Possibilities and ethical issues of entrusting nursing tasks to robots and artificial intelligence

Introduction

Elements of nursing practice that can be entrusted to robots and artificial intelligence (AI) are changing according to the times and social conditions. It is possible that in the future, robots will account for a greater proportion of tasks in nursing practice than ever before. If so, what are the things that cannot be replaced by robots or AI? Or is there no such thing in nursing practice?

Investigating these issues, this paper examined whether there are virtues or ethos that nurses are expected to possess but cannot be implemented by robots or AI, to clarify the elements of nursing practice that cannot or should not be carried out by robots or AI. Specifically, we will take typical ethical concepts in nursing (advocacy, accountability, corporation, and caregiving) ¹ as an example and discuss whether robots or AI can implement these concepts by analyzing them and the progress of current robotics and AI technologies. These concepts are complex and rich, each is still a subject of debate. Nor are they limited to the ethical concepts that are important in nursing. However, there is no dispute that these concepts are important in nursing practice, and there is also a common understanding to a certain extent, even if there may be disagreement on some details. Since the purpose of this paper was to examine the possibility of implementing important ethical concepts in nursing into robots and AI, these concepts were used as examples for analysis, and rather than examining each in detail, we examined them based on the common understanding.

Background

The science fiction film “Bicentennial Man,” based on the work of Isaac Asimov—a science fiction writer who is also famous for his advocacy of the three principles of robotics—is thought-provoking as it asks how robots with person-like abilities should be treated in the future. In the last scene of the film, the former robot Andrew, who has been granted various rights as a human Rebeing, draws his last breath. At this time, his female partner requests a nurse named Galatea to stop the life-prolonging treatment. Galatea acquiesces to her request. As audiences knew that Galatea had also been a robot (even a robot-like robot), in this scene, the audience realizes that the life-prolonging procedure was stopped by the robotic nurse.

Since the Three Laws of Robotics (Table 1) prohibit robots from harming humans, this scene can be interpreted in two ways. One interpretation is that the cessation of life-prolonging treatment is never harmful to a terminal patient. The other interpretation is that Galatea, who was given the same name as the life-giving goddess figure in Greek mythology, was also to be treated as a human, like the main character Andrew, and therefore made her own ethical decision and practiced nursing like a human nurse, without being bound by the Three Laws of Robotics.

OPEN IN VIEWER

Table 1.

Three Laws of Robotics proposed by Asimov

1. A robot may not injure a human being, or, through inaction, allow a human being to come to harm.

2. A robot must obey the orders given to it by human beings except where such orders would conflict with the First Law.

3. A robot must protect its own existence as long as such protection does not conflict with the First or Second Law

Currently, significant progress is being made in the research and development of robots and artificial intelligence (AI), and it has been suggested that these may play a part in medical and nursing practice in the future. However, some are of the contrary opinion and think it is difficult for them to perform some parts of the daily care provided by nurses,^2,3 such as taking temperature and pulse measurements every morning. Temperature and pulse checks are routine tasks performed by nurses, who could be replace by wearable devices and AI systems to monitor patients. In practice, research on the robotization of nursing tasks, known as “auto-nurse,” was conducted from the 1960s to the 1970s in Japan. However, this research was abandoned in the 1970s. It has been suggested that the reason for this is that automating temperature and pulse checks reduced the opportunity for nurses and patients to talk. ⁴ For those who argue that the nurses’ emotional connection with the patient is recognized in routine tasks such as temperature and pulse checks, and that these tasks are an essential part of the nurse’s role, it is not easy to accept that robots with AI could easily perform the nurse’s duties.

However, what was called an auto-nurse at that time was simply a system that automatically measured a patient’s breathing rate and temperature. Some of these systems are now already in practical use and have become part of nursing practice. Thus, it seems worthwhile for our discussion to preliminarily define the limitations of the nursing robot or AI system that we envision, in light of current developments in robotics and AI technologies.

