Table 2.
Meta analysis on end-user experiments of robot-assisted cognitive training.
| References | Participants | Country-term | Study design | Outcomes (after training) |
|---|---|---|---|---|
| Abdollahi et al. (2017) | 6 (1M) seniors with mild dementia and/or depression, aged 63–86 | USA; 4–6 weeks | One-on-one (robot vs. human) pilot study; Each individual had 24/7 access to robot. | Participants established rapport with the robot and greatly valued and enjoyed having the robot in their room. Subjects spent ~130 min per day interacting with the robot. |
| Agrigoroaie et al. (2016) | 1 male with physical disability and cerebellar ataxia, aged 73; 1 female with arthritis aged 83 | UK; One ~1-h session | Interaction with the robot in one partner care facility. | The residents' reactions were positive and they found the robot useful. |
| Alemi et al. (2016) | 11 children with cancer, aged 9.5 ± 1.63 | Iran; 18 days, 8 sessions | WOZ; Randomized into robot-assisted therapy group vs. psychotherapy control group | Children's stress, depression and anger were considerably alleviated during robot treatment. Significant differences were observed between two groups. |
| Ali et al. (2019) | 12 (11M) children with ASD, aged 3.7–10.4 | Pakistan; 6 months, 8 sessions for each intervention | Two different therapies of human-robot interaction, with and without inter-robot communication | Each participant showed improved eye contact duration over the experiments. In imitation module, participants actuated both robots almost equally in recurring experiments. |
| Alnajjar et al. (2020) | 11 boys with ASD, aged 9.03 ± 2.56 | UAE; one 5-min session in 1st week (pilot) and following 7 weeks with 1 session/week | Dynamic interaction scenario; Pilot study and long-term study | In long-term study, all 6 participants portrayed a trend of increasing attention scores. However, the therapist and system assessment trends were similar for most of the patients. |
| Begum et al. (2015) | 3 (3M) persons with ASD, aged 13–19 | USA; 6–10 days; 10–19 sessions; 2–4 min/session | WOZ | Metrics of skill execution and prompt dependency together created a highly informative picture of how well different participants performed. HRI metrics (Gaze, communication, and affect) were unable to measure the efficacy of the robot in achieving the goal of the therapy. |
| Bharatharaj et al. (2017) | 9 children with ASD, aged 9.33 ± 3.39; 9 children's parents; 1 pediatrician; 1 psychologist | India; 5 consecutive days; Three 15-min sessions/day | Pilot study; WOZ; The robot was taught in the presence of children, who are expected to be curious by the robot and compete with the robot. | Results indicated that children with ASD appeared attracted and happy to interact with the parrot-inspired robot. |
| Chu et al. (2017) | 139 (43M) seniors with dementia, aged 65–90 | Australia; ≥5 years; Mostly 1 trial, 4–6 h/trial | Observational study in real life; | Social robots can improve diversion therapy service value to PwD through sensory enrichment, positive social engagement and entertainment. |
| Clabaugh et al. (2019) | 17 children with ASD, aged 3–7 | USA; 41 ± 5.92 days, encouraging 5 sessions/week, 10 games/session | In-home SAR intervention; Single-subject design for subjective measures | Each child participant was engaged with most intervention and showed improved targeted skills and long-term retention of intervention content. The robot system was reported useful and adaptable by families. |
| Conti et al. (2015) | 3 (3M) children with ASD and ID, aged 11–12 | Italy; One 9-min session | WOZ; Robot-mediated imitated game | Suggesting that the robot can be effectively integrated in the ASD therapies currently used. |
| Costescu et al. (2015) | 40 children with TD, aged 5.4 ± 0.4; 41 children with ASD, aged 8.4 ± 2.2 | Romania; Not specified | Counterbalanced; Each participant went through a robot condition and a human condition for reversal learning task. | Children with ASD were more engaged in the task and seemed to enjoy more in the robot condition vs. human condition. Their cognitive flexibility performance was generally similar in the robot and human conditions. |
| David et al. (2018) | 5 (4M) children with ASD, aged 3–5 | Romanian; 20 days, one 10-min session/day | Single-case alternative treatments design; Rapid alternation of 2 treatments; WOZ | A very consistent pattern across all types of sessions: using more cues (i.e., gaze orientation, pointing, and vocal instruction) for prompting JA increased children's performance. |
| David et al. (2020) | 5 (3M) children with ASD, aged 3–5 | Romania; 20 sessions, 1 session/day, 5–15 min/session | Single-case alternative treatments; Robot-enhanced treatment (RET) vs. standard human treatment (SHT); WOZ | Most children reached similar levels of performance on turn-taking skills across SHT and RET, meaning that children benefit to a similar extent from both interventions. The Robot partner seemed to be more interesting to ASD children than human partner. |
