Taking Care With Caregiving Robots

It is important to get a clear sense of just how many robots are already out there and on the job.

Robots may never be as good at caregiving as humans can be, but—given the huge demand and low supply—they may be good enough.

Though Japan is leading the “carebot” charge, many countries all over the world are not far behind. These robots may range in size, appearance, function, and purpose, but they all serve to fill caregiving gaps. BeanQ, for example, is made by Roobo in China and has been called “childcare for the masses” [14]. Designed to be “an early educator, sharing some of the parental burdens,” BeanQ is reportedly already used in millions of middle-class Chinese households. The robot responds with simple words and phrases and an array of emoji facial expressions displayed on a large screen that serves as its face. It also has a “remote babysitting” mode, automatically taking snapshots of the kids and uploading them for their parents to see online. The Care-o-bot, from Fraunhofer IPA in Germany, is at work in a number of German-assisted living facilities [15]. Like the social directors, aides, and volunteers typically employed in such facilities, Care-o-bot does not just bring meals to the residents, but also plays memory games with them to help keep their minds sharp. The Dinsow eldercare robot, from CT Asia Robotics in Thailand, helps seniors remember to take their pills, tracks their health, and serves as a videophone for calls with family and doctors [16]. It entertains, too—doubling as a karaoke machine. ElliQ, from Intiriont Robotics in the United States, is a social robot companion that engages its owner throughout the day, providing appointment reminders and assisting with video calls to friends and family. It can suggest a walk when a human is overdue for some exercise. ElliQ can also share a spontaneous fact or humorous video [17].

There are many, many more on the way or already on the market. Pricewise, they range from a few hundred dollars to hundreds of thousands of dollars. As with every other high-tech product—from televisions to smartphones—each new version will likely be more alluring, more dazzling, and more affordable than the last. According to the McKinsey Global Institute, the average robot price fell by half in real terms from 1987 to 2018 and fell even further relative to the cost of human labor [18].

A growing and global demand for care that cannot be met the old-fashioned way, then, is being met in a new way. Is not that how it is supposed to work? Why not just lean into, “robots for everyone!” and get on with it?

Doing something extremely important in an extremely new way without giving it the necessary thought or testing before implementation tends to be a one-way road to disaster [19]. (Think, for example, of the unforeseen role of social media in political manipulation.) Some caregivers are already dead-set against the use of caregiving robots in any form. Others are skeptical and worried, but not unequivocally opposed. Several specific concerns have been raised by caregivers [20].

• Inability to give real love.
• Inability to fully decode emotions. One mother worried that robots will never understand “the jumble and mixed emotions [that] young children so often display and parents [only] slowly come to recognize.”
• Lack of human fragility. As one father explained, “A robot is intended to be 100% correct. But if [robot] caregivers don’t show any human fragility, then you are depriving the child of some human characteristics that will need to be learned later in life.”

There are a number of threats to automating caregiving, but perhaps the greatest underlying threat comes from modern society’s widespread failure to acknowledge how central it is to being human and living in human societies, to care for others, and to be cared for by others. Think of it this way: We may very well soon have the technology to keep the human race procreating at whatever level we want without anyone ever needing to have sex or for that matter become pregnant (think artificial wombs). But do we all really want that “convenience” as a normative default? In the same vein, if we rush headlong into the convenience (or seeming economic necessity) of robot caregiving—efficient, affordable, available—the empathy that is crucial to caregiving could be entirely lost. That will pose risks—some direct, others indirect and long-term—to the people being taken care of.

First, we have no right to experiment on a couple of generations of children so that the developers of caregiving technology can “perfect” their offerings. These technologies are simply too new to be proved. While each company will produce a pile of data to support its offerings when (if) regulation of the industry sets in, the science will necessarily be provisional; there are obviously no long-term studies. Even if regulatory agencies are eventually engaged in oversight, they do an imperfect job, sometimes missing real risk.