HOSPI®, an autonomous in-hospital transport robot developed by Panasonic, can automatically transport drugs and specimens within a hospital. It does not require a track on the floor, and its elevator boarding and alighting functions enable it to transport across floors. ⁵ Relay Robotics’ Relay+, ⁶ which has already been put to practical use in hotels and other facilities, is expected to be used as a similar transport robot in hospitals. The COVID-19 pandemic has strained medical resources, and the overwork and exhaustion of nurses has become a social problem, ⁷ for which nurse-assistive robots are expected to be the savior. Moxi, ⁸ a nurse support robot developed by Deligent Robotics, has been installed in a COVID-19 ward and is responsible for delivering and retrieving medical instruments, specimens, and rags, among other things. Moxi is a humanoid robot with a single robotic arm, and its head is equipped with AI and a learning function, which enables it to operate smoothly in new environments. ⁸

Recent developments in AI are expected to make important contributions to the future development of nurse-assistive robots. Although the development of AI for nursing is currently in the research and development stage, some research has focused on its social implementation.^9,10 Predictive modeling is often employed, along with AI technologies such as natural language processing, computer vision, speech recognition, and planning/scheduling. ⁹ Target nursing areas include nursing care planning, disease outcome prediction, health assessment, risk identification/prediction, detection/tracking/monitoring, documentation, and hands-free operation. ⁹ To give an example, a typical use of AI in the nursing domain is in monitoring. AI that predicts falls for older hospitalized patients through natural language processing of medical record information is already in practical use, ¹¹ and research and development on fall detection and prevention systems is also active.^12,13

Expanding the potential for forming more emotional connections with patients is an application of daily conversation AI to the nursing domain. ¹³ Moxi ⁸ can already make facial expressions, and thus, forms an emotional connection with patients; nonetheless, it is feasible to further incorporate conversational capabilities and AI for expressing facial expressions into robots. A wide variety of such communication robots are already available, such as the seal-shaped robot PARO created by Intelligent System Co. ¹⁴

Based on our assumption, robots that are expected to play an active role in the nursing field in the future will appropriately assist nurses in their routine tasks, such as carrying supplies, monitoring patients’ vitals, preventing falls, and having a watchful eye function that contributes to patient safety, as well as forming emotional connections with patients, including through verbal communication. Although such robots (with AI) have various limitations, the speed of their commercialization is expected to increase in light of current developments in robotics and AI technologies. In this paper, we assume that such robots (with AI) are nursing robots, and discuss their uses and changes in the nursing field.

The AI we assume indeed possesses broader capabilities than the so-called weak AI. However, we do not assume artificial general intelligence (AGI), which is capable of solving almost all tasks that humans can solve. ¹⁵ The minimum functionality required for AI in our discussion is a communicative function that allows forming an emotional connection with the patient. So-called strong AI, or AGI, do include such a communication function. Therefore, our discussion is more active and not limited to AGI, although it includes AGI in its scope.

Advocacy: can robots and artificial intelligence be a voice and advocate for patients?

Nurses are expected to be there for patients and their families and to be their advocates. For example, the Japanese Nursing Association Code of Ethics for Nurses states “Nurses should also act as advocates for the rights of clients, including speaking on their behalf, as appropriate.” ¹⁶ The American Nurses Association (ANA) also emphasizes that advocacy is an essential pillar of nursing. ¹⁷ According to a study by Abbasinia et al. (2019), ¹⁸ analyzing the advocacy concept, advocacy has five components: (1) safeguarding, (2) apprising, (3) valuing, (4) mediating, and (5) championing social justice in the provision of healthcare. ¹⁸ The challenge is whether it is possible to implement these elements in robots and AI.

Of the above five elements, it may not be so difficult to implement “Safeguarding,” which protects patients from medical errors as well as incompetency of or misconduct by co-workers and other members of a healthcare team, ¹⁸ and “Apprising,” which provides information to patients about their diagnosis, treatment, and prognosis, suggesting alternatives for healthcare, and providing information about the discharge program, ¹⁸ as functions for robots and AI. Of course, it is necessary to focus on how patients and their families are informed when medical errors are detected and how to inform patients about alternative medicine, for example. Some studies have shown that the way a medical professional handles and communicates a medical error can make a difference in whether or not the situation develops into a court battle. ¹⁹ However, considering that inexperienced nurses are not always able to deliver the information appropriately, robots and AI could possibly learn a suitable way of providing information in the same way that a new nurse becomes a skilled nurse through a combination of deep learning.

