An interview study on socially assistive robots and professional care relationships

Participant selection and recruitment

The sampling strategy was purposive. In light of the ongoing debate about integrating technology into care settings and the lack of empirical research on how care professionals experience and respond to this transformation, the focus of the study was on the perspectives of those employing SARs in inpatient elderly care and how this might impact relationships between caregivers and recipients. This group included (geriatric) nurses and care assistants who organize social interactions. We aimed for maximum variation of perspectives in terms of work setting, work experiences, time spent working with robots, and type of robots used. Institutions using SARs were identified through web searches. Snowball sampling was also used as a recruitment strategy.

Participants were contacted by email and telephone to arrange an interview. Before the interview, the participant received an email containing study information, an informed consent form, and a short socio-demographic questionnaire. A total of 16 interviews were conducted with 18 participants, as 1 interview involved 3 people from different organizational levels.

To understand how SARs are used today (and in line with a grounded theory approach), we included four people who were not directly involved in care but shape SAR integration. Specifically, they were from the fields of applied science, technical support, management, and caregiver training. In the German context, it is important to note that SARs remain largely piloted and are mostly introduced in the context of funded (research) projects.

Requests to participate were declined for a variety of reasons (by twelve institutions), including lack of personnel resources, high interest but commitments to other studies (and media engagements), because SARs were no longer in use due to practical barriers (e.g., unstable Wi-Fi), the termination of a funded project, or lack of technical support.

Data collection

The interview guide was developed using the process established by Helfferich ³² and discussed in a workshop with other qualitative researchers for validation. The guide focused on the practical experiences, implementation process, perceptions, and reflections of participants (see appendix). Due to the small number of interviewees available, it was not pre-tested but continually adapted. Interviews were conducted by CK (4 interviews) and MB (12 interviews). Interviews were audio-recorded using a voice recorder and lasted between 25 and 90 min. They were then transcribed, and all personal or identifiable information was pseudonymized. Initial notes and thoughts were collected during the interviews, followed by the production of a postscript.

Data analysis

Data analysis followed a constructivist grounded theory approach. ³¹ Inductive coding was the first stage of analysis: The data was read and re-read and then analyzed to identify units of meaning. Conceptual labels (codes) were used to divide the data (into words, phrases, or lines). Codes that related to or represented similar concepts in respondents’ experiences were combined into groups. The categories and the overall analytical framework were developed using codes. To deepen the analysis, the data were condensed in a second round to extract and refine the main categories and to highlight their relationship (focused coding). Additionally, theoretical coding was employed to contextualize the findings within existing concepts in Science and Technology Studies as well as nursing ethics, such as technological mediation, ²⁸ “good care,”^2,10 and critical perspectives on care technology. This enabled an analytical process that was both theoretically informed and empirically grounded. Therefore, it was possible to consider the transformation of caregiving in terms of the participants’ experiences and the broader relational and socio-material complex.

Data were coded by MB, CVK, and CK. In joint coding groups, codes were compared and relationships between categories were mapped (MB and CK, MB and CVK). In addition, categories were presented in interpretation groups (e.g., with nurses) and team meetings. MAXQDA24 was used to code the data.

Setting

The interviews took place in Germany between April 2023 and April 2024. Five interviews were conducted in person at the institutions, seven by video call and four by telephone. The interviewees mostly preferred the latter two formats, as they were less time-consuming. To get a feeling for the care setting, both interviewers carried out at least one interview at the interviewee’s workplace. One interview was conducted with three people as this provided the perspectives of the institution, the local caregiver, and the supporting academic institution.

Sample characteristics

The participating institutions were located across Germany and varied in type, from open day centers to closed institutions that specialized in dementia care.

The sample includes six trained nurses, three allied health professionals, and nine others (e.g., educational scientist, public health scientist, and technical support). There were 14 women and 4 men in the sample, aged between 26 and 64, with care experience ranging from 3 to 40 years, and experience with the SARs ranging from first approaches (a few days ago) to 8 years (Table 1).

OPEN IN VIEWER

Table 1.

Information about the sample.

