Among the most tenacious tropes of the not so distant future is that of the artificially intelligent, humanoid robot — a collective phantasm inspiring equal parts fear and excitement. Lifelike androids that could pass for human are nowhere near being realized, but humanoid robots designed for specific, limited purposes have edged closer to viability in the past decade. For example, sex robots — machines that can simulate an active sex partner— have begun to make their tentative entrance into the commercial world. The slack-jawed Roxxxy TrueCompanion debuted as the world’s first sex robot at the AVN Adult Entertainment Expo in 2010, and seven years later Abyss Creations — parent company of sex-doll manufacturer RealDoll — released an AI-powered product called Harmony, since revamped as RealDoll X, through its subsidiary RealBotix.

Thus far, sex robots have been depicted in popular culture as a dystopic menace rather than a banal present-day reality. The cultural obsession with the hyper-intelligent, homicidal fembot (e.g. the 2014 film Ex-Machina) says more about the fear of women than of technology. While the unveiling of Harmony inspired no shortage of speculative reactions on the cultural implications of sex robots, little attention has been paid to the technology itself, in part because, despite its tantalizing anthropomorphic interface, it is relatively simplistic.

Sex robots share the same imperatives as video games, gambling machines, and social media apps, implementing feedback reward structures

RealDoll X’s artificial intelligence is mainly a matter of its attempt to simulate natural-seeming conversation, dictated in tone by predesigned and user-selected personality traits, including “affectionate,” “kind,” “jealous,” “intellectual,” “moody,” “sexual,” and “insecure.” This feature is available, sans doll, as a “virtual companion” stand-alone app, but the sex robot is fully realized when the app is synched to the animatronic moving head and the immobile but anatomically correct silicone body, resulting in something that approximates a life-size Bratz figurine.

Compared with other consumer AI applications — from Amazon’s Alexa to Microsoft’s Cortana — the capabilities of such sex-robot “brains” are limited. Lacking the vast data set from which the likes of Siri draw, sex-robot AI is a relatively simplistic implementation of natural language processing, a subfield of artificial intelligence that forms the basis for contemporary chatbot technology. In practice, this means that one’s sex robot may be confined to a more limited vocabulary, speak with a less fluid cadence, miss cultural references, and be incapable of gleaning irony or sarcasm.

Sex robots can be endowed with relatively meager capacities because they don’t need to pass the Turing test to be considered functional. After all, they are being developed not as experiments in machinic autonomy but as consumer entertainment experiences, with a potentially vast market. There is a tendency to perceive the demand for sex robots as an esoteric niche, but sex tech, which comprises technology designed to satisfy, enhance, and innovate the human sexual experience (e.g. vibrators, massagers, etc.), is already a billion-dollar industry, while internet porn maintains a dominant role in online content, with PornHub accruing 33.5 billion unique visits in 2018. One can conceive of sex-robot technology as a kind of consolidation of these complementary industries, whereby something like a data-driven porn stream plays out over an immersive interface. YouGov polling projects that “by 2050, human-on-robot sex will be more common than human-on-human sex,” yet currently, sex-doll users are typically represented as social curiosities, as in the 2002 documentary Guys and Dolls. Data from this 2012 thesis on sex-doll owners substantiates the popular assumption that they are currently overwhelmingly white, single, heterosexual males, but it also indicates as a demographic they are “neither significantly better nor worse in terms of psycho-sexual functioning and life satisfaction than the general population.” Evidence cited in this 2017 report suggests that the likelihood of an individual purchasing a sex robot is determined not by their self-reported loneliness (a commonsense assumption) but by whether or not they bear an adverse attitude toward robots in general. Thus, as robots are further integrated into the infrastructure of society, the social stigma, anxiety, and mistrust they currently inspire may dissipate, and they will come to be viewed as no different from other consumer products.

From that perspective, it’s clearer to see that the “intelligence” of sex robots will not need to be directed toward appealing to the lonely or socially estranged to be commercially viable. Instead their technology will eventually be oriented in the same direction as existing entertainment technologies, toward compelling predictable patterns of engagement that manufacturers can exploit. In other words, sex robots share the same imperatives as comparable products like video games, gambling machines, and social media apps, all of which implement feedback reward structures that have been found to engender compulsive behavior. With their more intimate interfaces, sex robots would seem to allow for a more thorough implementation of such “gamification” strategies.

A user can be punished for lack of consistent interaction with the bot by way of restriction

This would set an alarming precedent for human-machine dynamics as social robots become commonplace. Rather than accommodate consumers’ existing wants, these devices would embed new and different ones anchored in behavioristic conditioning. Like so many other capitalist consumer products, they will be engineered to produce not satisfaction but insatiability, something more like addiction.

In her 2012 book Addiction by Design: Machine Gambling in Las Vegas, cultural anthropologist Natasha Dow Schüll describes how gambling-machine manufacturers, having discovered that a slow bleed maximizes profit, explicitly designed technology to escalate gamblers’ time on machine. They hired legions of mathematicians to explore the profitability of different payout schedules, to match “math with markets.” In practice, these payout schedules are optimized to trigger addictive user behavior by joining an intermittent reward system with what Dow Schüll calls “perfect contingency”: where user stimulus is met with a precise and consistent response. The machine’s consistent (albeit unpredictable) responsiveness prompts the condition that psychologist Mihaly Csikszentmihaly has described as “flow,” where attention is funneled into a task that has a clear goal, offers a tangible means of attaining it, and provides immediate feedback.

