1 Introduction
Along with biological and psychological factors, human sexuality is very much a socially constructed experience, where social behavioural changes are tied with technological advances. Being a basic human need and a fundamental form of social motivation, sexuality merits careful attention within a rapidly changing technological landscape. As the digitalization of life deepens at the start of the twenty-first century, all aspects of human society have already been fundamentally changed. Sexuality is no exception from this trend; for example it is now not unusual to choose a partner over Tinder, to enjoy socialization on YouPorn or to have sex in an immersive virtual reality, and even to have intimate contacts with sex robots (robots with artificial intelligence designed to interact sexually with humans). In fact, we witness the emergence of the term digisexuals , people whose primary sexual identity only comes out when interacting with digital technology (McArthur & Twist, 2017). We are now confronted with the need to explore these new dimensions of human sexuality that are not confined to human–human interactions.
Striking anecdotes about intelligent sex toys and much-touted examples of sex robots in the social media have painted a vivid picture of the imminent arrival of sex robots. This asymmetry between actual and perceived state of the art has created heated discussions around the pros (Devlin, 2015; Levy, 2007) and cons of introducing sex robots into human relationships (Richardson, 2016; The Guardian, 2015). Nevertheless, it has also served to draw attention to far-reaching social and ethical challenges that will be imposed on us as users of this new technology (Danaher & McArthur, 2017; Devlin, 2018). Despite the obvious scientific and social importance of this topic, systematic empirical research in this field is sparse (FRR Report, 2017; Nature, 2017). For this reason, we are not yet in the best position to answer many fundamental questions surrounding this topic, pertaining to the role of sex robots and the likely impact of this new sexual experience on individuals and on human society. In this outlook, rather than raising many more questions that we cannot answer at the moment, we will attempt to address the question of whether humans and machines are ready, from the available empirical evidences so far, for such interactions. We will also highlight the importance of a scientific transdisciplinary approach to the study of human sexuality in the twenty-first century.
2 Are They Ready?
Over time, robots have developed from being safely segregated and distal production facilities to being our interactive and proximal partners in everyday affairs. At the same time, roboticists increasingly design social robots to appear human-like because they believe that similarity supports the intended interactivity. Extrapolating this gradual encroachment of humanoid robots, we must eventually answer the question of whether social robots are ready to be our intimate partners. The answer is largely dependent on what precisely we humans expect these future social robots to do. Already Fong, Nourbakhsh, and Dautenhahn (2003) suggested that, in order for social robots to interact with humans in a successful manner, they should exhibit a set of ‘human social characteristics’. Specifically, they should express and perceive emotions; be able to communicate with high-level quality dialogue; have the ability to learn social skills, to maintain social relationships, to provide natural communicative cues such as gaze and gestures and have certain personality. It might still take decades for roboticists to enable humanoid robots to incorporate these socially intelligent properties; and even when this is the case, it will be unclear whether such skills suffice to sustain intimate or even sexual relations.
In the case of sex robots, Scheutz and Arnold (2016, 2017) conducted two consecutive surveys using the same measurement concerning expected capabilities, appropriate forms and functions of sex robots. In their two studies, a total of 298 participants (mean age = 33.76 years old, with 100 participants in the 2016 study) recruited via the internet platform Amazon Mechanical Turk (AMT) in the USA rated 14 potential capabilities of sex robots. Take initiative, recognize human emotions and has feelings were consistently rated as the least attractive capabilities, while the top four most attractive capabilities were specially designed to satisfy human desire, moves by itself, can be instructed and obeys order. Leaving aside concerns about the representativity of the participant sample and other methodological details of this pioneering study, these survey findings seem to suggest that ordinary people currently still see sex robots more as objects that can be controlled by their owners and less as autonomous beings that express and perceive emotions. This implies that it is currently hard to imagine that we eventually might attribute intentionality and agency to humanoids (cf. Kewenig, Zhou, & Fischer, 2018).