| Demetriadis et al. (2016) | 45 (9M) persons with mild CI, age not specified | Greece; ~8 weeks, Once per week, 45–60 min/session | Randomized: intervention group with programming tasks vs. control group | Significantly improved post-test performance in “Test of Everyday Attention” in intervention group vs. control group. |
| D'Amico and Guastella (2019) | 1 boy with impaired spatial abilities and WM, aged 15 | Italy; 1 week, 6 activities, 30–60 min each activity | The boy followed the RE4BES protocol. | Improvement in 4 WM abilities, no improvements in short-term visual memory span, a worsening in word span |
| 1 boy with ID and severe difficulty on focused attention, aged 10 | Italy; 1 month, 2 meetings/week | Single case quasi-experimental design | Significantly reduced problem behavior. | |
| Garcia-Sanjuan et al. (2017) | 40 (8M) seniors with no, mild and severe CI, aged 81.33 ± 8.48 | Spain; 3 tasks, ~10–50 s/task | Usability study; Each user performed tasks individually | It is usable and engaging for users with no or mild CI. It is less usable for persons with severe CI, but triggering positive emotional reactions among them. |
| Huskens et al. (2015) | 3 boys with ASD, aged 5–10; 3 healthy sibling aged 7–11 | Netherlands; 3–5 sessions; 30 min/session; | Concurrent multiple baseline design across 3 child–sibling pairs; 3 pairs were randomly assigned to different baseline lengths of three, four, and five sessions. | No statistically significant changes in ASD children's collaborative behaviors. |
| Ioannou et al. (2015) | 1 boy with high functioning ASD, aged 10 | Cyprus; Four 20-min sessions | Single-case study; The boy played game with the robot and therapist | From session to session, the boy became more independent, initiating interaction with NAO, directing his gaze and expressing affective feelings. |
| Javed et al. (2018) | 3 boys with ASD, aged 7–15; 3 (2M) neurotypical children, aged 4–9 | USA; Activity time not specified | Preliminary study; Test vs. control group; 2-stage activity targeted at sensory processing skills | ASD children initiated more physical contact with the robot on average compared to neurotypical group. Children from both groups waved and smiled at the robot, and displayed imitation by attempting to emulate the robot's dance. |
| Kajopoulos et al. (2015) | 7 (4M) children with ASD, aged 4–5 | Singapore; 3 weeks, six 20-min sessions | 3 phases: pre-test, robot training and post-test | Improved RJA skills after training. RJA skills were transferred from interaction with robot to with human experimenter. |
| Khosla et al. (2017) | 115 seniors with dementia, aged 65–90 | Australian; ≥1 trials; 4–6 h/trial; | Each trial involved 3 stages: introduction of robot, interaction with robot, and robot played games with users. | A statistically significant improvement in emotional, visual, and behavioral engagement of older people with social robots over the years. Their acceptance in the interaction with social robots was verified. |
| Kim et al. (2015) | 48 seniors without CI, aged ≥60 | South Korea; 12 weeks, 5 days/week, 90 min/day | Randomized: traditional CT vs. robot-assisted CT vs. without CT | Attenuation of age related cortical thinning in both CT groups. Less cortical thinning in the anterior cingulate cortices in robot group. |
| Kim et al. (2019) | 48 seniors with mild CI, aged ≥60 | South Korea; 4 weeks, 60 min/day | Single-blind RCT; Robot intervention group vs. control group | Greater improvement in attention in robot intervention group vs. control group. |
| Law et al. (2019a) | 10 (4M) seniors with no or mild CI, aged 75–101; 2 experts in aged care | New Zealand; 1–3 sessions, ~60 min/session | Quantitative and qualitative design to gather users' and observers' feedback | Both users and experts believed the potential of robot-assisted cognitive game to improve cognition in people with MCI. Many functional issues with robot needed to improve. |
| Lins et al. (2019) | 5 (3M) children with mild to moderate CP, aged 4–7 | Brazil; 2 months, 2 sessions/week; | Group sessions; 3-phase game where children manipulated the robot | All children improved their performances on at least one level of difficulty for the exercise, with only two children failing to reach the third and last level of difficulty. |
| Lopez-Samaniego and Garcia-Zapirain (2016) | 7 (3M) seniors with PI and CI, aged 78.0 ± 7.75 | Spain; Once every 3 months, 25 min/session | All subjects participated the same cognitive and physical exercise. | Users were satisfied with the system usability (mean SUS score, 79.29). |
| Manca et al. (2020) | 14 (5M) seniors with mild CI, aged 75.3 ± 4.5 | Italy; 12 sessions over 1 month, 2 days/week | Randomized in terms of technology familiarity; robot- vs. tablet-assisted music game. | Participants in the tablet group provided more correct answers during game than the robot group. The robot was received with more enthusiasm by the older adults. |