However, “Valuing,” which accompanies the patient’s own value judgments, and “Mediating,” which interfaces between the patient, family, and healthcare professionals, are expected to be more difficult to implement than the above two. Even if a robot or AI can learn a patient’s values, does merely tracing them mechanically mean that the robot or AI is attuned to the patient’s values? And while the robot or AI could play the role of transmitting those values to family members or other healthcare professionals, would it be able to play the role of a mediator in human relationships? There have been reports of pet robots in clinical settings, such as hospital wards and workplaces, that have contributed to the facilitation of human relationships, ²⁰ but further empirical research is needed to determine whether robots and AI can truly play the role of coordinating human relationships. There is also the question of whether robots and AIs can provide the kind of communication that nurses use in clinical practice to elicit the patient’s own thinking and sense of values.

Finally, as for “championing social justice in the provision of healthcare,” which calls for challenging the system on issues of injustice in healthcare, it is not difficult to implement if it is possible to program what conditions should be charged as injustice in practice. Of course, the implementation of such a feature would require more detailed research on injustice in healthcare to find an agreed and accepted standard. Conversely, if such criteria were established, robotic and AI nurses could be challengers of such injustices, unencumbered by personal feelings and social positions (which are sometimes troublesome for human nurses).

Accountability: can robots and artificial intelligence take responsibility for their actions?

Accountability is the ability to give a reasonable explanation for one’s actions. Nurses are expected to be accountable for their actions. It would be possible to implement this functionality in a robot or AI. As long as there is no malfunction, the actions of robots and AI are based on algorithms, and it is not impossible to explain the reasons (causes) for their actions in this respect. ²¹

There is the question of whether the explanation of an action output in such a way is comparable to that of a human agent. This issue is one of rationality of explanation. That is, the explanation of the action must be at least understandable to the human being and must indeed be the actual cause of the action. However, a similar problem can exist for human agents. For example, it is easy to imagine a situation in which a human nurse explains a false reason for her act for the sake of self-preservation when the real cause of the act is another. In this case, it is necessary to further examine whether accountability can be said to exist when the rational reason and the true reason do not coincide.

Additionally, a human nurse would have difficulty explaining the reasons for their own actions due to vanity or self-preservation, but a robot may be able to explain the reasons for its actions without such human emotions. In a sense, it is possible that a robot nurse can achieve accountability better than a human nurse.

Furthermore, even if a robot or AI could explain the reasons for its actions, there is the question of whether the robot or AI itself could or should be held responsible for its explained actions. In the field of robot ethics, the issue of the agency of robots is actively debated. ²² There are various positions on whether robots can be taken as agents or responsible subjects, but some believe that it is not necessarily impossible if our society accepts robots and AIs as members of a moral community. ²³ Depending on how we position robot and AI nurses in society in the future, there may be ample room for these entities to become responsible agents of action.

In considering the accountability of a robot with AI, especially if we understand it as narrow AI, the issue may concern the accountability of those who design it, rather than that of the robot itself. In this case, the question is whether a robot nurse is sufficiently accountable if its designer is accountable. Furthermore, when programming a robot to perform nursing actions, the source data must refer to the actions of actual nurses. If such normative nurses’ actions are sufficiently accountable, another question arises regarding whether the robots that refer to them are also accountable. In both cases, if the process is analogous to the process of a parent teaching a child what to do or a veteran nurse educating a new nurse, even if the designer and nurse being referred to are accountable, the robot itself will not be an accountable entity that guides the action thereon. In general, accountability requires the ability to explain the reasons for one’s actions, beyond merely following commands or imitating a senior nurse. In this respect, the question will be whether robots can do beyond following directions or imitation of commands and choose their own actions in the future.

