. 2021 May 24;16:941–971. doi: 10.2147/CIA.S282709

Table 2.

Characteristics and Summary of the Included Research

First Author, Date	Robot/Agent; Name; Description of Intervention	Setting; Country	Sample Mean Age (Range)	Sample Gender (f: Female; m: Male)	Sample Size	Study Design and Duration	Loneliness Measure	Results	Technique	Effect Size (Cohen’s d)
Abdollahi, 2017⁵⁹	Robot; Ryan Companionbot Home companion that converses, facilitates memory work (shows photographs of past), plays cognitive games and shows videos.	Homes in retirement villages; USA	n/s	n/s	6	Observational, mixed methods; longitudinal 3–6 weeks	Recorded conversations with the robot; survey on likeability and acceptability; caregiver feedback	● The robot established a connection with participants. ● Participants liked interacting with Ryan and accepted it as a companion although it cannot replace human companionship. They spent much time with the robot, liked speaking with it and believed it helped them maintain their schedule, improve their mood, and stimulate them mentally. ● Participants were sad to see the robot go.	Direct companion	n/a
Banks, 2008²⁴	Robot; AIBO Animal-assisted therapy with AIBO for 8 weeks (30 min sessions weekly) may help to decrease loneliness and attachment may occur.	Long term care facility; USA	n/s	n/s	38	Experiment (RCT): control (no animal assisted therapy) and real dog comparator 8 weeks	UCLA; modified MLAPS	● Those who received animal-assisted therapy (AAT) were significantly less lonely than those who did not. ● The loneliest individuals improved most with AAT and the most lonely in the control group became more lonely. ● The UCLA showed the control group was statistically different from the AAT (AIBO/dog) groups, but there was no statistically significant difference between the AIBO and Dog groups. ● The MLAPS showed significant attachment in both AAT (AIBO or dog) groups. These were not statistically different.	Direct companion	n/s
Barrett, 2019⁶³	Robot; MARIO Facilitated sessions 3 times weekly (60mins) for 4 weeks. Functions: games, music, memories (their photos), news and calendar (eg social activities, birthdays).	Nursing home; Ireland	83	f: 7; m: 3	10	Experiment; single group, pre-post 4 weeks	Bespoke questionnaire; observations	● Participants’ social connectedness increased through more frequent engagement with others (eg staff or other residents) while using MARIO. This included initiating interactions to tell others about MARIO or show the applications. ● MARIO was a topic of conversation between participants and their family, and had an impact on those not involved in the study (eg other residents danced to MARIO’s music).	Catalyst for social interaction	n/a
Bott, 2019³²	Agent; Care Coach Conversational animal avatar used in hospital to re-orient users (to mitigate delirium risk), elicit toilet needs (prevent falls), and engage in social interaction.	Community hospital; USA	76	f: 52; m: 43	95	Observational; case control pre-post test. Control: daily 15-min visit from a nursing student n/s	3-item UCLA	● Participants with avatars experienced a significant decrease in loneliness, when compared with those in the control group.	Direct companion	UCLA– d=0.36
Broadbent, 2014⁶⁴	Robot; iRobiQ and Cafero Personal service/homecare robots to improve quality of life, reduce depression, and improve medication adherence.	Retirement village; New Zealand	85.23 (72–94)	f: 15; m: 14	29	Experiment (randomised cross-over trial); control: no robot 12 weeks	Interviews	● 17 responses were positive and included the robot being a friend. ● 7 reported positive effects, of which 2 included companionship. ● Participants also found the Skype function useful.	Direct companion; Facilitate communication	n/a
Chen, 2020²⁵	Robot; PARO 24/7 use of harp seal Paro by individuals (at home) may help to reduce depression and loneliness and improve quality of life.	Long-term care facility; Taiwan	81.1 (65–93)	f: 13; m:7	20	Experiment; mixed methods; control: no robot 16 weeks	Interviews; version 3 (20-item) UCLA	● Results from the Paro intervention (from T2 to T4) revealed significant positive changes for loneliness. After the intervention (from T2 to T4), there were significant differences in every time point comparison. ● Paro provided opportunities for social interaction with others, acted as a companion, brought comfort and helped participants live ‘meaningful lives.’ ● Participants thought that Paro could reduce loneliness through direct interaction (eg stroking, petting, conversation) or indirect interactions (eg putting Paro beside them).	Direct companion; Catalyst for social interaction	UCLA- Pre-Paro vs end of 8 weeks- d=2.50