| Marino et al. (2019) | 14 (12M) children with ASD, aged 4–8 | Italy; 10 sessions, twice a week, 90 min/session | RCT; Randomized in terms of gender; Robot-assisted intervention vs. control group; Group sessions | Substantial improvements in contextualized emotion recognition, comprehension and emotional perspective-taking through the use of human-assisted social robots. |
| Mois et al. (2020) | 11 (3M) seniors with forgetfulness, aged 74.64 ± 6.02 | USA; 4 weeks, 30-min session per week | WOZ | Engaging with the SAR improved participants' cognitive function across multiple domains |
| Otaki and Otake (2017) | 6 seniors with coimagination, aged 73 | Japan; 1 session; session duration not specified | WOZ | The robot could fulfill its role as a moderator, but the impression of robotic motion was not so good and the robot did not extend the topic by the question. |
| Pino et al. (2020) | 21 (11M) seniors with mild CI, aged 73.45 ± 7.71 | Italy; 8 weeks, weekly 90-min meeting | Group format; Training conditions robot- vs. human- assisted | Robot-assisted memory training increased patients' visual gaze and reinforced therapeutic behavior. |
| Rudovic et al. (2017) | 36 (30M) children with ASD, aged 3–13 | Japan and Serbia; One 25-min session | Exploratory analysis; WOZ; 2 Groups of Japan and Serbia | Statistically significant differences in engagement displayed in the two groups. |
| Salvador et al. (2016) | 11 (9M) children with high-functioning ASD, aged 9.8 ± 2.9 | USA; 5 weeks, 1 session/week | 2 initial baseline sessions; 3 robot assisted intervention sessions. | There is correlation between reinforcer preference and age. |
| Sandygulova et al. (2019) | 14 (12M) children with ASD and ADHD, aged 3–8 | Kazakhstan; ≤ 6–15-min sessions | Iterative interaction design; 2 Phases; Design involving therapists, doctors and parents | Robot-assisted play had positive outcomes for most children. |
| Santatiwongchai et al. (2016) | 6 (5M) children with ASD, aged 3–10 | Thailand; ≤ 6 blocks of imaging matching game | Preliminary experiment; The robot as a medium for children with ASD and their parents in the game | Results varied among the children. Generally, response time and the number of incorrect answers decreased. Children often lost concentration during experiment. |
| Scassellati et al. (2018) | 12 children with ASD, aged 9.02 ± 1.41 | USA; 1 month, 30 min/day | Home-based intervention; Child-robot-caregiver interaction | The system maintained engagement over the 1-month deployment. Children showed improved JA skills with adults when not in the presence of the robot. Caregivers reported less prompting over time and overall increased communication. |
| Shukla et al. (2017) | 30 (12 M) persons with ID, aged 45.24 ± 11.28; 6 caregivers, aged 38.6 ± 9.24 | Spain; 2 days, 1 session/day, 10–20 min/session | Groups with robot-assisted cognitive stimulation vs. only caregiver | A significant reduction in caregiver workload in robot group. Disadvantages of robotic technical limitation. |
| Shukla et al. (2015) | 6 (1M) persons with moderate to severe ID, aged 33–67 | Spain; 3 months, 15–30 min/trial | Case study; 4 categories of participant-robot interactions | Participants showed 33 (out of 54) perfect responses. Irrespective of their mental condition all the participants were able to engage fully with the robot during interaction. All participants showed either a reduced or at-least same level of disability behavior during robot interaction trials comparing to normal situation behaviors. |
| Sung et al. (2015) | 16 seniors with social interactions problems, aged ≥65 | Taiwan; 4 weeks, two 30-min sessions/week | Robot assisted therapy in group session | Significantly improved communication and interaction skills (z = −2.94, P= 0.003) and activity participation (z = −2.66, P= 0.008) in participants after therapy. |
| Taheri et al. (2015a) | 2 twin boys with ASD, aged 7 | Iran; 6 weeks, two 30-min sessions/week | Individual and group sessions; Robot-Patient and Robot-Patient-Brother/Parent | Both participants showed greatly improved joint attention, pointing, and gaze shifting. |
| Taheri et al. (2018) | 2 twin boys with ASD, aged 7 | Iran; 6 weeks, two 30-min sessions/week | Single subject design using WOZ; Robot-Child or Robot-Child-Brother/Parent/Therapist interactions | The JA scores of both participants vs. treatment time showed linear shape of 0.3704 and 0.2589 (p = 0.02). A decrease in autistic and maladaptive behaviors in child with low-functioning ASD. The communication of both participants with each other improved. |
