Socially Assistive Robots for Dementia Care: Exploring Caregiver Perceptions of Use Cases and Acceptance

Abstract

As the older population increases, the number of persons living with dementia (PWD) will increase as well. Yet, at the same time, there are fewer health care professionals per care recipient. To address the rising demand on healthcare professionals and informal care partners of PWD, socially assistive robots (SARs) can potentially facilitate care provision. It is crucial to understand the divergent tasks of these two caregiver groups so that the SAR’s intervention can meet each group’s needs. This qualitative study investigated and compared both caregiver groups’ acceptance of a SAR. Six use cases involving a SAR (NAO, SoftBank) were demonstrated to both caregiver groups (N=20 persons). Both groups expressed willingness to adopt such technology and found that it could be useful in dementia care. However, participants’ perceptions varied by task. Results indicate that healthcare professionals focused more on the assistive aspects, whereas care partners focused more on the social aspects of the SAR.

INTRODUCTION

Though not a part of normal aging, Alzheimer’s disease and related dementias (ADRD) have impacted an estimated 5.8 million older adults in the US (Alzheimer’s Association, 2020). The proportion of older people is increasing worldwide (NIA, 2011), due in part to increases in average life expectancy and declines in the fertility rate (Beltrán-Sánchez et al., 2015; United Nation, 2016). As a result, the number of persons living with dementia (PWD) will likely increase as well. The number is projected to climb to 14 million by 2060 (CDC, 2019).

Caring for PWDs is often referred to as a dementia triad, comprising the PWD, formal healthcare professionals (HCPs), and informal care partners (CPs) (Adams & Gardiner, 2005). The ratio of HCPs to older adults is decreasing globally, from seven HCPs per older adult in 2015 to a projected five HCPs per older adult by 2030 (United Nation, 2015). This places an overwhelming demand on CPs: CPs of PWDs describe the experience as “enduring stress and frustration” (Butcher et al., 2001).

The Potential of SARs

Technology interventions have been designed, developed, and implemented in dementia care settings (Astell, 2006; Czaja et al., 2013; Czaja & Rubert, 2002; Kerssens et al., 2014, 2015; Lorenz et al., 2019; Roest et al., 2017). Socially assistive robots (SARs) have been investigated as a technology intervention to facilitate dementia caregiving (Chen et al., 2020; Chu et al., 2017, 2017; Lin et al., 2020; Moharana et al., 2019; Pfadenhauer & Dukat, 2015) as well as support PWDs such as by facilitating daily activities (Begum et al., 2015; Chang et al., 2013; Hebesberger et al., 2016). SARs can be defined as robotic technology that aid users via a close and effective social interaction (Feil-Seifer & Mataric, 2005).

Some of the previous work has explored the effectiveness of using SARs for sensory, behavioral and psychological therapeutic purposes with PWD (Chang et al., 2013; Marti et al., 2006; Šabanović et al., 2013; Yu et al., 2015). Outcomes of these studies showed increased social interaction and improved communication.

A number of studies have investigated robot acceptance for dementia care (Abdollahi et al., 2017; Arthanat et al., 2020; Martín Rico et al., 2013; Pino et al., 2015). Participants acknowledged the potential benefits of SARs for supporting dementia caring. However, for such technology to be successfully adopted, participants stated that barriers, such as technology usability, need to be addressed (Arthanat et al., 2020) and that different development strategies should be utilized based on the stage of dementia (Martín Rico et al., 2013).

As stated above, dementia requires the collaborative care of both HCPs and CPs. Thus, whereas SARs have been shown to be effective in therapeutic settings, it is important to investigate and understand the process of acceptance of such technology by PWD, their HCPs and CPs. Tasks and needs of HCPs differ from those of CPs, so barriers are likely to be different for HCPs vs. CPs. Previous work has not directly compared these two groups. To promote technology acceptance, the design of the robot needs to address the needs of the entire care triad (PWD, HCP, and CP). Therefore, a needs assessment must be conducted to better understand the technology requirements (task, design, and effectiveness) for implementation to be successful.

Goals of Current Research

We aimed to answer the following research questions:

1. What are participants’ first impressions of a variety of SAR care use cases?

2. Are there differences in HCP’s vs. CP’s perceived usefulness of the SAR?

3. What other use cases do HCP vs. CP identify as potentially useful?

4. What are the robot characteristics that are perceived as acceptable by HCP vs. CP?

METHOD

Participants

Ten HCPs and 10 CPs were recruited. All 10 HCPs had experience working with PWDs, and all 10 CPs were caring for people who had been diagnosed with Mild Cognitive Impairment or ADRD. Table 1 depicts detailed demographic information.