Chi, 2017⁴⁷	Agent; Care Coach Conversational animal/pet avatar used at home that provides companionship, entertainment, reminders, and assistance	Homes, retirement complex; USA	78.3 (68–89)	f: 10, m: 0	10	Observational; mixed-methods; longitudinal 3 months	Interviews	Most stated that the agent provided companionship. Some reported the relationship as superficial because of the agent’s limited conversational ability. Some wanted to bring the agent outside as it could help them to initiate conversations with others. Most enjoyed introducing the agent to friends/families and reported that it enhanced their social interactions with others.	Direct companion; Catalyst for social interaction	n/a
Coşar, 2020⁵⁸	Robot; TIAGo (with embedded home IoT) Homecare robot that delivers cognitive games, healthy tips, facilitates exercise, provides an agenda (calendar), enables calling and presents the weather/news.	Care homes; Greece, UK, Poland	n/s (66–90)	f: 7; m: 4	11	Observational; qualitative; longitudinal 10 weeks	Interviews	● The robots presence (eg waiting at home) helped to alleviate and cope with loneliness and made the environment more friendly.	Direct companion; Catalyst for social interaction	n/a
D’Onofrio, 2019⁴⁸	Robot; Kompa ¨ı 2 (MARIO) Companion and home support robot for people with dementia. Plays music, helps users to reminisce, plays games, shows news and calendar activities and facilitates social engagement.	Residential care, hospital, community; Italy, UK, Ireland	77.08 (55–93)	f: 24; m: 14	38	Observational; quantitative; pretest, posttest n/s	MSPSS; observations	● There was no statistical difference when comparing the total MSPSS before and after interaction. When looking at only the posttest score, those aged 68–76 years perceived that they had major social support and friends to support them. The people aged 77–85 years perceived they had major family support. ● During interactions with MARIO, participants spoke about various things, including MARIO, staff, other residents, themselves, the researchers and other. In Ireland, Italy, and the UK the topic of conversations was most often MARIO.	Direct companion	MSPSS (before vs after)- d=0.61
Fields, 2019⁶⁰	Robot; NAO The robot delivers drama therapy (participatory Shakesparean arts activities) one-to-one to help reduce loneliness.	Residential care; USA	85.80 (77–92)	f: 11; m: 4	15	Observational; quantitative; pretest, posttest; longitudinal 3 sessions (each 10mins)	Revised (3-item) UCLA	● Loneliness significantly decreased across six time periods. ● The decline was slightly greater in those without dementia, compared to those with dementia.	Direct companion	UCLA (pre-post test x 3) Session 1- d= 0.45 Session 2- d= 0.12 Session 3- d= 0.25
Gross, 2019⁴⁹	Robot; SYMPARTNER Home companion robot that provides functional support. Finds and greets users, wakes them, gives reminders, suggests activities, presents the news/weather forecast, helps video call, shows photos/videos, and responds to emergencies.	Residential complex; Germany	74 (62–94)	f: 16; m: 4	20	Observational; qualitative; longitudinal 1 week	Interviews	● Majority developed a personal relationship with the robot companion. ● The robot has potential to be accepted as a meaningful social companion. It provided support in everyday life (eg structure/reminders), provided cognitive stimulation, provided support to keep users on the go and provided emotional stimulation (eg humour). The personal addressing by the robot was helpful.	Direct companion; Reminder of social interaction	n/a
Gross, 2015⁵⁰	Robot; Max Homecare and companion robot that finds users, gives reminders, acts as a fitness coach, helps video call, monitors vital signs, shows photos/videos, presents the weather forecast and responds to emergencies.	Resident complex; Germany	n/s (68–92)	f: 6; m: 3	9	Observational; qualitative; longitudinal 3 days	Interviews	● Participants talked about a co-presence of the robot. Most treated the robot like a social being, although they were aware that it was a machine. ● 8/9 named it and were sad to see it go. ● Participants praised it, felt sorry for failures, ranted to it, cared about it and asked for its opinion. All spoke to it, although it cannot understand them. ● Participants frequently touched/stroked it. This was part of the emotional bonding. ● The robotic companion helped to cope with boredom and feelings of loneliness.	Direct companion	n/a
