their association with play did not imply “that these automata are useless curiosities.”³⁹ Like philosophical toys, automata served as a field where present and possible future interactions with machines were tested and presented to the public. The entertaining character of games, moreover, went hand in hand with the third and final aspect of human–automata interaction examined in this chapter: the spectacular dimension of automata.

3 Exhibition and Spectacle

Since their emergence in the mid-twentieth century, computers participated in the long-standing tradition of exhibiting technological achievements within a spectacular frame.⁴⁰ Public events in which the potential of computing was the center of public-facing demonstrations traversed the entire history of modern computing, from the exhibition of early computers as mechanical brains to IBM’s Watson competing in the TV quiz Jeopardy!, from the transformation of the Turing Test into a public contest at the Loebner Prize to the human–machine duels won by Deep Blue and AlphaGo.⁴¹ Today researchers, companies, and entrepreneurs continue to make substantial efforts to emphasize the capability and achievements of digital computers and related technologies through spectacular means: In robotics, for instance, carefully staged demonstrations of the anthropomorphic robot Sophia have been organized in public venues and broadcasted on TV, and Boston Dynamics’ videos of dancing robots circulate widely online.⁴²

Nineteenth-century automata do more than share this culture of technological spectacle: While usually computers are produced primarily to perform some tasks although they might be adapted to spectacular demonstrations, automata were often created and circulated specifically to entertain their users and audiences. Automata were, therefore, built with the eyes of the beholder in mind, and this deeply informed their design and their performances, as well as the public reception. The “user” to which these machines adapted was more often a spectator than an interacting user in the strictest sense of the word.

As a result, automata that provided little or no possibility for interaction were nevertheless constructed so that they could create an impression of engagement among the audience. An automaton that played music, described in an 1863 article in the New York Times, was for instance programmed so that it glanced and bowed at the audience.⁴³ Also, the Turk occasionally moved its head and arms, as if commenting upon one of the chess moves in the game.⁴⁴ These gestures are reminiscent of recent explorations in HCI and especially in social robotics, where theatrical movements have been incorporated into the design of autonomous machines to make them more pleasing to their users.⁴⁵

The idea behind this work (and behind the domain of social robotics in general) is that the behavior of the machines impacts how users perceive them. It follows that social robotics, and, in general, all AI technologies programmed to communicate and interact with humans, incorporate an element of deception, since their design can activate specific kinds of responses, including empathy and socially minded behavior. AI and robots are in this regard indeed the modern equivalent of the Mechanical Turk and other “fake” automata: In contrast to the Turk, they might really be able to play chess, but they will still invite users and audiences to overstate their ability in areas such as empathy, creativity, or sentience.⁴⁶

In the nineteenth century, audiences’ interactions with automata were often described as stimulating emotional reactions such as astonishment or fear, and commentators frequently evoked the cultural proximity between automata and magic.⁴⁷ In fact, stage magicians such as Jean-Eugène Robert-Houdin and John Nevil Maskelyne built and exhibited automata in their shows.⁴⁸ The spectacular power of automata was seen as originating in the spectator’s uncertainty about the technological process that allowed these machines to move and act, much like how, in a spectacle of stage magic, the nature of the trick is hidden from the audience, inspiring their curiosity.⁴⁹

The gullibility of the audience was also a matter of substantial reflection. Many dismissed the public’s engagement with the Turk as evidence of their lack of critical sophistication.⁵⁰ Others, however, gave a more nuanced interpretation: They realized that automata were objects that allowed one to investigate and explore the limits and potential of trickery, illusion, and fraud, in continuity with a newly emerging industrial show business that programmatically explored audiences’ desire to discover if an attraction was authentic or not.⁵¹ Some reports on automata, indeed, underlined that people enjoy being deceived, and that this taste for deception “is a principle of human nature.”⁵²

