Illustration: Jim Cooke, Gizmodo
The sex robot community — the people who make the sex robots, and the people who want to have sex with the sex robots — suffered a blow this past week, when the Houston City Council voted to preemptively ban what would’ve been the first sex robot “brothel” in the US.
But even those council members must know that their gesture was futile. Soon the stigma will fade, and Target will sell these things in 60 different flavours. Which of course means that, sometime in the future, you’ll almost certainly be able to buy a BDSM robot.
As repeatedly pondered over Twitter, before you can get yourself sexually trussed, whipped or choked by a large piece of machinery, we as a culture will need to reckon with — among many, many other things — Isaac Asimov’s Three Laws of Robotics. These laws state explicitly, right at the top, that “a robot may not injure a human being”.
An originalist interpretation would lead one to conclude that Asimov would not be down with BDSM sex robots — but it’s hard to imagine he had them in mind when he drafted his famous rules.
Asimov’s been dead for a quarter-century, so for this week’s Giz Asks we surveyed lawyers, ethicists, computer scientists and philosophers on whether or not a BDSM robot would violate his first rule. Robots are perhaps too dumb at the moment to engage in fetishistic nuances of the human psyche, but the question becomes more complicated as technology advances, as technology always does.
Research Fellow, Ethics + Emerging Sciences Group, California Polytechnic State University, whose research examines how people interact with emerging technologies
The nature of my job is to predict ways humans will interact with a given technology — in this case, a robot designed for a sexualised situation.
One of my first questions about a robot designed for BDSM interactions with humans would be to ask about the definitions of injure and harm from the robot’s perspective, and how the robots will accurately identify those conditions in people. Both of those words have multiple meanings that include very nuanced physical and emotional ramifications.
The robot will need to understand human emotions, which can be communicated as vascular response in different parts of the face, body language, voice tone/tempo/loudness, paralinguistic use, and all the many ways that people communicate beyond speech that are integral to sexual experiences.
But all of these things still don’t equate to a humanlike cognitive process in the AI, one that understands things like situating human emotion within different cultures, or the broad spectrum of human sexuality.
The foundations of healthy, happy, satisfying and pleasurable sexual experiences are trust, effective communication, and of course, consent between people. The role of consent for the human in any situation is physical and psychological safety.
At the same time, human sexuality includes many behaviours that rely on someone’s interior life and what uniquely excites them, and for some people that includes role playing and other creative interactions that sometimes involve testing and teasing physical and emotional limits of the body with their trusted partner(s), or practising things that society may consider taboo.
However, in this case we are discussing scenarios between a (let’s say non-sentient) robot and human(s), so the idea of consent should be human-centred, as in the Laws.
Furthermore, in BDSM the robot or the human could be in roles that call for behaviours that involve risk to themselves or the other — again, those risks and their potential outcomes are different for humans versus robots, so the AI needs to differentiate between those concepts, too, a boundary between its robot Self and human Others.
How will this robot encourage trust with the human? How will the robot recover if they have violated that trust, so they can recognise that failure and work toward the goal of re-building that trust? How will it recover in a way useful to the human if they have misinterpreted their state of being (injury/harm or safe)? How will a robot identify consent? Or play? Or acting?
Will this AI favour able-bodied people in its interpretation of injury or harm? How will the robot evaluate for individual fragility of people? For example, will the model of injury/harm evaluation be different for adults 18-35 versus seniors, or can the robot learn to predict potential outcomes based on individual health histories?
These concepts are all related to the central idea of accurate communication between the human and the robot. All of this is a lot to ask of current AI.
If a robot could understand ideas of harm and injury in a nuanced and humanlike way that included all of the adjacent concepts outlined above, then no, a BDSM robot will not violate the Laws from the human’s perspective in a consensual and safe situation.
But, lacking these abilities, or deep understandings of the people they interact with, then the robot may violate the first Law because it inaccurately assesses the situation.
Cansu Canca, PhD
Founder and director of AI Ethics Lab
BDSM sex robots do not violate Asimov’s First Law of Robotics unless we think that either people who engage in human-human BDSM harm themselves and others, that or robot-human BDSM would have different effects than human-human BDSM.