Hudson, 2020⁵¹	Robot; Joy for All companion pets Robot companion pets used at home by self-identified lonely individuals, with the intention of alleviating loneliness.	Homes; USA	76 (65–90)	f: 10; m: 10	20	Observational; qualitative; longitudinal 4 weeks	Interviews	● Participants with less social connections (and less opportunity) described more engagement with the robot. They were more likely to keep them close and touch them. ● Some felt comfort/companionship when taking the robot with them during the day. ● Participants named their robots and showed them to others. This increased opportunities to connect, especially unknown people. The robot helped forge new connections for those who were shy/felt uncomfortable interacting with new acquaintances. ● A widow explained the pet helped her adjust to her husband’s absence, while those who lived alone used the robot as a proxy for a conversational partner. ● Many thought the robot’s presence positively influenced their feelings of loneliness. ● The robot was as a friend/companion with whom residents formed a strong attachment.	Direct companion; Catalyst for social interaction	n/a
Jøranson, 2016⁵²	Robot; Paro Group sessions to reduce apathy and increase social interaction for residents with dementia.	Nursing homes; Norway	84.65 (62–92)	f: 16; m: 7	23	Observational; quantitative; longitudinal 12 weeks	Video recordings of sessions	● Smile/laughter toward other participants increased; conversations with Paro on the lap showed a decrease during the intervention. ● Conversation with and without Paro on the lap were each registered about 10% of the time, resulting in a total conversation time of 20%. This suggests that participants do not need to have Paro on their lap to contribute to conversations and could indicate that Paro is suitable for increase interactions in groups. ● Smile/laughter toward Paro and toward other participants/activity leaders was registered more than 2% of the time.	Catalyst for social interaction	n/a
Kanamori, 2003⁵³	Robot- AIBO Interaction with pet-type robot AIBO to help maintain or improve quality of life.	Nursing homes; Japan	68.2	f: 5; m: 1	6	Observational; quantitative; pre-post test; longitudinal 7 weeks	AOK Loneliness Scale	● The posttest scores of AOK Loneliness Scale with pet-type robot were significantly lower than the pretest scores. ● The robot acted as a catalyst for social interaction by stimulating sociability.	Catalyst for social interaction	AOK scores (pre to post difference)- d= 1.32 Baseline- M=1.00 ± 1.26 Post- M= 3.33 ± 2.16
Khosla, 2019⁵⁵	Robot; Betty Home-based social robot that facilitates communication and helps to support care services in people living with dementia.	Homes; Australia	n/s (75–85)	n/s	5	Observational; mixed methods; longitudinal 3 months	Video recordings of engagement; survey about attitudes	● Positive emotion: participants expressed positive emotion during 6%–10% of the interaction with their robot. ● Two participants strongly agreed that they wanted Betty to come back to them again, whereas others responded neutrally. ● 4/5 agree/strongly agree that Betty is a friend.	Direct companion	n/a
Khosla, 2013⁵⁴	Robot; Matilda Improves emotional wellbeing (mental and physical health outcomes) and supports healthcare workers in personalizing care.	Residential care facilities; Australia	n/s (71–98)	f: 28; m: 6	34	Observational; mixed methods; longitudinal 2–5 months	Video-recordings of interactions; interviews	● The games and sensory enrichment activities facilitated acceptability of Matilda and encouraged reciprocity/cooperation among older people through their participation in one-to-one interaction activities (eg diet, calendar reminders, and playing quiz). ● The games facilitated interaction with others and bought residents out of social isolation.	Catalyst for social interaction	n/a
Kidd, 2006⁶⁹	Robot; Paro Group sessions in a nursing home to generate social interaction.	Nursing home; USA	n/s	n/s	23	Experiment; Observational; mixed methods; 3 groups (Paro turned on, Paro turned off, No Paro) 4 months	Observations; Social Interaction Questionnaire about Paro	● Sessions where Paro was turned on were quieter than those where dialog was prompted. However, these sessions were more lively than other activities in homes, which were often silent. Evidently, Paro (whether on or off) is an improvement over normal settings. ● When Paro was turned on many interactions consisted of short exchanges and silence. ● Those in the robot on condition felt more encouraged to interact with one another and were more likely to play with the robot. ● The novelty effect appeared to wear off in the robot off condition. ● In the on condition (with a caregiver present), there is continuing interaction with the caregiver and other residents. ● Some residents expressed an attachment to Paro. They named it and conversed with it, using its noises.	Catalyst for social interaction	n/a