Another key element in the spectacular exhibition of automata was the subtle play between hiding and revealing the mechanisms inside the machine. Much has been written about modern computers as “black boxes,” that is, technologies whose actions and outcomes are apparent, while the mechanisms that lead to these outcomes remain opaque to their users.⁵³ In spectacular events that included automata, a significant component of the show revolved around the act of “opening up” the automata’s black box to show their inner mechanisms.⁵⁴ In cases such as the Turk, however, this was a theatrical act more than an authentic effort to show the functioning of the machine: The vision of the complex mechanisms, whose exact functioning remained mysterious, was meant not to educate the audience but to inspire their wonder toward the machine. The tradition of magic is, again, relevant in this regard: Stage magicians usually do not reveal the nature of their tricks, but they might suggest that these are of a technological nature, thereby evoking the sensational character of modern technology and its association with magic.⁵⁵

It is not surprising, in this sense, that the nineteenth-century literature on automata largely consists of exposés of their trickery or of attempts to illuminate their inner functioning.⁵⁶ Similarly to the exposés that debunked the trickery of spiritualist mediums in the Victorian age, making spiritualism a subject of public controversies and boosting its impact on the public imagination, this body of literature did not so much reduce the public fascination with automata as enhance and support it.⁵⁷ “The faith-inducing dramaturgy of technology,” as Mark Sussman called it, exploited not just faith but also willingness to find out if there was a trick, and how this trick functioned as a technology in itself.⁵⁸ The fact that the automaton chess player became “an object of intense curiosity,” as Poe put it, was therefore related to its status of technological novelty as well as to the uncertainty about this status.⁵⁹

4 Conclusion

In histories of AI, automata are commonly indicated as progenitors of the field. Given AI’s obsession with the dream of recreating the human mind by mechanical means, it is unsurprising that the relationship between automata and modern computers is usually discussed in such terms. Yet the case of automata shows arguably more about the way humans react to machines exhibiting intelligent behavior, than about the functioning of these machines and their relationship with the human. The “human–automata interaction” perspective that I developed in this chapter places emphasis on the role of users and audiences. It was their enthusiasm about seeing a machine move and behave like a human, their faith or their skepticism regarding the machine’s performance, their curiosity to understand how it worked, that drew large masses of people to demonstrations and exhibitions of automata in the nineteenth century and beyond.

Elsewhere, I have argued that a similar perspective is essential to understand the impact of contemporary AI. Whenever AI technologies interact with users, in fact, the outcome of this interaction is given by the functioning of the machine as well as by the contribution of the user, who adds their interpretation and perception to what is happening. Voice assistants such as Alexa and Siri, for instance, work well for the many users who have adopted them, not just because their hardware and software are sophisticated and efficient, but also because specific design choices have been made to facilitate their integration in everyday life and environments, starting from their humanlike, gendered voice.⁶⁰ Automata provide a prehistory and an archeology of these dynamics: As I have shown, long before the introduction of digital computing, automata stimulated reflections and discussions of how users reacted to machines.

Each section of this chapter has moved back and forth between the histories of automata and of HCI. Such an unusual approach is based on the persuasion that investigating “human–automata interaction” entails more than just finding a precedent or a prehistory for HCI. More than pre- or posthistories, this chapter attempts (in cybernetic fashion) a feedback loop between moments and areas in the history of technologies that are rarely examined in conjunction with each other. We may learn a few things about HCI through the study of automata, and we may learn a few things about automata through the study of HCI.

In conclusion, human–automata interaction reminds us that the impact of technology is never due to technology alone, but also to its integration in social and cultural environments. As every practitioner of HCI knows well, technologies that enter into contact with human users need to be adapted and prepared for them. The construction of automata was not different: They were built with their users and audiences in mind; they were invented to be seen, admired, sometimes manipulated by humans. Their history is a history of how ingenious mechanisms were conceived and put together, but also of how their ideation and construction moved from reflections and anticipations of how people would perceive, react, and interact with them. The science of automata, in other words, was always located at the interface between the human and the machine.