No.	Position/Role a	Robotic system used	Years of experience in elderly care b	Interview alone/jointly
1	Care assistant (“Betreuungskraft”)	Paro	Intermediate	Alone
2	Care assistant (“Betreuungskraft”)	Paro	Intermediate	Alone
3	Social education worker (organizational development)	Paro	Senior	Alone
4	Public health scientist (project coordination)	Pepper	Junior	Alone
5	Healthcare support worker (“Heilerziehungskraft”)	Pepper, JAIme, and YANNY	Junior	Alone
6	Nurse (nursing home management)	Pepper, JAIme, and YANNY	Senior	Alone
7	Social education worker (social assistant)	Pepper	Junior	Jointly
8	Nurse (organizational development)	Pepper	Senior
9	Researcher (socio robotics)	Pepper	–
10	Geriatric nurse (nursing home management)	Pepper	Senior	Alone
11	Nurse	Navel	Senior	Alone
12	Leadership (organizational development)	Navel	Junior	Alone
13	(Palliative) nurse (nursing home management)	Lio	Senior	Alone
14	Nurse	Lio	Junior	Alone
Additional interviews
15	Organizational development (technical support)	–	–	Alone
16	Researcher (social informatics)	–	–	Alone
17	User experience specialist (robot implementation)	–	–	Alone
18	Social education worker (professional education)	–	–	Alone

The robotic systems used by participants included Lio (2x), Navel (2x), Paro (3x), Pepper (7x), JAIme (2x), and YANNY (2x). As some facilities employed more than one system, the total number of systems is higher than the number of interviews conducted.

Reflexivity

Data collection and analysis were mainly undertaken by CK and MB. CK was the co-project lead and is an interdisciplinary health scientist with a focus on empirical bioethics. MB worked as a researcher on the project and is a sociologist specializing in interpretive research methods. Both researchers had previous experience of conducting qualitative research, particularly semi-structured interviews with health professionals. In addition, CVK, a research assistant trained in life sciences, coded all the interviews. The project and parts of the analysis were discussed in internal team meetings, colloquiums, and in an interdisciplinary interpretation group that included among others nurses and physicians. Their practical insights greatly enriched the interpretations.

Ethical considerations

The study was approved by the Ethics Committee of the University of Potsdam (No. 84/2022). Written informed consent was obtained from each participant prior to inclusion in the study. Data were pseudonymized, and original audio recordings were deleted after transcription to protect participants.

Initial phase: Fears of displacing the human element of care

Initially, caregivers expressed concern about SAR implementation, which they perceived as a potential threat to relational care. This resistance is rooted in a dyadic model of caregiving, in which the caregiver–recipient relationship is seen as fundamental to providing “good care.” From the caregivers’ perspective, both they and care recipients alike expressed concern about the erosion of the human connection.

A recurring theme in the interviews was the fear that SARs could replace human caregivers, undermining interpersonal relationships:

“Sure, the question always comes up in the second sentence [with caregivers]: ‘Are they going to replace us living people?’ And we always answer: ‘We hope not’. (…) Because of course the biggest fear of the seniors is that the human connection will disappear altogether” (Mehl). ¹

Underlying such concern is the conceptualization of the SAR as a potential new partner in a dyadic relationship. Such a development is feared, as it implies substituting the human element with a robotic one.

Furthermore, concerns were raised about the economic motivations behind the implementation. Caregivers feared that SARs would be used as a cost-saving measure, prioritizing efficiency over quality of care:

“I'm worried about robots being misused - residents being left with robots just because it's cheaper. That would be a nightmare. Care is about human warmth, about seeing what someone needs in the moment. A robot can't do that, no matter how many sensors or cameras it has” (Stein).

Caregivers’ relational understandings of care were contradicted by this supposed economic rationale. They believed that maintaining care as relational work with a “human connection” (Mehl) is a normative obligation that political and ethical institutions should address:

“I'm just afraid that more and more people will say, ‘Yeah, we've got a robot here. So we don't need a human’. And that is a political decision (...). That's why I think it's really a matter for the German Ethics Council (...) that care remains human. (...) As I said, I'm a fan of all this, but as an addition and not as a replacement” (Nau).

This skepticism was associated with the previously mentioned ethical commitment to “good care.” Other narratives referred to further reasons, such as fear of job loss or an increased workload, a lack of interest in and inability to cope with modern technology, and concerns about fundamental changes to the nursing profession:

“Care has changed – suddenly you also must watch out for the robot and incorporate it into your work. You're no longer just a care assistant working with people. Suddenly, it all becomes technical. Some were very skeptical, saying: ‘Technology – that's just not my thing’” (Palm).