Liang, 2017⁶⁵	Robot; Paro Harp seal companion Paro is used at home and in a day care setting to investigate affective, social, behavioural, and physiological effects.	Dementia day cares; New Zealand	n/s (67–98)	64% female	30	Experiment (RCT); mixed methods; control: standard care 12 weeks	Observations	● Participants in the Paro group showed significantly more positive facial expressions and talked to staff and researchers more, compared to those in the control group. ● No significant differences in negative facial expressions and other social responses were observed between participants in either condition. ● Some had conversations with Paro, where they would interpret Paro’s noises and respond. ● Paro encouraged interactions with visitors and other family members. It was a talking point and participants enjoyed showing it off.	Catalyst for social interaction	n/a
Moyle, 2014⁶¹	Robot; Giraff Telepresence robot that enhances engagement between family and a person with dementia living in long-term care.	Long-term dementia care facility; Australia	85 (79–89)	f: 4; m: 1	5	Observational; mixed methods; longitudinal 6–8 weeks	Video recordings; interviews	● Giraff reduced social isolation and increased connection by enabling residents and their families to virtually visit each other. This was important for those who lived far away or may not have seen each other for some time. ● Being able to see the family members’ face was important in maintaining connection and enhanced communication compared to phone calls.	Facilitate communication	n/a
Niemela, 2019⁵⁶	Robot; Double Telepresence robot that facilitates communication at a residential care home.	Residential care facility; Finland	n/s (83–93)	f: 2; m: 1	3	Observational, pre-post test 6–12 weeks	Interviews; observations	● Video connection increased feelings of family members’ presence by opening connection to family not at the facility. This connection was felt stronger than on the phone.	Facilitate communication	n/a
Orejana, 2015²⁶	Robot; iRobi Robot helps older adults in rural communities (who are often isolated) with day to day healthcare and provides companionship.	Homes; New Zealand	85.5 (76–94)	f: 4; m: 1	5	Observational; case studies; longitudinal 3–12 months	Interviews	● There was no influence on external social connections as Skype was not widely used. Residents already communicated with family. ● It felt like they had a companion at home and they did not feel so alone. ● They felt that the robot lighting up was its way of interacting with them. ● All reported that they would miss the robot when it went, with one patient describing the robot as being part of the family.	Direct companion	n/a
Ring, 2015³¹	Agent; Tanya Conversational agent to provide longitudinal social support to isolated older adults.	Homes; USA	65 (56–75)	f: 11; m: 1	12	Experiment; longitudinal; between subjects’ design (proactive vs passive agent) 1 week	UCLA; self-reported state loneliness, comfort with agent, and relationship with agent (5-point Likert); diary entries; semi-structured interviews	There was a non-significant trend towards greater reductions in UCLA loneliness scores in the proactive group, compared to the passive group. Diary measures showed participants reported feeling less lonely with the proactive system than the passive system. Significant correlations found between time spent interacting with agent and change in UCLA scores, state loneliness, degree of comfort with agent, and degree of relationship formed with agent. Participants reported the agent gave them social support and acted as a friend or pet.	Direct companion	UCLA Change scores across groups- d=0.92. Diary measure between groups d = 0.51
Robinson, 2013⁶⁸	Robot; Paro Paro mimics animal assisted therapy in a rest home setting and facilitates social interaction.	Residential care facility; New Zealand	n/s (55–100)	f: 27; m: 13	40	Experiment (RCT); control: activities as normal (eg interacting with resident dog) 12 weeks	UCLA; observations	● There was a significant difference between groups in loneliness change over time, whereby loneliness decreased in the Paro group, but increased in the control group. ● When comparing social interactions with Paro or the resident dog, residents touched and talked to the robot significantly more. The number of residents who stroked Paro was higher when accounting for the number of people present. Residents talked to each other significantly more about Paro. Staff also started conversation more about Paro, but they did not talk to the robot significantly more than they talked to the dog. ● More residents talked to each other in the Paro group than in normal activities and when compared to activities with the dog.	Catalyst for social interaction	ηp2= 0.18 Change scores (pre-post)- differences between 2x groups- d=1.11