Second, we have to keep in mind that caregiving robots are being developed largely by for-profit interests or by educational institutions with an eye toward commercializing them. Their primary motives are growth and profits—period. When it comes to presenting data on the robots’ efficacy and side effects, they will cherry-pick the best outcomes and suppress any ominous results as long as they can. From cigarettes to seat belts, from vaping to addictive iPhone apps targeted at children, there are innumerable examples of this type of corporate behavior. We would be crazy to trust that these companies have society’s best interests at heart.

Third, there can be incredible risks when technologies are used in unanticipated ways. Television was not invented to keep kids quiet. Facebook was not created as a news source, let alone a political power broker. Similar unintended possibilities for caregiving technology are endless, ranging from the benign if misguided—too much screen time—to the downright evil—direct marketing and political or social indoctrination.

Finally, one overall advantage of human care is the well-documented fact that someone else caring about you can itself have healing powers [21]. Can the fact that some thing (rather than some one) takes care of you have that kind of effect? We certainly do not know and we would be wise to assume not until proved otherwise.

Robot caregiving needs to be approached with humility and patience, which have so far been lacking in the introduction of far-reaching technologies. Online social networking was introduced entirely on the basis of its creators’ (often narcissistic) dreams. Some of those dreams were quite high-minded, but none was buttressed with any large-scale testing or accumulation of experience. The result: social media has already taken over much of our collective life, with enormous social, emotional, political, and ethical issues barely addressed. For example, we have grown comfortable with dramatic increases in rates of mental health ailments and social isolation among vulnerable populations in our society [21]. Social media are not entirely guilty, but they certainly are not entirely innocent [22], [23]. Given that robot caregiving touches on matters every bit as fundamental to human life as social networking does, we can expect it to present us with equally fundamental issues. What say this time around, we address them before it is too late?

Business priorities will not encourage a holistic development and roll-out of caregiving robots. For businesses, growth and profit reign supreme. If society does not make its own decisions about robot caregiving intentionally—and mighty quickly—the tech industry will have made them for us without our even realizing it.

What will become of the natural—and noble—human impulse to take care of the needy if technology is always there as the first, easiest, and cheapest “solution”? Do we really want to outsource this cornerstone of our humanity?

Consider again the “demand” that seems to be driving the supply of robot caregiving. How real is this imbalance? Yes, the global demographics of some aging societies are real enough, but is the supply of caregiving really as inadequate as it seems? To some extent, it is actually artificially low. There could be a lot more human caregiving than there is—even if that still left a shortfall to be filled by robots—if only we treated caregiving as highly valuable and compensated it accordingly. Nothing gets labor supply off the sidelines like paying more for what you want. Good caregiving is a skill—or a gift, however, you want to see it—and it is one of great value. Spend some time in a nursing home or daycare. Watch the best child care workers, nurses, therapists, aides, and teachers. Pay attention to how loving they are to people they do not even know or have only known a short time. Try to understand how they figure out what someone who cannot speak comprehensibly wants and needs. Imagine how they keep their good nature even when someone is being belligerent. Then ask yourself: Could I do all that?

If our society provided greater support and our employers were not quite so greedy with our time, it would be easier to see that there is not a true caregiving shortage. If we hand caregiving over to robots without any change in the way society values caregiving, is there any reason to think that government and employers will have higher standards for the care provided (say, by insurance) to our friends, family, and loved ones than they do now?

As long as we keep the risks in mind, robots can play a healthy role in caregiving. They can—and should—wash the dirty diapers, but they should not give the baby the hug before plopping her on the changing table, nor should they be the ones to kiss her little belly button when she is all clean. In short, robots should support human caregivers, not supplant them. One caregiver cut right to the heart of whether human caregivers and robot caregivers are true substitutes for each other: “There is a difference between care and caring!” [24] an IV drip takes care of you. The nurses who change it cannot do what the IV does, but they care about you, which the IV drip cannot. The very best caregiving may result when both are present.

Robot caregiving needs to be approached with humility and patience, which have so far been lacking in the introduction of far-reaching technologies.

People are prone to empathize with humanoid robots, much as we empathize with other people [25]. We can even come to feel that robots love us. Though this may sometimes be comforting, it leaves us very vulnerable to the robots’ creators and programmers.

