Book Review: Deceitful Media: Artificial Intelligence and Social Life after the Turing Test

social through the para-social relationships between those playing and watching, reshaping what it means to watch someone play games. There are also ‘Twitch drops’ where in-game content becomes available if you watch enough of a game streamed on the platform. Twitch is also used to play games collectively. In 2014 ‘Twitch Plays Pokémon’, a stream where people’s comments determine how the protagonist moves and acts began. Collectively playing Pokémon is often chaotic, with multiple people inserting different instructions simultaneously. Through the madness, a community based around a distinct way of playing the game, with its own interests, practices and its own cultural lore emerged. While a bit of an outlier of a case, it does demonstrate how influential Twitch has been in terms of its social aspects. Despite these two areas needing elaboration, Arditi’s demonstration of social change and the development of a hyper-fast capitalism in streaming is compelling and informative and will be of use in classes focused on media politics and popular culture. With the development of streaming, some things are lost, and others are gained. Streaming Culture is most certainly a gain.

social through the para-social relationships between those playing and watching, reshaping what it means to watch someone play games. There are also 'Twitch drops' where in-game content becomes available if you watch enough of a game streamed on the platform. Twitch is also used to play games collectively. In 2014 'Twitch Plays Pokémon', a stream where people's comments determine how the protagonist moves and acts began. Collectively playing Pokémon is often chaotic, with multiple people inserting different instructions simultaneously. Through the madness, a community based around a distinct way of playing the game, with its own interests, practices and its own cultural lore emerged. While a bit of an outlier of a case, it does demonstrate how influential Twitch has been in terms of its social aspects.
Despite these two areas needing elaboration, Arditi's demonstration of social change and the development of a hyper-fast capitalism in streaming is compelling and informative and will be of use in classes focused on media politics and popular culture. With the development of streaming, some things are lost, and others are gained. Streaming Culture is most certainly a gain. Artificial intelligence (AI) is not actually intelligent. According to Simone Natale, we are deluded into believing that it is. Deceitful Media positions AI as both a trickster and an ethical problem and, ultimately, as a technology whose primary function is that of deception rather than of knowledge production and consciousness. The long-standing question of whether AI can think or not, the author argues, distracts us from what is really at stake: how have our relationships with AI and the construction of an algorithmic imaginary (Bucher, 2017) informed the development of intelligent machines? Natale's book focuses on this latter question. Whether AI is real or fake does not matter so much as the aura of intelligence these algorithmic technologies possess. Over its six main chapters, the book considers how the design and deployment of communicative AI has been contingent on creating illusions about what AI is and what it can do. Today, what has emerged is not a 'strong' AI that resembles human intelligence, but a 'strong' illusion of the capacity of AI to surpass human intelligence that exploits our very human proclivity to be deceived.