Robinson, 2016⁷⁰	Robot; Paro Paro mimics the psychosocial effects of animal assisted therapy in a rest home setting. This can include increased socialising with both Paro and other residents	Residential care facility; New Zealand	84.4	f: 14; m: 6	20	Part of an experiment (RCT); qualitative analysis of Paro only group 12 weeks	Observations; interviews	● Companionship: some residents developed an emotional attachment to the robot. They treated it as a pet or friend. ● Participants who enjoyed Paro were also observed to enjoy sharing it and talking about it. Participants wanted to show the robot to friends/family and other residents. ● Staff thought that Paro would be good for reducing loneliness, comforting distressed residents or entertaining residents. ● The reduction in loneliness reported in the RCT was believed to be a result of increased social communication than of companionship from the robot.	Direct companion; Catalyst for social interaction	n/a
Sidner, 2018⁶⁶	Robot and agent; AlwaysOn System The AlwaysOn System is used in the homes of older adults to decrease loneliness and increase happiness.	Homes; USA	66 (55–91)	n/s	44	Experiment; pretest, posttest; 3 groups: robot, virtual agent, control: nothing 30 days	NSSQ; UCLA; Relationship Closeness Inventory; 5-point likert scale of relationship with agent; interviews; open-ended questions	● The were no significant differences between the groups for the UCLA and social support. ● During the interviews, multiple participants said that the agent provided social support and companionship. The agent’s personality allowed it to be supportive but not judgmental. ● Multiple found that the agent “kept [them] company.”	Direct companion	n/s – Insufficient information (no SD)
Vardoulakis, 2011⁶²	Agent; Tanya This embodied conversational agent was controlled via Wizard-of-Oz and provides social support to isolated older adults.	Homes; USA	62 (56–73)	f: 10; m: 2	12	Observational; Mixed method; pre test 1 week	UCLA at baseline only; agent satisfaction questionnaire; interviews	Participants conversed about many topics including weather, family and future plans. All had something positive to say about their experience. Many reported that the agent provided a sense of companionship, while a few felt no connection. One participant felt worse when talking to the agent as she became aware of her lack of friends.	Direct companion	n/a
Wang, 2017⁵⁷	Robot; Ed Robot for people with Alzheimer’s disease. Provides stepwise prompting to complete at- home activities.	Simulated home; Canada	77.8 (59–88)	f: 6; m: 4	10	Observational 1 day	Interviews	● Participants gave examples of how the robot may improve interactions and relationships by avoiding conflict, decreasing worry/anxiety, and moderating frustrating interactions. ● Perceptions ranged from the robot being a friend to a machine. ● The robot was seen to provide companionship to someone who lives alone, as they can talk to it, without being judged.	Direct companion	n/a
Zuckerman, 2020⁶⁷	Robot; Prototype robot Robot cognitive word game for healthy older adults, with social interaction as secondary function.	Laboratory; Israel	75	f: 21; m: 18	39	Experiment; Mixed methods; 3 conditions: Companion; Game; and No-Function 1 day	Interviews	● Participants associated the robot’s gestures with feelings related to being-seen. This theme was similar across the conditions. The robot’s movements/gestures showed it was paying attention. ● Some felt they had formed a connection. ● 14/39 participants described the robot’s potential to relieve loneliness, emptiness, and sadness.	Direct companion	n/a

Abbreviations: n/s, not stated; n/a, not applicable; UCLA, University of California, Los Angeles Loneliness Scale; MLAPS, modified Lexington Attachment to Pets Scale; MSPSS, Multidimensional Scale of Perceived Social Support; AOK Loneliness Scale, Ando, Osada and Kodama Loneliness Scale; NSSQ, Norbeck Social Support Questionnaire; yellow highlight, positive qualitative evidence for effectiveness; green highlight, positive quantitative evidence for effectiveness; blue highlight, mixed evidence or no evidence for effectiveness.