But robot caregivers may have particular strengths compared to human caregivers, too. First, people are susceptible to sentimentality. This, of course, can be a positive aspect of caregiving, but it also has its downsides. Sentiment can override other concerns, sometimes even the wishes of the individual being cared for. Humans are also capable of negative sentiment, but robot caregivers can likely—though not at all assuredly—be programmed to not take a dislike or bias to any particular person or group of people and treat them unfairly. They will not feel insulted, put upon, or unappreciated.

Of course, to view robots’ lack of sentiment relative to human caregivers as a virtue is a complicated call. Humans exercise great discretion in their caregiving; for example, allowing an older adult occasional splurges outside their prescribed diet or allowing a child to stay up late for something special. There is no form of AI yet capable of that mix of discretionary behavior. And even the display of negative sentiment may, in some caregiving situations, be valuable, helping those cared for to learn the limits of what they can (and should) expect from others. Yet, there are situations—for example, late-stage Alzheimer’s disease—in which human sentiment can be so stretched that some support from an unsentimental robot caregiver might be a valuable aid to the caregivers and the care recipient.

A second potential strength of caregiving robots relative to human ones is the fact that robots are no more liable to error than humans—and less liable in many cases. Robots do not get tired, cranky, or bored. They can and do malfunction, but they do not get sick. They can work through the night.

Given a sufficient supply of robots in a given facility, the same robot could attend to a particular patient indefinitely, which could never be the case for, say, a hospital nurse or nursing home aide. They could have access to all medical records all the time and, in fact, to more or less infinite medical data. Properly programmed robots will not forget one task due to a sudden rush of other tasks; unlike most of us, a robot could be suddenly diverted to assist in an 8-h surgery and then remember to go back to a patient’s room and pick up a Kleenex off the floor. They can also “move-in” to someone’s home much more easily than a human caregiver can, which may make it easier for more older adults to stay out of the nursing homes they often dread.

Robots can assist human caregivers with heavy physical tasks, such as rolling someone over in bed or even just picking up a baby. They are much less prone than humans to mechanical strain and, in any case, they can easily be repaired or replaced should they break or malfunction. They also cannot catch any diseases from humans or medical equipment such as needles. Depending on their outer materials, robot caregivers would still be capable of carrying and passing on at least some communicable diseases, so even robots might need to wear gloves, but they would likely never find them annoying or forget to change them or try to hoard them.

These advantages are mostly technical, but we also need to consider some more profound issues: what are the ways in which robots can replace human caregivers and what are the ways in which they cannot—or should not? Looking at another way, the issue is: how do we safeguard a role in caregiving for humans, in a world of caregiving robots?

The best approach to future caregiving is not to ditch the humans or forgo the carebots, but to create a more holistic robot-assisted caregiving model. The idea is that all tools—up to and including the most sophisticated robots—should augment human caregivers, not replace them. Robots helping out with physical tasks would free up more of the human caregivers’ time and energy for caring and human interaction. Robots might support the mental health of caregivers who may, for example, be struggling to cope with a loved one who has dementia or who is in great pain or very near death. Robot caregivers might provide coverage for human caregivers to take much-needed short breaks or could just lighten the load by changing the bed linen, or doing the darn laundry.

With the idea of robot-assisted caregiving in mind, a number of guidelines with respect to the design and use of the robots themselves should be followed.

The best applications of technology empower individuals rather than bypass them. To that end, the best caregiving will not infantilize care recipients. Robots should be designed to react to the humans receiving care, not to intrude on them or be overbearing. Imagine, for example, a robot that would give older people a reminder checklist of tasks to do in the morning. This would be a mistake. Instead, the robot should notice if its user skips a critical daily routine, such as taking prescribed medication and offers a prompt. In this way, the people receiving assistance from the robot remain as independent as possible, aided by the robot but not subject to it. Overbearing care by robots poses the same risk as overbearing care by a human: a cycle of dependency, in which a care recipient (young or old) constantly needs to be told what to do and how to do it.