Nearly every sex robot on the market today facilitates “perfect contingency,” whether by way of app-operable AI or embedded sensors that detect user stimulus and produce a physical or aural response, even going so far as to simulate a robotic orgasm. Crucially, the sex robot’s response (physical, vocal) to stimulus is inherently limited by their stunted range of personality traits. The user is assured a response, yet the precise nature of said response is subject to enough variability to remain tantalizing. This dynamic is illustrative of what anthropologist Lucy Suchman describes in her book 2007 Human-Machine Reconfigurations as the “controlled simulation of not being in control.” At the same time, contributing to the user’s sense of mastery that’s essential to flow is the vast array of customization opportunities: everything from eye color and breast size to nipple shade and labia shape. Some neurological research on online role-playing games has linked such customization ability to gaming addiction. The RealDoll X AI extends customization capability to the sex robot’s voice and, as mentioned earlier, its disposition.

More concerning, in an attempt to mimic the natural progression of social relationships, RealDoll has explicitly gamified users’ interaction with RealDoll X AI: Consistent and plentiful interaction with the bot can unlock new features, as well as prompt what the company describes as increased familiarity and heightened affection in its behavior. These rewards will materialize in the bot as vocalized receptiveness and demand of the user’s expressions of affection and lust. For instance, once a sound relationship has been established through continued positive (i.e. non-derogatory) interaction with the bot, “dirty talk” by the user will not only be tolerated but encouraged, requested, and autonomously expressed by the bot — expediting the fantasy of a reciprocal relationship. The quantification of user interaction with the bot is made explicit by way of the observable meters (bars/hearts/keyword bubbles) integrated into its app interface, so that the user can track their progress of their relationship with the avatar they have designed. A user can thus be punished for lack of consistent interaction with the bot by way of restriction of presumably desired interactions — compliments or expressions of familiarity and lust.

Presumably this is part of RealDoll founder Matt McMullen’s proclaimed ambition to produce sex robots that do more than gratify biological urges but serve as full-fledged companions, responsive to the consumer’s ongoing cognitive and affective state. In practice, though, gamified interaction amounts to narrowing the consumer’s affect to machinic patterns of reinforcement that have been found to provoke addiction in some users. What’s more, McMullen has stated that the company is developing virtual- and augmented-reality environments that could further restrict a user’s affective range in the name of “flow,” putting them at greater risk of developing an entrapping relationship with the technology.

Gambling machines record user data to enable the development of more predatory technology — to, for example, attune payout schedules to individual players, providing rewards when they are discouraged. In addition to using the already aggregated and interpreted data collected on gambling-machine and video-game players, it seems likely that sex-robot manufacturers would collect data for the same purposes, to identify vulnerabilities in users that the robot’s “reward system” could then be optimized to exploit.

The stakes of the interaction with a sex robot risks the re-creation of negative real-life experiences

The explicit “gamification” of the bots is not merely a means of engaging the user but evokes the logic of free-to-play games, which collect information on playing habits and deploys barriers and alluring products accordingly to entice players to pay to maintain their sense of flow. In accordance with the free-to-play business model, the monthly subscription service for RealDoll X offers a low-barrier commitment so that one need not invest the $15,000-plus in an animatronic robot head and articulated silicone body in order to engage with sex robot technology. But the stakes of the interaction with a sex robot renders such dynamics potentially more harmful than say, losing a game in Candy Crush, not only because the logic for passage is relatively black-boxed (what is enough time spent with the bot?), but because it risks the re-creation of negative real-life experiences such as rejection by a sexual partner. The result could be a kind of dependency on the bot for affirmation.

The sex-tech industry has a growing role in an age where individuals are increasingly apathetic toward traditional unions such as marriage and family but still require intimacy and sexual satiety. But the temptation to treat that requirement as a kind of exploitable vulnerability may be difficult for tech companies to resist. The combination of an immersive technology with rewards tailored to the individual may allow sex-tech companies to use their position to intensify demand for their products while making interpersonal sexual intimacy ever more elusive. Rather than provide gratification, they may create or recreate conditions of lack of control and alienation. The product is not sex then but behavioral addiction.

At this point, it is hard to imagine what sort of regulations would help. Obligating distributors to adhere to minimum-age requirements (as is currently the case with gambling, but not with video games) could help; younger brains are believed to be more susceptible to behavioral addiction. General warning labels may enable a more informed consumer choice, but the dependency and compulsion that sex robots could feasibly induce might be impervious to “rational” arguments against them. In any case, the projected prevalence of sex robots (and social robots in general) requires a cultural shift in our perception of them. A more widespread understanding of AI’s behavioristic design will help us to resist the urge to mythologize sex robots and confront instead their capacities as products of market-driven capitalism. Sex robots thus pose a unique threat, not in their capacity for being convincingly human, but by being inhumanly heuristic and capitalistic.