In addition, caregivers were skeptical about the ability of SARs to authentically emulate emotional and social behaviors. They feared that artificial emotional engagement could create misleading or inauthentic connections with care recipients, ultimately damaging trust. Nevertheless, all interviewees were open to try the new technology, but only a few were starting with a positive view: “From the beginning, I was very excited. ‘Oh, a robot! I really want to try that out’” (Kaya). Additionally, caregivers saw the potential for improving the status of the nursing profession through SAR implementation:

“Robots are already versatile and can be used in many different ways. They are definitely interesting, and even attractive, to young people as a way of learning the profession” (Stein).

In summary, most caregivers were skeptical in the early stages of SAR implementation for various reasons. While they recognized the potentials of SARs, they emphasized the irreplaceability of human empathy, warmth, and judgment in caregiving. Their reservation was rooted in a dyadic model of care, which views the human relationship as fundamental for the notion of “good care.” Implementing SARs in their view runs the risk of replacing human caregivers with robotic ones, which also triggers fears of job loss and fundamental shifts in the nursing profession. However, as the next phase shows, this view begins to shift with practical experience in working with SARs.

Usage phase: Integrating SARs and recalibrating the relationship model

In retrospect, the professionals described how their initial expectations changed as they began to work with SARs. They observed that SARs had limited autonomy, could not fully engage with care recipients, and required constant supervision. This realization led to a recalibration of the relationship model: instead of replacing caregivers, SARs became part of a collaborative triad in which their effectiveness depended on caregivers’ ability to adapt and integrate them into daily practices:

“At first, we thought that if we just put Pepper in the room, something might happen. But we learned that if you just put the robot in the dining room, it will be ignored. It just doesn't work that way. As a person, you really must build a bridge to the elderly” (Pieper).

SARs were primarily used for “communication and memory training, and to promote physical activity” (Pieper). They therefore facilitate social interaction and assist with specific routine tasks. There was some debate as to whether they relieve caregivers. While they could free up time when used for routine tasks, at the same time they often require the presence and supervision of caregivers:

“It's important that robots don't replace staff. Robots should only support us, like reading recipes to seniors who want to cook, while I provide one-on-one care to someone who really needs it” (Dorn).

After understanding the limited functionality of SARs, professionals had to adapt their conceptualization of them as a replacement technology. Instead, SARs were reinterpreted as a tool to support certain care tasks that were then carried out by a triad constellation. While the care relationship was mediated by SARs, effectively turning it into a triad, the roles within the care constellation are marked by power and responsibility differentials. Professional care providers, and to a lesser extent care recipients, need to make an effort to integrate SARs into the care relationship. It is often up to the caretaker to decide when and how to use it and when to end the interaction. For example, the task of calming down an agitated dementia patient can be realized by three entities working together: the care recipient interacting with a robotic seal Paro and the care professional who facilitates this interaction:

“The seal [Paro] has a calming function. (…) The care recipients can stroke and can practically get rid of everything they have inside them for a moment. Then the facial expression relaxes, the body relaxes” (Nau).

However, in the end it is up to the care professional to place Paro in the lap of an agitated person. Other tools or the help of care provider (his/her voice etc.), can also be employed. In this triadic model, caregivers maintained their relationships with care recipients because they facilitated interactions with SARs. Their role ensured that SARs were perceived as a tool to enhance, rather than replace, human connection. SARs thereby broadened the care-toolkit but did not fundamentally change care or professional identities.

While SARs are mostly conceptualized as tools, it is still a relationship that is formed as the following statement illustrates:

“In the beginning we also had to deal with some prejudice: now Pepper comes and replaces the caregivers. No, he doesn't. He is an addition. But now he really is an integral part, and somehow, he belongs. Well, that had to develop. It wasn't like that from the beginning” (Pieper).

Unlike traditional care objects (books or memory games), SARs were perceived as social agents capable of eliciting attachment. As such, their integration into care facilities required intentional effort from staff to foster a sense of familiarity and connection among both caregivers and residents. For example, in one case, the SAR was introduced as a “new trainee” (Erich) and then gradually became part of the team: “But now he’s just part of the team and that’s a good thing” (Erich). Care recipients formed bonds with the SARs, often treating them more like companions than tools:

“The seniors like to shake his hand, just say hello and then he says something. (…) The dementia patients react very positively to this and then talk to him or are simply happy that he looks cute and comes” (Pieper).