Currently, the expectations of the general public and those of robotics researchers regarding expected capabilities and social characteristics of sex robots do not seem to match. Further empirical data is urgently needed in this area to understand the views and attitudes of various groups of people (e.g. the public, psychology and robotics experts and potential users, such as therapeutic patients and socially isolated individuals). Their views can then inform discussions about how sex robots can best meet human needs and interests.
We are still some distance away from a sex robot that can replicate the full experience of human-partnered sex. It is a myth of the so-called strong AI view currently prevalent in the media that humanoid robots with social intelligence will be widely viable within a few years. If we turn to the empirical evidence from robotics research and development, the fact is that the level of intimacy-related behaviours of artificially intelligent humanoid robots is still far from what the popular press has portrayed. On the other hand, there is a great deal of diversity when it comes to the experience of intimacy. Some people already use artificial companions to offer them satisfactory social and intimate experiences despite their limited level of social intelligence.
Compared to other applications of social robotics, such as health care and education, potential sexual applications of artificial intelligence, including human–robot intimate interactions, have not yet received sufficient consideration from the research community. Yet, there are important lessons to be learned from the social robots in existence today. All social robots invite humans to physically and/or socially interact with them, either through their morphology or their activities (Breazeal, 2002; Steels & Hild, 2012). Research shows that during interaction with social robots, humans have a tendency to attribute human features to robots, a phenomenon called ‘anthropomorphizing’ (Duffy, 2003; Melson, Kahn, Beck, & Friedman, 2009). For example, young children are more attached to toy robots than to dolls or teddy bears and even consider them as friends (Tanaka, Cicourel, & Movellan, 2007). This is a result of the robot’s ability to produce evolutionarily formed communication signals, such as trusting sounds and eye contact (Cangelosi & Schlesinger, 2015). Similarly, elderly care home inhabitants were reported to have a lower level of loneliness and higher level of social interactions as a result of interaction with a pet-like robot called Paro (Robinson, MacDonald, Kerse, & Broadbent, 2013). However, currently available social robots still have very limited abilities of social interaction.
The above examples show that researchers can fruitfully apply basic research from social psychology to understand the mechanisms supporting more efficient and effective human–robot interaction (Breazeal, 2003; Duffy, 2006; Heerink, Kröse, Evers, & Wielinga, 2009). In turn, the latest generation of intelligent humanoid robots is perceived to be an excellent tool for the study of variables influencing human social behaviour. This is due to the fact that humanoid robots will perfectly reproduce any given behaviour as often as needed for a study, while human interaction partners will vary from trial to trial and thus create unwanted error variance (although this absence of variability may itself create problems for the human interaction partner).
3 Are We Ready?
Currently the two most common but perhaps also most challenging social behaviours that roboticists are trying to model are social distance (Mumm & Mutlu, 2011) and social gaze (Ruhland et al., 2015). Roboticists argue that, in order to enable robots to be better integrated into human society, robots should show appropriate social distancing behaviours based on human norms. Researchers systematically observe how humans approach each other and try to model these approach behaviours in robots and then test the robot models in robot–human interactions (Avrunin & Simmons, 2014). Because social distancing behaviours (the so-called proxemics) in human interactions are influenced by many factors, including likeability, gender, culture and age, creating robot models that exhibit similar behaviours is challenging. Social gaze in human interactions functions as an important cue for trustworthiness. Robots are modelled for different gaze acts (long, short and rapid gaze shifts) to signal their level of engagement while interacting with people. Human–robot gaze cueing has also been studied in relation to the behavioural and neural signatures in the human brain (Wykowska, Chaminade, & Cheng, 2016). Due to their still limited language abilities, nonverbal behaviours such as gaze, posture and gesture become important means by which robots convey and also detect emotions. Although robotics researchers have made progress in these challenging areas, a recent review highlighted the general lack of knowledge about the mechanisms that encourage communication between humans and robots (Royakkers & van Est, 2015). The authors concluded that ‘this research discipline of human–robot interaction is still in its infancy’.