Currently, there is a growing concern about the accountability of AI in general, and especially for medical AI, for which there is a strong demand to provide explanations for AI decisions. ²⁴ Various types of explainable AI are currently being developed to challenge the black box of deep learning, and the methods of explanation are mixed, including those that explain perceptively and mathematically. ²⁵ As development progresses, it is expected that robotic nurses equipped with such explainable AI will have at least some of the accountability that is required of human nurses. However, the development of an explainable AI will not be easy, because it must face philosophical and epistemological issues along with technical ones. The first issue concerns resources. Spending cognitive resources on explaining judgments may distort the accuracy of judgments in the first place. London (2019) argues that opaque decision making is common in medicine to begin with, and that AI should focus more on the accuracy of judgments rather than explainability. ²⁶ The second is the problem of infinite regression of explanations—that is, an explanation provided by an Explainable AI requires further explanation of that explanation—to which Explainable AI is not immune. Third, there is a question regarding the type of explanation. In the context of epistemic justification, causal explanation is considered. However, are causal explanations the kind that patients need in a nursing setting? The ancient Greek philosophical tradition organizes causes into four categories: matter, form, agent, and purpose. These concepts of cause are broader and richer than the modern concept of causation. We are still in the process of exploring what explanations are truly needed by patients in the nursing field. Therefore, at this stage, we cannot say that explainable AI will solve the problem of accountability for robotic nurses. Rather, paradoxically, it will open up epistemological horizons specific to nursing, namely, what explanations are needed by patients in the nursing setting.

Cooperation: can robots and artificial intelligence collaborate with other healthcare professionals?

In medicine today, nurses must collaborate with physicians, other healthcare professionals, and, of course, other nurses who are their colleagues to provide medical care. Therefore, nurses need to collaborate appropriately with others and engage in nursing care to achieve medical and nursing goals; from a “cooperation” perspective, the question is whether robots and AI can collaborate with human nurses and other healthcare professionals.

As mentioned in Section Background, the fact that machines, once called auto-nurses, have already become a part of nursing practice makes it quite possible for nursing tasks to be partially mechanized and for human nurses to work together, and this is already happening in reality. Systems that allow real-time mechanical checks of electrocardiogram, heart rate, respiratory rate, blood pressure, and body temperature using bedside monitors, and even centralized monitoring at the nurse’s center, are already routine in many healthcare settings. However, it will not be possible to build a relationship that can be called “cooperation” simply by working together. If we can build such a relationship only by respecting and acknowledging each other, one of the keys here will be how we perceive robots and AI. That is, if we recognize robots and AIs as members of the community, as mentioned in the previous section, they will be fully qualified to be cooperators with human nurses and other healthcare professionals.

The formation of trust between humans and robots has been the subject of considerable research in recent years. These studies have reported that collaborative outcomes are better achieved when trust between robots and humans is established.^27,28 However, these studies have been limited to experiments performed in closed spaces in behavioral economics and have not been fully tested in actual clinical nursing practice. Therefore, we believe that further research is needed to determine whether the formation of a nurse-robot trust relationship is possible in actual clinical nursing situations and, if so, how it affects nurses’ performance.

Concerning this issue, the appearance of robots and AIs that perform nursing tasks may also be a point of contention. For example, in Japan, a seal-shaped robot has been introduced into nursing homes to provide comfort and healing. ²⁹ The nursing robot in “Bicentennial Man,” introduced in Section Background, was depicted as having an appearance indistinguishable from that of a human being, and in the real world, robots and AI with appearances similar to those of human beings could be used to facilitate cooperation. However, if a robot or AI is implemented with an appearance that resembles humans too much, there is a possibility that it will impede cooperation. This is the so-called “uncanny valley” phenomenon. ³⁰ Of course, the uncanny valley itself is a phenomenon that has been evaluated in various ways, ³¹ but further research is needed to determine what kind of appearance robots and AIs should have to be able to efficiently cooperate with human medical personnel and nurses. In any case, the kind of appearance that should be implemented in robots and AI to enable cooperation with human medical professionals and nurses will be the subject of further research.

Caring: can robots and artificial intelligence become care-providing entities?

“Caring” has been emphasized as a particularly important ethical concept in nursing. However, the concept of caring is ambiguous, and it is not necessarily universally endorsed. ³² Again, the purpose of this paper is not necessarily to formulate the concept of caring in detail but to examine whether it is possible to implement “caring,” including its rich content, in robots and AI. Therefore, for the moment, as Tronto argues, we will divide “caring” into four aspects: (1) caring about, (2) taking care of, (3) caregiving, and (4) care-receiving. ³³ Then, we will consider whether robots and AI are feasible for each of these aspects. However, from the perspective of taking care of and caregiving, the question arises whether robots and AI can be responsible care providers, but this point has already been discussed in Section Accountability, so we will avoid duplication and discuss mainly taking care of and care-receiving.