By design, acts in BDSM might involve physical and/or emotional constraints and discomfort. However, equating that to overall harm would be a very limited understanding of the concept. Dental procedure or endurance sports also often come with physical/emotional discomfort but benefit peoples’ overall physical and mental well-being.
Consensual BDSM uses physical pain as a way to achieve pleasure, satisfaction and happiness. For that reason, in the absence of counter empirical evidence, we’d say that people who engage in BDSM do not harm each other, themselves, or others in the society.
Following the same line of thought, a robot that engages BDSM would not be harming anyone. There are, however, a number of points where the involvement of the robot might make a difference.
One of them is the nature and effects of engaging in BDSM with a robot. While one could expect human-robot BDSM — not involving another human — to be safer and perhaps more liberating, it might also have negative mental effects by taking away the social and interactive aspects of the sexual engagement and excluding interpersonal trust.
Another one is a robot’s "consent". If robot-human BDSM does not ensure that the robot is programmed to act as if the act is "consensual", then the interaction might slip from BDSM to the human engaging in a form of rape. That would take us to a whole different set of ethical questions.
And finally, we have to remember that with robots come opportunities to collect, store and share intimate data about people. Ensuring privacy is extremely important in building systems that engage with people in intimate ways.
Professor, Virtual Reality and Artificial Intelligence, Heriot-Watt University
We should remember that Asimov wrote a whole set of stories around the ambiguities and conflicts inherent in his three laws. In the first law the word ‘harm’ stands in for many centuries of philosophical discussion, and so an answer here hinges on what we mean by that word.
One definition of artificial intelligence reads: ‘the ability for a computer to carry out tasks which are said to require intelligence when humans do them.’ So one could rephrase this as ‘harm is defined as any action carried out by a robot which would be called harmful if carried out by a human.’ By this definition, if BDSM is not harmful between people, it is not harmful between a person and a robot. The protocol humans use is to have a key word which allows their partner to distinguish between masochistic appeals to stop which can be ignored, and a real request to stop. In principle, that could be carried over.
However the suggested definition ignores that fact that a robot is a piece of machinery and not a person. In general, using a person like a machine in sexual interaction does cause social as well as individual harm since it allows classes of people—usually women —to be dehumanised and reinforces self-centered and selfish views of sexual gratification. From that angle a BSDM sexbot is no different from any other sexbot. A sexbot is after all merely a mechanised version of the blowup doll.
Finally, I doubt whether anyone close to the technology would feel happy about risking tender parts of their anatomy against the profound limitations of such machinery. Actual robots are pretty incompetent compared to the fantasies of film, and one might not want to allow one’s bodily safety to depend on the accuracy of a speech recognition system.
Bethany Monaghan Wilson
Graduate student, Law, University of Birmingham, whose thesis focused on on sex robots and the law
You can argue that there’s a difference between “pain” and “harm”, and that BDSM seeks to cause the former. In that sense, a BDSM robot would not violate the first law of robotics, which specifically mentions harm.
But sex robotics as they exist now are problematic, because we’re not yet at the point where a sex robot could see the difference between the two. The most advanced sex robots implemented with artificial intelligence can only remember simple facts about their users. Therefore it could be fair to say the robot wouldn’t know any limits, and so would violate Asimov’s first law.
We’d need to see significant advancement in AI for BDSM robots to function properly. The sex robots we have today only attempt to move their joints and head; human beings are largely more powerful than they are (at least, as of now) so physically a sex robot could not injure a human being, at least not beyond the extent that the human being wants to be injured.
If sex robots do evolve to be more powerful than humans, they might violate the first law — but at the same time they would then be advanced enough to distinguish between pain and harm, which would prove a barrier to violating it.
Psychologically, though, a BDSM robot could violate the first law, as the user might become isolated from the "real world", depend too much on their robot for pleasure, and stunt their social development. On the other hand, the BDSM robot could empower the user to feel pleasure that they cannot with a human being.
Though some scholars who are against the use of sex robots (Campaign Against Sex Robots) would say that a BDSM robot violates the first law by its very nature.
Associate Professor, Law, University of Washington, and faculty co-director of the University of Washington Tech Policy Lab
As far as I know, Asimov was not trained as a lawyer. If he were, he might have drawn a distinction between a prima facie or “on its face” violation of the Laws of Robotics and a violation that ultimately requires a remedy. What you describe may be an example of the former but not the latter.
