Human Rights of Users of Humanlike Care Automata

For Ash-like mechanical beings, see Hanson 2011; for various kinds of cyborgism, see Hughes 2004, Maxey 2013.

Constructing fully biological humanlike robots seems to pose more blatant ethical problems (of experimentation and experimental development, for example), which respond more readily to ethical assessment. I also leave out discussion of ethical and policy issues arising from the third type, cyborg construction, because of the enormity of problems that their middle ground invokes.

See Miller 2015 for assessment of many of these works, particularly those assuming that tools should not be treated as objects created for a purpose.

For more information on the definition, characterization, and investigation of maximally humanlike automata, please see Miller 2015, 2017.

This openness does not deny the uncanny valley phenomenon (see subsection labeled “Four: The Mori Effect” below), by which there is a phase of technology development whereby an automaton’s humanlikeness repulses the user, but when the automaton becomes extremely humanlike, the effect starts to reverse. See Ferrari et al. 2016.

One may object here that as techniques become more sophisticated and humanlike, there may be less ethical worry for care automata because the machines will be better made, more competent in handling the cared-for, less prone to error, and more ethical themselves, In fact, they will be as loving as any human can be, as conscious, properly emotive, companionable, and kind. Thus, the machines would need only face the same ethical and legal constraints as any human. However, as seen later in this article, this increasing humanlikeness still brings in ethical and legal issues for Homo sapiens societies over and above those that Homo sapiens must face, if machines are created for a specific purpose, for which humans and their societies were not created (if societies—contrasted with polities—are created for a purpose). I believe this caveat will become clearer in the course of the article.

For the ease of discussion, assume that the elderly person is not, say, a science-fiction fan who longs to live in a highly technologized, fantastical future, but is of more mainstream tastes. Studies have begun into the degree to which cared-for elderly would welcome automaton care (at the current technical level) (Van der Plas et al. 2010), and there is plenty room for further such research.

Hanson (2011) makes an art-for-art’s-sake case for MHAs, while adding that via such development, scientists can have a means by which to understand the human being in ways not possible without such automata research.

A practical issue then arises from the possibility that a high percentage of elderly under care may not want automata, thus diminishing the program’s economic justification. I bypass this side issue here.

Even if the cared for were sufficiently apprised of the caregiver’s ontology and still acceded to the automaton caregiver, a deception would be involved, although the cared-for would not disallow this deception.

Coeckelbergh (2012) covers a slightly different angle concerning emotional deception in automata and is worth further attention. While he is considering automata with built-in emotional responses that may merely appear to the user/patient as bona fide emotion when it may not be, this article is concerned about the emotional responses in the user/patient that are induced by the automaton’s overall biological and behavioral humanlikeness.

This statement need not counter notions of technological mediation as proposed in post-phenomenological works on philosophy of technology, contending that proper understanding of a technology cannot be had by looking at the purposes and motivations behind a design, but rather through how the object mediates the person and the world. (See Verbeek 2005, Ihde 2009.) The point here is that the designers and manufacturers commonly fashion a design to—in their minds—facilitate a human activity.

I discount for now the possibility that some persons in elderly care would be fascinated by having automata caregivers than human.

There may be exceptions to such social organization, such as some ethnographically reported Caribou Aleut groups (Burch and Csonka 1999), which seemed to be organized in societies running up to the hundreds in population but spread over large territories. Arctic cultures, given their harsh environments, tended toward inegalitarian, hierarchical organization and may have tended, more than other types of foraging cultures, toward infanticide and geronticide (Lee and Daly 1999).

The correlation is indicated as indirect because, as the National Cancer Institute’s literature observes, studies about direct link between cancer and stress are unclear; but more clearly, high stress often leads to patients’ turning to carcinogens such as tobacco or high alcohol intake, thus confirming an indirect link.

Ironically, these values likely came to be valued in their origin for the sake of serving life in and of itself, although finally taking priority over life itself.

Before proceeding with rights and policy, I should mention companion automata, for friendship or romance, only highlighting the rights issues here. Hauskeller (2014) and Miller (2016) inquire whether these devices invoke human-rights issues. Certainly the United Nations Declaration of Human Rights (UNDHR, 1948) strongly justifies a right to choose whom one wants as a companion. There are problems, though, with invoking a right to an MHA companion. It remains unclear whether one has a right, tout court, to something that does not exist and may at best only conceivably exist. (The same problem arises for whether one has a right to an MHA caregiver.) No companion MHA exists. It would be absurd to insist that in general we have rights to things that do not exist, as there are a potentially infinite, nonexistent ways to ensure, say, health, and it is beyond the call of rights for every person to deserve these infinite nonexistent measures. During the interval from now till the day of successful MTT and MHA, there is no valid rights claim to such a nonexistent object. It is also not clear whether one has a valid rights claim to bring such an entity into existence and deploy the particular kind of techniques required. I have contended that caregiver MHAs could bring to the surface and instantiate values that threaten human rights. Deployment of these MHA techniques could thereby engender rights violations. The problem would be that the techniques for the care MHA would be so similar to those for companion MHA, that manufacturing the latter type would be tantamount to enabling deployment of the former, which I contend could be deleterious for the rights of the cared-for. (See next section.)