First, is it possible to implement the ability to care about patients in robots and AIs? In recent years, there has been a great deal of research in the field of robotics and other fields that attempts to read a person’s emotions from his or her facial expressions. ³⁴ Of course, there is a philosophical question as to whether robots can understand human emotions in the same way that humans do, but there is still a possibility that robots will be able to understand such emotions, at least externally, and respond appropriately. Understanding needs is also possible, as discussed in Section Advocacy.

The more difficult question is about care-receiving. In other words, the question is whether patients will accept robots and AI as care providers. As described in the auto-nurse failure case study in Section Background, one problem for patients was that the time spent by human nurses providing care could not be replaced by the time spent by the auto-nurse. Of course, this may simply be a matter of habituation, or it may be a problem that can be remedied. For example, if the patient did not want to receive care with the cold, inorganic hands of a robot, it would be quite easy to cover the robot’s hands with a soft material to keep them warm for human skin. In this regard, empirical research on patients’ attitudes toward robots and AI when they become widespread in medical care will be required. Naturally, these studies will require evaluation of the quality and validity of algorithms for decoding the facial expressions of patients—sometimes including those who have lost the ability to speak—and these should be carefully monitored as future research in this field progresses.

So far, we have discussed whether robots and AI can become “caring” entities. On the other hand, there is a concern that trying to implement “caring” in robots and AIs may, on the contrary, reduce “caring” to a one-dimensional concept. ³⁵ While such criticism needs to be taken seriously, it may be an inevitable but not constructive criticism, since the concept of caring itself is ambiguous and cannot be defined in a way that many people can agree on. Rather, in the context of robot ethics, some argue that by teaching (=programming) ethics to robots, we may gain a better understanding of our own ethics. ³⁶ Similarly, we may gain a better understanding of what “caring” is in the process of trying to implement it in robots and AI. Perhaps, as a result of such a process, we may discover that there are important elements of “caring” that cannot be implemented in robots or AI after all, but such a quest does not necessarily seem to be an exercise in futility.

Discussion: How can artificial intelligence and robots be introduced sustainably in the nursing process?

We have been examining the feasibility of implementing ethical concepts considered significant for nurses in robots and AIs. According to our analysis at this point, we believe that it is possible to implement those properties in robots and AIs. In the future, there may be a possible scenario in which nursing robots or AI that imitate virtuous nurses through deep learning will be introduced into nursing practice. From the moral-ethical viewpoint that one can acquire virtues by imitating virtuous agents, it may not be a dream to gradually create virtuous nursing robots and AIs by having them learn the actions and attitudes of virtuous nurses. ³⁶ Although further philosophical and ethical discussion is needed to determine whether such robots and AIs are truly virtuous, they would be extremely valuable in the healthcare setting, where there is a chronic shortage of human resources.

However, even if it were possible to create such a robot or AI, should we create it and put it to work in the nursing profession? Bryson (2010) ³⁷ argues that robots should be slaves and that we should not make them any closer to humans than necessary. ³⁷ In the nursing context, highly developed robots and AIs such as Galatea, introduced in the introduction, may be unnecessary for several reasons.

First, in terms of cost-effectiveness, the creation of a robot that implements nursing virtues, such as caring, is currently still expected to require a significant amount of capital. Although a more detailed cost–benefit analysis is needed, considering the cost of creating a single nursing robot and its maintenance, it may be far more realistic to train and employ a new nurse. Of course, the advantage of robots is that once a manufacturing method is established, the manufacturing cost can be gradually reduced, and the possibility of improving this point in the future cannot be denied.