Consider the question of whether a boxer has committed battery against her opponent during a match. The law defines battery as unwanted physical touching that is harmful or offensive. Neither boxer wishes to be hit and of course it hurts to be. But no court would find battery because, while the act of hitting another boxer may appear to be battery on its face, the defendant would quickly note that her opponent expected and consented to being hit by stepping into the ring.
Similarly, while the robot may be inflicting pain in contravention of the admonition that a robot “not injure a human being”, ultimately the robot is doing so at the behest of a person and for the ultimate purpose of pleasure. (I will withhold judgement on whether robot BDSM implicates Zeroth Law).
Meanwhile, note what happens if we don’t make this distinction: A robot would not be able to drag an injured person out of a burning building because being moved is painful. That can’t be right, or at least it shouldn’t be how we program our machines.
Professor, Politics and Philosophy, University of Warwick
Sado-masochistic sex between consenting adult people can involve injury, and the injury can be enjoyed by participants. What is more, injury isn’t just incidental to sado-masochistic sex, but what sets it apart from tame varieties. It is certainly conceivable that injuries inflicted by a BDSM robot at the command of someone who likes sado-masochistic sex would be enjoyed by the person commanding it.
Some philosophers (e.g. Thomas Hobbes) say that what is done with one’s consent can’tbe an injury. But this view makes it hard to state the distinguishing features of sado-masochistic sex. It is more natural to say that sado-masochistic sex often involves injury and that robots commanded to engage in it cannot at the same time obey Asimov’s laws.
Professor of Philosophy and Director of the Ethics + Emerging Sciences Group at California Polytechnic State University
Technically, yes, anything that a robot does (or fails to do) that harms a human would violate Asimov’s first law of robotics. But this is true only if we understand “harm” in a naive, overly simplistic way. Sometimes, the more important meaning is net-harm.
For example, it might be painful when a child has to have a cavity drilled out or take some awful medicine, but we understand that this is for the child’s own good: In the long term, the benefits far outweigh the initial cost. We’re actually trying to save her from a greater harm.
This is easy enough for us to understand, but some obvious concepts are notoriously hard to reduce into lines of code. For one thing, determining harm may require that we consider a huge range of future effects in order to tally up the net-result. This is an infamous problem for the moral theory of consequentialism that treats ethics as a maths problem.
Any harm inflicted by a BDSM robot is presumably welcomed, because it’s outweighed by a greater pleasure experienced by the person. A BDSM robot would seem to inflict harm onto you, but if you had requested this, then it wasn’t wrongfully done.
If the robot were to take it too far, despite your protests and without good reason, then it’s wrongfully harming you because it’s violating your autonomy or wishes. In fact, it’d be doubly wrong, since it violates Asimov’s second law, too.
But assuming the robot is doing what you want, the pain inflicted is only technically and temporarily harm, but it’s not harm in the commonsense way that Asimov’s law should be understood.
A computer, of course, can’t read our minds to figure out what we really mean — it can only follow its programming. But ethics is often too squishy to lay out as a precise decision-making procedure, especially given the countless variables and variations around a particular action or intent. And that’s exactly what gives rise to the drama in Asimov’s stories.
Professor, Science and Religion, St John’s University and the College of St Benedict
First, I’m guessing a good BDSM robot will be challenging to produce. The sex robots of today are rather passive. I’d say there’s yet to be a good male robot that isn’t much more than a fancy vibrator.
For a BDSM robot to take the dominant role in a sexual interaction — well, I just don’t see that as being right around the corner. Being a good dominatrix takes inventiveness and a certain degree of sensitivity to the submissive’s reactions and desires.
To the question: Whatever happens is consensual so it’s not hurting the submissive, but that’s a bit too simplistic. Most BDSM relationships have a safe word (let’s say “eggplant”) that the submissive can use to stop whatever is occurring (since “no” or “please stop” is often part of the role playing). This would be easy enough to program in.
However, there are always unforeseen circumstances where the submissive can’t use the safe word and depends on a certain degree of trust in the dominatrix. This could result in an unsafe situation with a robot. One might elect to use such a robot under surveillance (say in a brothel), but who wants to have kinky sex under surveillance?
That said, I’m not a big fan of Asimov’s laws. They don’t work.