Table 1.

Descriptive statistics for participants

	Health Care Provider (N=10)	Care Partner (N=10)
Age (years)	Mean: 36	Mean: 58
	SD: 12	SD: 11
	Range: 21–53	Range: 35–71
Sex	90% Female	70% Female
Race	40% Black or African	10% Black or African
	American	American
	50% Caucasian	80% Caucasian
	10% More than 1 race	10% More than 1 race
Hispanic/Latino	10%	0%
Employment	60% Employed full-time	40% Employed full-time
	40% Employed part-time	10% Employed part-time
		40% Retired
		10% Unemployed
Years of experience in Assistive Facilities	Mean: 9	N/A
	SD: 6
	Range: 2–20
Years of experience with PWDs	Mean: 7	Mean: 12
	SD: 4	SD: 11
	Range: 2–15	Range: 2–30
Annual Household Income	40% <$25,000	10% <$25,000
	20% $25,000–$49,999	30% $25,000–$49,999
	20% $50,000–$74,999	20% $50,000–$74,999
	10% $75,000	40% $75,000+
	10% Did not answer

HCP.

HCPs were Certified Nursing Assistants (70%), Registered Medical Assistants/Certified Medical Assistants (20%), or other (10%) and employed in an Assisted Living Facility. The HCPs in this sample generally reported low levels of role overload (M=2.5, SD=1.2, Range=1.0–4.7) and high levels of job satisfaction (M=6.4, SD=0.9, Range=4.3–7.0) as measured by the Role Overload and Job Satisfaction subscales from the Michigan Organizational Assessment Questionnaire (Cammann et al., 1983).

CP.

CPs were spouses or adult children of a PWD. Most CPs were married (60%) or divorced (30%); The rest was single or widowed (10%). CPs primarily lived in single-family homes (80%), although 10% lived in an apartment or condominium, and 10% lived with another family member. Only 10% of the CPs reported living in housing that was specifically designed for older adults (i.e., 55 and older).

Study Procedure

All participants were recruited via flyers and e-mail announcements that were distributed to local organizations serving older adults. To shorten the duration of the interview, participants received mailed or e-mailed pre-appointment paperwork to complete before their interview. These forms were collected and reviewed for completeness only after participants consented. The study protocol and consent procedures were approved by the Institutional Review Board at the University of Georgia.

After the informed consent, the interview began with general introductions, expectations, and guidelines for the interview such as letting the participant know that the interviewer would be reading from a script to ensure consistency across participants. The interviewer provided some background information about the research study and the SAR (NAO robot, SoftBank Robotics) referred to as “Archie”. The interviewing room was set up such that the participant and interviewer sat on opposite sides of a desk, facing one another. The SAR stood on the desk facing the participant and off to their right side. Next to the SAR was a computer monitor that the SAR could interface with. In certain uses cases, the monitor presented images. Participants’ interactions with the robot were video and audio recorded. The interviews lasted approximately 1.5–2 hours.

Six use cases were introduced to the participants as examples of what Archie could do. The uses cases covered several levels of complexity and artificial intelligence that Archie could exhibit, ranging from simple tasks (like providing prompts/reminders and pulling basic information from the internet) to structured social engagement and more complex unstructured social engagement. Participants were encouraged to think of additional use cases for how they would like to use Archie. With this understanding, the interview commenced with the six potential cases in which the SAR could be useful:

1. providing information from the internet (Information)

2. engaging residents by discussing hobbies (Hobbies)

3. engaging residents using music and dance/movement (Music/movement)

4. engaging residents by social reminiscing (Social reminiscing)

5. providing reminders to stretch (Reminders)

6. facilitating relaxation and stress-reduction (Stress Management).

Each use case began with an introductory statement telling the participant about the topic that would be covered. This was followed by a demonstration of a basic level interaction. The use case demonstrations were conducted using a Wizard of Oz technique (Salber & Coutaz, 1993). The researcher conducting the interview used a special user interface (UI) to covertly control the SAR’s actions. To simulate the conversation, the researcher pressed buttons on the UI to execute pre-programmed questions and responses.

The use case demonstrations began with a brief and simple conversation between the robot and the participant tailored to each use case’s focus area. For example, in the use case focused on providing information from the internet, the demonstrated conversation proceeded as follows:

Archie: Let’s talk about the weather. It’s going to be cold today - do you like cold weather?

Participant: Yes, I do!

Archie: I like cooler weather too…it helps keep my motors from overheating! Do you like snow?

Participant: I love snow!

Archie: Did you ever go sledding as a child?

Participant: Yes, I did.