| Taheri et al. (2019) | 4 boys with ASD, aged 6–7 | Iran; 11 weeks (11 sessions), 20–30 min/session | Case study design; WOZ; pre-, post-, follow-up test | As a tool and facilitator, the robot was able to teach musical notes/rhythms to participants with high-functioning ASD. The severity of children's autism as well as the stress of the parents decreased somewhat during sessions. Noticeable improved social/cognitive skills in all participants. |
| Taheri et al. (2016) | 4 boys with ASD, aged 6 | Iran; 11 sessions, 20–30 min/session | Single subject design study; WOZ | All participants showed improvement in playing rhythm. The program affected positively on ASD severity, fine movement and communication skills. |
| Taheri et al. (2020) | 20 (14M) children with ASD, aged 4.95 ± 2.01; 20 (10M) children with TD, aged 5.30 ± 1.95 | Iran; Not specified | Counterbalance condition; Random order of robot-child interaction and human-child interaction; WOZ; | While the TD group showed a significantly better imitation performance than the ASD group, both ASD and TD groups performed better in the human-child mode than the robot-child mode. |
| Tariq et al. (2016) | 3 (3M) children with ASD, aged 3.5–7 | Pakistan; Four 15-min sessions | Exploratory study of robot-mediated play protocol | Increased execution, duration of target behaviors and social development (i.e., communicative competence, turn taking, and eye contact) of children with ASD with the robot-mediated play. |
| Tleubayev et al. (2019) | 3 (2M) children with severe ASD and ADHD, aged 5–8 | Kazakhstan; 21 days, 4–6 sessions on different days, ~15–20 min/session | Exploratory repeated-measures study | Sub 1: interested with the robot, and comprehension of tasks evolved throughout the experiment. Sub 2: Less noticeable dynamics in behavior. Sub 3: Significant improvement in eye contact with the robot and people outside the experiment. |
| Tokunaga et al. (2019) | 21 (12M) healthy seniors, aged ≥65 | Japan; 1 session, Session duration not specified | User study; Individual session. | Robot's appearance was acceptable; Participants had difficulty remembering story (correct rate ≤ 50%) |
| Valent́ı Soler et al. (2015) | 101 (Phase1); 110 (Phase 2) | Spain; 3 months, 2 days/week, 30–40 min/session | Controlled clinical trial of parallel groups; Randomized by living units, stratified by dementia severity: CONTROL vs. PARO vs. NAO (Phase1) and CONTROL vs. PARO vs. and DOG (Phase2). | Phase 1: Improved apathy in patients in robot groups; Declined MMSE (but not sMMSE) scores in Patients in NAO; No significant changes between the robot groups. Phase 2: Increased QUALID scores in patients in PARO. |
| van den Heuvel et al. (2017) | 17 children with severe physical disability, aged 2–8; 7 professionals | Netherlands; 2.5 months, 6 sessions, 2 individual sessions/week or 1 group session/week | Exploratory pilot study; WOZ; Children interacted with the robot in individual or group sessions. | A positive contribution of the robot in achieving therapy and educational goals. Sessions with robot were indicated as playful. The robot can contribute toward eliciting motivation, concentration, taking initiative and improving attention span of children. |
| Wong and Zhong (2016) | 8 (6M) children with ASD, aged 5.3 ± 0.5 | Singapore; 5 weeks, one 45-min session/week | Between conditions and within subjects design. Randomized to control condition and robot training condition | 90% of children achieved some or all of individual pre-set aims. Significantly improved turn-taking skills and JA, and longer duration in eye contact engagement in children in robot condition. |
| Yun et al. (2016) | 8 children with minimum competency level of age-appropriate cognitive skills, aged 3–5 | South Korea; 8 sessions, 30–40 min/session | 8 sessions were executed using iRobiQ and CARO equally; Child-therapist-robot interaction | Highest accuracy of 85.7% by robot in eye contact recognition; Gradually declined total eye contact rate during sessions. Progressively increased correct answer rate (≥72.25%) in reading emotions in participants. |
| Zheng et al. (2016) | 6 boys with ASD, aged 2.8 ± 0.37 | USA; 4 sessions across 32.5 days; Session 5 and 6 the same day | User study; 4 sessions of one-target interventions; 2 sessions to evaluate JA skills after 8 months | This autonomous robotic system was able to elicit improved one-target JA performance in young children with ASD over 8 months. |
| Zheng et al. (2015) | 4 children with ASD, aged 3.83 ± 0.54; 6 children with TD, aged 3.61 ± 0.64 | USA; Four 3-min sessions | User study; 2 human-administered sessions and 2 robot-administered sessions for each participant | The robotic system drew more attention from the ASD children and taught gestures more effectively compared to a human therapist. |
M, male; RCT, Randomized Control Trial; PI, Physical Impairment; CI, Cognitive Impairment; CT, Cognitive Training; WOZ, Wizard-of-OZ robot control; WM, Working memory; ID, Intellectual disability; JA, Joint Attention; RJA, Responding to JA; TD, Typically Developing; CP, Cerebral Palsy.