Let’s now turn to the social behaviour of interest, namely sexual interactions between humans and robots. Despite a current lack of empirical evidence regarding sex robot research and development, the sex industry, some robot technologists as well as members of the academic community (Nature, 2017) have come to believe that sex robots can be a major driving force for the development of social robots and for human–robot interaction research generally.
Historically, there have been huge changes in people’s attitude, from denial to acceptance, regarding various sexual practices, such as homosexuality and masturbation (Brenot & Coryn, 2016; Levy, 2007). Changes in sexual thinking and sexual attitude reflect changes relating to culture, religion and historical periods. We believe that the acceptance of digital sexuality as a mainstream expression of human sexuality will depend on (a) the speed of research and development in sexual applications of artificial intelligence and (b) the speed of social adaptation to these technologies. In order to better understand trends in future human sexuality, the next important question is whether attitude and behaviour correlate (Ajzen, 1991) and how well people’s sexual attitudes predict their sexual behaviour (Reiss, 2006). In the case of sexual interaction with intelligent machines, what are the barriers preventing this change? What would be the appropriate process to bring about this change? And why would it be useful? We do not have definite answers to these questions at the moment. Nevertheless, there is a huge scope for research to explore the relationship between sexual attitude and sexual behaviour.
Recent surveys among Western cultures revealed variable numbers regarding people’s positive attitude towards robot sex: a whopping 86% in the USA (Scheutz & Arnold, 2016) but only 20% in the Netherlands (De Graaf & Allouch, 2016) and 17% in the UK (Nesta FutureFest, 2016). Individual differences, lack of rigorous definition of sex robots as well as the framing of the questions might have contributed to these variable numbers. Despite the inconsistent figures in attitude surveys, experimental studies have revealed some evidence that we humans might be aroused by artificial sexual partners. Using affective priming tasks, researchers at University Duisburg-Essen (Szczuka & Krämer, 2017) found that for heterosexual men the concept of attractiveness was equally strongly associated with women and with female-looking robots. The authors explained this result by arguing that the visual cues of the depicted robots (e.g. breasts) activated deeply rooted perceptual mechanisms and associated reactions towards unambiguously female cues. Along the same line, touching a socially less accessible part of a humanoid robot (e.g. the equivalent of its genital area) was found to be physiologically more arousing than touching a more accessible part (e.g. its hand) (Li, Ju, & Reeves, 2017). These initial findings in favour of a possible emotional tie between humans and humanoid robots will need rigorous further testing before any firm conclusions can be drawn. Many basic questions, for example whether human–robot intimate interaction can elicit physiological arousal similar to that when we interact with human sex partners, can be subjected to empirical testing.
4 Need for a Scientific and a Transdisciplinary Approach
By a scientific approach, we mean an empirical approach that respects observation, experimentation and critical reasoning in understanding human sexuality. Alternative approaches such as religion, philosophy and political institutions will also try to understand this phenomenon by applying their own set of assumptions and procedures because this topic is unprecedented and emotional. We believe that it is time for human–robot intimate relationships to be set up as a serious topic for scientific investigation and encouraged for pursuit. This is not only to satisfy our scientific curiosity, as we know little about this phenomenon and cannot currently make evidence-based decisions. It also concerns the fact that public debates, robotic design and clinical interventions all need guidance from empirical evidence. Systematic funding is urgently needed to support establishment of a research community and transdisciplinary empirical research projects. By a transdisciplinary approach, we mean that all parties involved, including psychologists, philosophers, computer scientists, ethicists, industry professionals and media representatives, need to work together. Furthermore, the role of industry professionals and the adult entertainment industry in advancing and shaping social behavioural changes, as well as their role in acting as potential funding bodies, cannot be underestimated.
Scientists recently held round-table discussions at the University Duisburg-Essen in Germany with the aim to develop a Special Priority Programme in the DFG (Deutsche Forschungsgemeinschaft, German Science Foundation) on the theme of ‘sexuality in the digital context’ that includes human–machine interaction in a sexual setting. It has been a very exciting development for scientists who are interested in the development of this area of work that the importance of this topic has been finally recognized.