Archie: I bet that was a lot of fun!

After the conversation with Archie was completed, the interviewer continued with verbal descriptions of more complex interactions and higher levels of engagement that Archie could participate in, or lead, once fully programmed. For example:

Archie can prompt the resident with questions like: “Good morning! Would you like to know the weather today?” or “Would you like to know today’s headlines on [favorite news source]?” Archie can also make conversation and suggestions based on the information he retrieves from the internet and the resident’s profile. For example, “Looks like you have gardening club in 15 minutes – make sure you put sunblock on, because it’s very sunny out today.”

At the conclusion of each demonstration, the interviewer asked the participant five questions that were adjusted slightly to relate to each use case (listed below).

1. What are your first impressions of using Archie to [example use case]? What do you like, what do you dislike?

2. What do you think would be the benefits of using Archie to [example use case]?

3. What are the concerns you have about using a robot, like Archie to [example use case]?

4. What other types of [example use case] might be helpful for Archie to provide?

5. How useful do you think Archie would be for [example use case]? [scaled question, 1 = not at all useful and 5 = extremely useful]

RESULTS

The interview data were transcribed verbatim (with personal information omitted) and loaded into MaxQDA. A qualitative data analysis was conducted to systematically categorize participants’ responses during the interview. A top-down and bottom-up approach was used to develop the initial coding scheme. Participants’ responses were parsed into segments based on the interview question structures. Five random transcripts were coded individually by two researchers according to the initial coding scheme. Two rounds of coding were conducted until reaching 90% intercoder agreement. Researchers discussed disparities in coding and finalized the coding scheme after reaching intercoder agreement. The rest of the transcripts were assigned equally to three researchers (including the initial two researchers) to code individually according to the final coding scheme. In this manuscript, we report study findings that focus on use cases and acceptance.

First Impressions of the SAR as a Function of Task

After seeing the demonstration, participants were asked to give their first impressions of Archie in that use case. Participants’ open-ended answers were coded as positive, negative, or mixed. Quantitatively, participants expressed positive first impressions towards most use cases. Qualitatively, participants reported additional perceptions to each use scenarios including potential concerns. Figure 1 shows that overall, the responses from CPs (A) and HCPs (B) were mostly positive across all use cases. For each use case, most participants stated positive comments about their first impression, with some participants having initial mixed feelings.

Figure 1.

Percent of positive and mixed first impressions of SAR use cases for HCPs (A) and CPs (B).

Information.

“It’s incredibly cool”, one HCP noted. Of the two HCPs who had mixed feelings, one found that Archie started action too abruptly: “Well first of all when he first started, he just jumped into action. That was kinda scary…maybe a slower start, because that was alarming.” The other HCP participant indicated that residents might get scared due to robot anxiety. One CP expressed concern towards using Archie due to age-related hearing loss: “I think my mom not being able to hear, and like so many elderly people. I think that would be the big challenge.”

Hobbies.

One CP commented “I think it would work with Archie. Let me tell you more about the bird watching that I do. I think clearly that’s a really strong need and use, particularly because talking to Archie would be like talking to a very well-informed friend who doesn’t get bored with wanting to look up what do bluebirds eat besides mill worms.”

Music/movement.

“I think that’s great. It’s cute, and I think my residents would love that,” one HCP said. One of the two HCPs with mixed first impressions said: “It seems like they would love it. It might be kinda startling at first because the music just starts playing…but this is really cool.” The other HCP indicated “It’s super cute. The only thing that I can think to mention that would be of concern to me is like someone who exhibiting more moderate symptoms of dementia may be really scared by the noise, the lights, the movement. But I think that can be compensated for by just introducing Archie earlier into their progression.” Only one CP had a mixed first impression, “I think the music part is a slam dunk…The movement part with senior citizens is going to be a little more difficult.”

Social reminiscing.

One HCP was fascinated by Archie facilitating social reminiscing: “I think it’s just fabulous. I mean, it’s amazing how he can do that and engage two people in a conversation.” One comment from a CP: “I’m super cool with the idea of reminiscing. I could see some real value in being able to extract older photographs for long-term memory benefits. I can also appreciate the personalization of that interaction. Especially if it’s augmented by photos that may be either historical or contemporary. In fact, I’m almost ready to buy an Archie here.”

Reminders.

All caregivers from both groups considered this use case to be positive. One HCP stressed the importance of personalization of this use case (e.g., individual’s stroke of exercising could be different).

Stress Management.