Robots should be designed and deployed with great consideration for another kind of dependency: emotional dependency. It will be very natural and even unavoidable in some cases, for people to become emotionally attached to robots and this impulse could be magnified and exploited by design. With wide, child-like eyes or soft fur, many ultracute caregiving robots are purposefully designed to elicit attachment. Such emotional dependence raises deep ethical considerations—Is feeling loved by a robot meaningful?—as well as legal considerations—Can an elderly person, for example, leave a bequest to the robot who cared for him or her? [26] When children are the care recipients, other ethical and legal considerations arise [27]. For example, when—if at all—should a child should know that his or her caregiver is a robot and not a person or a pet? Does a child have a legal right to contact a former and beloved caregiving robot after it has “moved on” to a new child?

Caregiving robots should be customizable. For example, a clear benefit of robots is that they are tireless. But is this trait always desirable? If the person receiving the care is supposed to “graduate” at some point to needing less or no care, the robot itself has to change what it does and does not do, and how much—calibrating care as a human caregiver would.

With no long-term studies of the effects on children or the elderly, robot caregiving should be introduced carefully, a little at a time, and studied closely all along the way. These studies must be performed by qualified researchers (not necessarily academic) who are not connected in any way with the robots’ manufacturers or to other institutions with motivations to justify the investments of the manufacturers’ top managements and boards.

The best approach is not to ditch the humans or forgo the carebots, but to create a more holistic robot-assisted caregiving model.

As we start down a long path of AI caregiving, privacy must be taken seriously this time. One of the major selling points of many caregiving robots is their ability to monitor and record, 24/7. Is that caregiving—or jail? Do we want every moment of our earliest years captured, recorded, and potentially analyzed? We may want our doctors to know about what we eat, our exercise regime, and our compliance with prescriptions, but do we want Acme Robots to know all that?

Whenever robots are used, let us make sure they are part of a larger plan of socialization for the care recipient, a plan that should be routinely reevaluated. Robot caregivers should be part of a broader commitment to ensure that people interact with other humans as much as possible. Many research studies show that human-to-human, face-to-face contact plays a critical role in people’s health—improving mood and lifespan [28] —something we likely cannot get from a robot.

Caregiving robots, when working alongside human caregivers, should have boundaries and there should be a clear hierarchy in place. The human caregiver, not the robot, should be in the driver’s seat. Making a human caregiver subordinate to a robot caregiver not only raises a host of ethical concerns, but also puts their respective strengths precisely backward. Robots are very good at repetitive and mundane tasks. Humans need to be engaged in activities that demand attention and preferably are interesting—those aspects of caregiving should, for the foreseeable future, be left to human caregivers. A human caregiver should be charged with making sure robots are being used wisely, as a tool or medium for support, not vice versa. Robots, even “smart” ones, should not be left to decide what is in the best interest of their charges.

Last, remember that, just like humans, robots are not infallible. We need to have a tolerance for robot malfunctions and to expect and plan for such malfunctions. Humans are trusted with other human’s lives despite being demonstrably imperfect. So are all sorts of medical devices today—and caregiving robots will be tomorrow.

We have a new opportunity to use AI to help reduce disparities and improve care overall. Will a given robot be able to make people of various genders, races, sexual orientations, and other backgrounds equally comfortable? Or will it be designed by a homogeneous group of engineers for a dangerously narrow group within our society? Human caregivers are prone to some racial bias in recognizing (and as a result, fully treating) patients’ pain [29]. Will this bias be carried over into caregiving AI? We already know, unfortunately, that facial-recognition systems are more likely to misidentify people of color than White people [29]. Such failure becomes extraordinarily troubling in the case of caregiving robots. Technologies must be designed for the diverse constituencies making up our society [30].

Finally, class dynamics need to be considered. The digital revolution was heralded by many as a great equalizing force. Yet we have seen one digital divide after another emerge, with the wealthy having much greater access to everything from high-speed internet to technologically advanced cars. Is it possible that those who can afford human care will receive it, while those who cannot will be stuck with robots? Or will working-class children grow up raised by low-end, $ 100 caregiving robots, while wealthy children are raised by a mechanical Mary Poppins?

It is likely to be a rocky road. Have you tried to pick out a television lately? Or decide on the right phone? Imagine what it is going to be like picking out a caregiving robot to take care of your loved ones. We would do well to take precautions now, as the technology—as it is prone to do—is moving fast.