Second, as mentioned frequently in the previous sections, there is the question of whether patients, their families, and colleagues, such as physicians and other nurses, will accept nursing care by robots and AI. There may be an issue of habituation, of course, but patients and other healthcare professionals should not be expected to become accustomed to such changes, and it is also possible that as our understanding of caring advances, more serious reasons beyond mere habitual issues could be identified. From the perspective of human nurses, there is also concern that they would lose their profession. With the current concern that many professions will be replaced by robots and AI, although nurses are relatively difficult to replace, there are reasons for nurses to reject such replacements when the possibility becomes apparent in the future. However, from the perspective of an aging society and the increasing sophistication of nursing practice, there remains the possibility that nurses would welcome the entry of new colleagues who would not be fatigued. In such a case, human nurses may have the time and energy to devote to tasks that only humans can perform, or would be preferable for them to perform. Accordingly, the entry of robots and AI into nursing would be a desirable outcome for both patients and nurses.

Third, the introduction of robots and AI into nursing raises the issue of authenticity. This is the fear that the introduction of robots and AI into nursing practice will replace what we originally hoped to achieve in nursing with something else. In other words, the provision of person-to-person, compassionate care to patients, which is considered essential in nursing practice, may be altered. Even if, as a result of various improvements, many patients, physicians, and other nurses accepted nursing care provided by robots and AI, it may only be a substitution for human nurses. There is an unresolved concern that patients are reluctantly accepting care provided by robots and AI, even though they prefer to be cared for by human nurses. For example, Sparrow describes a scenario in which an elderly person with dementia endlessly repeats the same story to a care robot—a tragic example of how care for the elderly can be entrusted to robots. ³⁸ When robots and AI replace care that was traditionally provided by human nurses, the inherent richness of nursing practice provided by humans may be reduced. ³⁹ For example, among the rich content of caring, only the aspects that can be implemented in robots and AI may be emphasized, and as a result, caring, which we consider important, may be unintentionally transformed.

Despite the problems mentioned above, there is certainly something exciting about the development of nursing robots and AI with implemented capabilities such as accountability and caring. First, as mentioned above, if these can be successfully utilized in the extremely busy nursing environment so that human nurses can truly focus on nursing, this is a desirable future. In addition, if the robots and AIs take over the workload and have the ability for advocacy and caring, and promoting cooperation and accountability in the medical setting, this would be the best scenario for patients, physicians, and human nurses alike. In addition, as mentioned in the previous sections, if practical and philosophical understanding of caring can be deepened in attempts to implement caring in robots and AIs that perform nursing tasks, we should at least be proactive in our research of such robots and AIs. Thereafter, our society should further discuss the pros and cons of introducing this into practice.

Summary

In this paper, in anticipation of the future advancement of robots and AI in nursing practice, we have considered whether it would be possible to implement the ethical concepts required of real-world human nurses in robots and AI nurses. Consequently, it was thought that, although some difficulties could be expected in Advocacy, Accountability, Cooperation, and Caring, it is possible. Of course, elements that require emotional communication with patients—such as valuing and mediating, which are components of the advocacy concept—are not easy to implement. To facilitate emotional communication based on the patient’s values, the current robotic nurse will have to undergo drastic development. The emergence of explainable AI would seemingly bring the accountability required of human nurses to robot nurses. However, the concept of explanation faces the problem of infinite regression of explanations and the attribution of responsibility. Although overcoming these difficulties is not easy, a future in which robotic nurses are welcomed as members of a community in care settings and collaborate with human nurses and patients is possible to some extent.

However, even if these could be implemented in the future, further consideration is needed as to whether such robotic or AI-based nursing care should be introduced. In such discussions, it will be necessary to involve not only ethicists and nurses, but also a wide range of society members. Such studies also include further elaboration of those concepts, such as the concept of caring, and we believe that those will certainly contribute to the future development of nursing practice.

This is a pioneering paper that focuses on nursing practice and discusses the ethical pros and cons of the introduction of robots and AI. However, although we examined four important ethical concepts in nursing—advocacy, accountability, cooperation, and caring—these are not the end of the list of ethically relevant concepts in nursing. Moreover, it is beyond the scope of this paper to examine whether these concepts are essential for robots and AI to play a role in nursing. Therefore, further research should be conducted from the viewpoints of both ethics and nursing on the points to be considered when introducing robots and AI into nursing. If this paper can serve as a foundation for such future research, it can be considered to have fulfilled a certain role.