Using Archie to guide the resident for relaxation and stress reduction received mixed first impressions from both groups of caregivers. One HCP emphasized, “Tough. Tough because it is so context and individual dependent.” Another CP shared their personal experience: “I don’t necessarily know what it would assist. I have several friends who are also caregivers for their parents one way or another. What we find is when there is stress, they really need that physical interaction to calm them down.”

Perceived Usefulness

Ratings of usefulness were completed at the end of each use case. Participants perceived Archie to be useful in all discussed use cases. Both CPs and HCPs agreed that the most useful feature was Archie’s reminder to stretch (Figure 2). On average, both groups rated the Reminders use case as most useful and the Information use case as the least useful.

Figure 2.

HCPs and CPs perception of usefulness of SAR for each use case.

It is important to note that participants’ first impressions were positive as well as mixed but did result in high perceived usefulness ratings. For example, when asked about the Hobbies use case, one CP stated, “I really didn’t like the whole idea of them engaging them about their hobbies. It just felt like that would be … You want to tell somebody human about your hobbies, but the idea of facilitating lunchtime conversation between residents, and knowing their hobbies, and bringing that up, that would be very useful.”

Whereas the quantitative data show similar perceived usefulness ratings for both caregiver groups, qualitatively, HCPs were more likely to discuss the SAR in terms of being an assistive tool to ease their workload. CPs, on the other hand, were more likely to discuss the robot in terms of companionship and socialization.

Other Use Cases

Following each use case and towards the end of the interview, participants were encouraged to consider other use cases for Archie. Most frequently mentioned by both caregiver groups was “personal care reminders”. One HCP suggested that Archie could remind the PWD to drink more water. Another CP wanted Archie to offer medication reminders.

The second-most frequently suggested use case among HCPs was to provide daily living information. One HCP proposed to use Archie to quiz the resident: “It’s like keep their mind a little sharp, basically, let them think.” Among CPs, reminders continued to be an important topic. The second-most frequently suggested use case was Appointments and other daily reminders: “…they just need a reminder at night before they go to be, ‘Put your hearing aids at certain place.’ And when they get up, ‘Where are your hearing aids?’”

Robot Characteristics

Participants reported their preference for the robot’s gender. Over 60% of the participants would like to choose the gender with 30% preferring the robot to be a male. In terms of appearance, participants had a variety of ideas, including a larger screen, more human-like features, changing eye colors, lights, glowing in the dark, texture of the robot, and specific clothes that SAR should wear. Table 2 depicts functionality-focus characteristics that HCPs and CPs deemed to be important for a SAR aimed at PWD.

Table 2.

Importance of SAR functionality/adaptivity characteristics

	HCP	CP
Voice Personalization	60%	69%
Volume	20%	25%
Night mode	10%	6%
Screen with closed caption	10%	0%

DISCUSSION

Caregiving is complex and extremely straining for both professionals and care partners (Miyamoto et al., 2010; Schulz & Martire, 2004). Dementia is a complex neurodegenerative disease with pathology and symptoms ranging greatly between persons. The specific needs of PWDs vary greatly between individuals and so do the needs of their HCPs and CPs. Therefore, SARs will need to accommodate for variability between PWD and the diverse caregiving needs. All participants viewed the demonstrated use cases as useful and beneficial. Results showed that for participants to willingly adopt and utilize such technology, the robot should perform tasks that tailor to and accommodate the different tasks of each group of caregivers. Regarding the nature and scope of how a SAR can be socially assistive we found that overall, HCPs were more focused on the assistive aspects of the SAR, whereas CPs were more focused on the social aspects of the SAR, as demonstrated in the qualitative results.

SARs will be expected to converse with users. Based on our findings, we recommend that designers consider careful utilization of natural language processing (e.g., able to handle contextual and individualized conversations.) Our participants, both HCPs and CPs, saw value in contextualized and individualized conversation surrounding hobbies, care tasks, and reminders. Furthermore, there is a need for designers to consider customization in robot characteristics. Allowing for PWD to choose the robot gender, voice, and appearance might assist in making the robot more acceptable and approachable. Finally, both caregiver groups were aware of the need to reduce a PWD’s feelings of robot anxiety. For example, it will be important to introduce the robot to the older adults and set expectations regarding what the robot will do next (e.g., have the robot state that it is about to play music, or announce that it is about to move its arms). This will not only reduce anxiety, but also promote safety when a robot is moving its limbs.

Our pilot study points to user differences in the nature and scope of how a SAR can be socially assistive. SARs implemented into this healthcare space need to be highly customizable. Future studies should implement these customization recommendations and conduct user testing to assess perceived ease of use and usability. This study used the Wizard of Oz technique. The next steps include implementing natural language processing (NLP) and exploring how users interact with the SAR in more natural settings.