ORCID iD
For readers unfamiliar with the history of AI, Natale provides an accessible overview of the inception, developments, failures and achievements of AI from the mid-twentieth century onward. To frame this discussion, he invokes perhaps unexpectedly for those unacquainted with his previous work (Natale, 2016;Natale and Pasulka, 2020), late nineteenth-century spiritualist séances as a metaphorical precursor to AI. He explains that how people inadvertently deceive themselves into experiencing supernatural phenomena (inanimate objects moving on their own accord or spirit communication with Ouija boards, for example) is analogous to how users readily participate in the co-creation of the illusion of intelligent machines.
The book's chapters discuss seminal case studies on relationships between humans and computers: ELIZA, the first chatbot, developed by Joseph Weizenbaum at MIT in 1964; the infamous and largely forgotten Microsoft Bob virtual assistant software for computer novices launched in 1995; the Loebner Prize competition for AI developers occurring annually since 1990 that has made an academically discredited albeit entertaining public spectacle from the Turing Test; and the recent advent of domestic AI voice assistants such as Apple's Siri, Amazon's Alexa and Google's Assistant which are reconfiguring users' access to and relationship with information. Throughout these analyses, Natale draws on human-computer interaction (HCI) theory (Turkle, 1995) to demonstrate that it is not only AI that is being adapted to us-we are adapting our own behaviours and language to AI.
These cases explore the genealogy of AI in the context of games and play, in which deceptive design is not only necessary for enjoyment, but functions to obscure relational discrepancies between user and computer. Informed by actor-network theory (Akrich, 1992) and, in particular, the 'computers are social actors' framework (Neff and Nagy, 2016;Reeves and Nass, 1996), Natale contributes a novel perspective by bridging the histories of AI and HCI and by stressing the role of the user as the agent who renders AI 'intelligent' through a 'symbiotic relationship' (p. 101) with the machine. In other words, AI is only intelligent insofar as it is perceived as such.
With this book, Natale demystifies the deceitful nature of AI while reframing our understanding of deception from a peripheral to central question in Communication Studies, not unlike Finn Brunton's (2013) and Elinor Carmi's (2020) studies on spam. This discourse has wider implications beyond AI. We are naturally predisposed to deception, he explains, but not necessarily in the sense that we are susceptible to malicious trickery. Not only is 'deception is an inescapable fact of social life' (p. 4), but 'deception implies some form of agency' (p. 4) too. We allow ourselves to be deceived, he contends, so that we may be able to coherently engage with the world and to maintain relations with people and things. This mundane suspension of our disbelief is an innate feature of human cognition which he dubs 'banal deception' in contrast with straight-out, deliberate deception.
AI developers, as Natale suggests, mobilise banal deception with 'elements such as appearance, a humanlike voice, and the use of specific language expressions [which] are designed to produce specific effects in the user' (p. 6). He further demonstrates how computer scientists, realising this early on, subsequently developed AI upon the premise that 'users' perceptions of computing systems could be manipulated in order to improve interactions between humans and machines' (p. 48). In fact, Natale traces this deception in AI back to Alan Turing, who with his eponymous 1950 Turing Test proposed that a machine's intelligence can be determined by its ability to convince human participants that they are communicating with another human.
However, as Natale elaborates through an anthropological lens (Appadurai, 1986;Gell, 1998), long before Turing and the advent of intelligent computation, this banal deception existed as an art of historical and dramaturgical tradition and as a necessary function of media technologies. Only when we accept illusion as reality can we engage meaningfully with art, literature, theatre, cinema, recorded sound and telecommunications, for example. What is implied throughout the book, and from which Natale's argument would benefit if articulated explicitly, is that what is different about AI is our unprecedented inability to disengage with the illusion of AI as we do when we cease to interface with art and other media.
Yet Natale confusingly oscillates between warning and reassuring the reader about this banal deception. While at times he claims that 'our vulnerability to deception is part of what defines us' (p. 132), he is quick to caution that 'banal deception may finally bear deeper consequences for societies than the most manifest and evident attempts to deceive' (p. 7). Is banal deception a natural feature of human perception and sociality or is it the most insidious problem posed by AI? This remains unclear throughout the book.
The book's strengths, nevertheless, lie in its succinct thesis and eclectic frame of reference that extends across media studies, science and technology studies, religion and media, social anthropology and media archaeology. Deceitful Media makes a meaningful contribution not only in the broader arena of humanities perspectives on emerging AI technologies but also within a smaller cohort of scholars (Benqué, 2021;Lazaro, 2018;Marenko, 2019;Steyerl, 2018) writing about the magical, spiritual or occult lineage and implications of AI.
Finally, it is important to note that much has happened in the realm of AI in the year since the publication of this book. Since the latter half of 2022, we have witnessed the seemingly sudden emergence and evolution of text-to-image generators available to the general public from deep learning AI models such as DALL-E, Stable Diffusion and Midjourney, as well as OpenAI's articulate and omniscient ChatGPT software. While Natale's focus in Deceitful Media is specifically on the legacy of communicative AI and recent developments in the area of speech processing, his concept of banal deception may offer a compelling approach for the study of such generative machine learning technologies that rely on natural language processing models.
How is the instantaneous ability of these technologies to produce, from a short descriptive prompt phrase, an elaborate and accurate one-of-a-kind image or a coherent and credible original report in any discipline of knowledge contingent on the manipulation of human intelligence through illusory design and user interaction? The rapid proliferation of these technologies in mainstream popular culture not only poses unprecedented ethical and legal challenges but also represents the first widespread cognisant interfacing between ordinary users and generative AI models. As a nascent and critically transformative site for human-machine relations, what is at stake here is not simply that we are being half-wittingly tricked by AI, but that there are fundamental shifts occurring in our own consciousness as we continue to engage with these technologies.
As Natale cautions, 'it will become increasingly difficult to maintain clear distinctions between machines and humans at the social and cultural levels' (p. 129). Thus, a main challenge for scholars of HCI and developers of AI will be to determine how these technologies are redefining what it means to think, communicate and relate with other humans, as well as with the non-human and more-than-human. Such research will inevitably have to contend with the de-centring of intelligence from that which happens inside the human brain to that which emerges from relationships and interdependencies between heterogeneous agents (Bridle, 2022). By analysing generative AI models through a framework of banal deception, we can, perhaps, determine not only how humans and machines are in the co-creative process of transforming what intelligence is, but also to identify emergent and potential forms of planetary intelligence.