Furthermore, empirical evidence indicates there are other threats to the cared-for and by extension users of MHA companions. A study by Ferrari et al. (2016) involving psychological tests found that the more humanlike the automaton, the more subjects felt their human distinctiveness threatened, with “androids”—those exactly resembling humans—evoking the strongest repulsion. Furthermore, if such threat could be considered as a loss of security, an MHA could be further construed as offering yet another type of threat to human rights. (At least in terms of resemblance; as Ferrari et al.’s two studies were based on physical appearance, not behavior.) Certainly, there could be a factor of humans’ becoming habituated to the presence of more and more humanlike automata over the years, so they would feel less threatened. (See Vallor 2016 and discussion of the Mori effect above.) But it is unclear why designers and manufacturers would feel compelled to keep teasing (to put it lightly) or baiting human responses over the years, just in case humans were to habituate. Rodogno (2016) raises a further factor for MHA companion proposals, bringing out how a type of action viewed only in isolation may not evidence any apparent moral problem. But once an action is undertaken by massive numbers of agents, a moral problem can start to arise. He provides the example of the automobile, which in 1900, with so few around, could evoke scant moral concern. But over a century later, with upwards of a billion cars in operation and the massive impact on health and the environment, moral concerns spring up. One or two companion MHA may pose no evident moral concern; billions of them in use could.

A third problem with the case for companion MHA is enigmatic. It is hard to account for just what “consent” of the automaton would amount to in practical terms (Miller, 2017). If the user buys the MHA from a supplier who programs it to ensure giving informed consent to—or even falling in love with—the buyer, then the question of obtaining informed consent is moot; the machine has no choice at all but is made to agree to the buyer’s advances. However, further, abusing the machine also seems questionable, even if programmed to accept abuse (Fox 2018): The problem of what, in a human/MHA relationship, consent would consist in points up how obfuscating and unconvincing it is to defend, so prematurely, potential users’ having a prima facie right to a companion MHA.

Mechanical military MHA are another conceivable type of MHA, although for what purpose the military should make an MHA is not currently clear. (See Maxey 2013.)

There are some differences here between basic rights and what I called “foundational” rights. I suggest that rights to life and liberty are even more basic than his basic ones, even by his definition, as one cannot enjoy the right to assembly or expression unless one is alive and free to attend the assembly or write articles or to eat and have property (even if in oneself) to be protected.

Alternatively, it is possible that machines such as MHA themselves should be granted some rights, though not the full panoply of human rights. As Darling (2012) and other writers point out, there could be good socio-psychological case why allowing an owner to an MHA to abuse it may in itself create more serious moral ramifications. (One idea about the moral concern here is much like that in the early years of animal-welfare protectionism, when it was observed that abusing animals is a good route to abusing humans. See also Rollin 1989.) If the device is indeed sentient, it would be immoral to abuse it. To sum up a topic that deserves an entire article, see Miller 2017).

Fissile substances such as uranium-238, nuclear-construction materials, and nuclear reactors, as well as smallpox and other dangerous microbes, are controlled by national and international bodies, partly to protect the global human right to security, life, and health. These are materials and techniques designated as not fit for private, everyday use; and one would be hard-put to contend that private citizens’ rights to possess these materials overwhelm the basic right to life and health. Rights to do what one does in private are limited if that activity must entail uses of materials and techniques that may endanger others’ rights.

One problem with the MTT is that, even if the tested automaton passed, the machine may still not be conscious, intelligent, rational, or even sentient but only well-imitating behaviors that in humans we believe are driven by such traits. The idea of the original Turing Test was that if the machine uses language so well that only (artificial) intelligence equal to or surpassing a human’s could explain the result. However, this linguistic-only test merely begs the question of what intelligence consists in and the degree to which language-use reflects it. The MTT is supposed to help eliminate “cheating” by mere imitation, but perhaps even sentience and consciousness may be merely imitated.

Even if these forces have grown more formidable in recent years, if it is indeed morally right to seek policy control, counterforces should not spell a reason to forsake policymaking. Nevertheless, the appearance of groups such as the Future of Life indicate how many eminent persons are coming to see the need for oversight of such new types of powerful techniques.

One further ethics issue to consider: If automata are economically more efficient than human caregivers but the cared-for person is repulsed by an automaton caregiver, would this person be penalized for having “expensive tastes”?