While initial concerns focused on job displacement which could be dissolved by experience, these fears often gave way to disappointment with the SARs’ functional limitations. Many caregivers found it frustrating that SARs lacked autonomy and required constant human intervention to perform effectively. For example, they needed to manage technical issues, such as software errors. One participant contrasted the limited capabilities of SARs with inexpensive consumer technologies:

“The need for constant supervision was a recurring theme. (…) With Pepper, there always must be someone with him – a caregiver or a housekeeper – to do something with him. It's a bit of a disappointment, because you'd expect it to be able to do more, you know? When I think about Alexa at home, it costs 30 euros, and you can almost have a conversation with it. With Pepper you can't” (Kaya).

This comparison between low-cost consumer technology and the expensiveness of SARs led to skepticism about whether robotics in their current state could meaningfully reduce workload or improve quality of care.

Our findings show that through their SAR experiences, caregivers gradually adjust their view of how care relationships with SAR will look from a traditional dyadic structure to a more flexible triadic one.

Reflective phase: Establishing normative boundaries for SAR use

Caregivers consistently emphasized the need to establish boundaries for the SAR use. While some boundary-making practices were rooted in ethical reasoning, such as maintaining human-centered values that define their profession, others were more pragmatic, such as acknowledging the unreliability and lack of contextual sensitivity of SARs. Both can be morally relevant, not because of their intent but because of their impact on care practices and values.

As mentioned in the first phase, the fundamental principle articulated by caregivers was that SARs should augment, but never replace, human caregiving. This belief shaped how caregivers integrated and used SARs. When used for social tasks, such as displaying images during biographical work or group gymnastics, they fostered social interaction and indirect conversation, as well as group identity building. However, they were firmly excluded from roles requiring emotional depth or relational care, which seems to drive some tension between the imagination of designers and policy makers and the expectations of caregivers. ² Therefore, one caregiver explained the importance of maintaining a balance between technological integration and human presence:

“And that could be the limit, because sometimes individuality can get lost in the daily grind. But that is our task and our challenge for the future, to ensure that this is the norm in care. (...) When it comes to educating people about data protection and how to handle it, I really don't think there are any major barriers. It's more about how you deal with it, how you implement it, how you integrate it. So that you don't feel like there are just robots running around and they can't really talk to anybody. So, the social aspect must continue” (Dorn).

Caregivers consistently established normative boundaries. They did this in situations that required responsibility, empathy, or direct human involvement. This could include administering medication or assisting with personal hygiene. Such scenarios are not currently feasible with existing robotic systems, highlighting the discrepancy between caregivers’ ethical concerns and the actual capabilities of present-day SARs:

“But I always think that wherever responsibility or human affection is involved, I think that robotics simply cannot and should not, in my opinion, replace humans. So, when it comes to giving pills or food or going to the toilet or things like that, I think a robot has no place there” (Pieper).

This boundary setting underscores the belief that while SARs can play a supportive role, they must not compromise the relational and ethical core of caregiving. However, caregivers recognized that these boundaries are not fixed.

Despite this, end-of-life care emerged as a normative boundary seemingly non-negotiable:

“I would never use [a robot] in a palliative setting. It's a time when you need human closeness. A dying person or someone in palliative care needs so much care and attention that no robot can provide. Palliative care and end-of-life care, I always do myself” (Erich).

Another caregiver expressed similar sentiments but acknowledged potential future challenges that might make a shift in attitudes necessary:

“In the last phase of life, like end-of-life care, there is so much empathy needed. I think if there's ever a situation where we don't have enough staff, maybe a robot would be better than nothing. But for now, these moments require a human presence” (Mehl).

These quotes show that caregivers negotiate the normative boundaries of SAR use. While SARs were welcomed for certain routine and logistical tasks, caregivers maintained firm limits regarding relational and emotional aspects of care, particularly in palliative and end-of-life settings. However, their awareness of future challenges suggests that ethical boundaries are not static but require ongoing dialogue as both the challenges and the role of technology in care continue to evolve.

In the reflective phase, professionals critically evaluated the triad. They sought to ethically allocate professional responsibilities and maintain relational integrity.