=Paper=
{{Paper
|id=Vol-2804/paper8
|storemode=property
|title=Social Robots in Cognitive Interventions. Advances, Problems and Perspectives
|pdfUrl=https://ceur-ws.org/Vol-2804/paper8.pdf
|volume=Vol-2804
|authors=Olimpia Pino,Giuseppe Palestra,Berardina De Carolis,Valeria Carofiglio,Nicola Macchiarulo
|dblpUrl=https://dblp.org/rec/conf/aiia/PinoPCCM20
}}
==Social Robots in Cognitive Interventions. Advances, Problems and Perspectives==
https://ceur-ws.org/Vol-2804/paper8.pdf
Social Robots in Cognitive Interventions. Advances,
Problems and Perspectives
Olimpia Pinoa , Giuseppe Palestrab , Berardina De Carolisc , Valeria Carofiglioc and
Nicola Macchiaruloc,d
a
Department of Medicine and Surgery, University of Parma, Via Volturno, 39, Parma, Italy
b
Hero S.r.l., Martina Franca, Italy
c
Department of Computer Science, University of Bari, Via Orabona, 4, Bari, Italy
d
Exprivia S.p.A., Molfetta, Italy
Abstract
Social Assistive Robots are being used in therapeutic interventions for elderly people affected by cognitive
impairments. The present paper reports our research lines aiming at investigating the role of a social
robot in aiding therapists during cognitive stimulation sessions for elders with Mild Cognitive Impairment
and Mild Dementia. We review our studies whose results show that social robots have been positively
accepted by the seniors in different experiments. Participants were very attentive and involved in the
sessions’ tasks and their experience was mainly positive. Our data suggest that this technology can be a
valid tool to support psychotherapists in cognitive stimulation interventions emphasizing the need of
multidisciplinary approaches combining assessment of behavior and robotics.
Keywords
Social Assistive Robots, Cognitive Stimulation Therapy, Elderly Care
1. Introduction
Robots serve various tasks and purposes in the health and social care sectors and are becoming
one of the most important technological innovations of the 21st century. Robotics could partially
fill in some of the identified gaps in current healthcare and home care/self-care provisions for
their possibilities of engaging, stimulating and managing the users through social interactions
and support them with a large range of functions and services in daily life including intelligent
communication, safety, assistance, therapy, and cognitive stimulation. For these and other
promising applications in support of elderly therapy and home assistance, we expect that
robotics can play relevant roles in the future in the field of smart home technologies and
social/companion robots.
The range of available robotic applications is extremely vast, different and continually grow-
ing, from robots used in minimally invasive robot-assisted surgery [1] and in rehabilitation
[2], to robots designed to function in hospital/care homes and personal robots serving as mo-
tivational coaches or assisting older people [3, 4]. Socially engaging robots and interactive
Workshop on Artificial Intelligence for an Ageing Society (AIxAS 2020)
Envelope-Open olimpia.pino@unipr.it (O. Pino); giuseppepalestra@gmail.com (G. Palestra); berardina.decarolis@uniba.it
(B. D. Carolis); valeria.carofiglio@uniba.it (V. Carofiglio); nicola.macchiarulo@uniba.it (N. Macchiarulo)
© 2020 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR
Workshop
Proceedings
http://ceur-ws.org
ISSN 1613-0073
CEUR Workshop Proceedings (CEUR-WS.org)
technologies aimed at provide people with long-term social and emotional support [5, 6, 7].
Recently, with the purpose to help people with special needs living healthier lives, connecting
with others, the challenge of research is designing empathic robots able to assess and recognize
the users’ affective status. Currently, robots cover solutions at different stages of development,
whether commercially ready and available on the market or still at the various stages of research
experimentation and prototyping.
As far as elderly care is concerned, robots are being used as assistive technologies for people
suffering from Mild Cognitive Impairment (MCI) to help them to remain in their familiar
environment for as long as possible [6, 7, 8]. The increasing use of assistive technologies in this
context is partially due to the fact that aging population is increasingly placing pressure on an
already burdened healthcare system. Neuro-cognitive stimulation programs implemented in
care settings are beneficial only when followed regularly and consistently. Due to financial and
resource constraints, it is often difficult to arrange regular exercise sessions with a therapist or
trainer. One way to potentially address this problem is to use a robot to fill these gaps.
The spectrum of required assistive functionalities of such a robot companion is broad and
reaches from reminding functions (e.g., taking medication or drinking) and cognitive stimulation
exercises, up to detection and evaluation of critical situations like falls. In particular, people
suffering from dementia and cognitive impairments show deficits in memory, thinking and
behavior, and symptoms usually develop slowly and get worse over time [9, 10] with devastating
effects on the psychological well-being and quality of life.
Social humanoid robots seem promising since they can support more engaging interactions
with users. Moreover, recent work exploring the use of robots for aiding cognitive treatments
[11] has shown their potential in this field. They are being effectively used in dementia care, and
several commercially available robots have been employed with satisfactory results in cognitive
stimulation and memory training [9, 12, 13].
The integration of robotics into both formal and informal MCI care opens up new opportuni-
ties for improving the patient’s quality of life relieving the caregivers and healthcare services
burden. Early studies have shown that Social Assistive Robotics (SAR) has the advantage of
enhancing mood, social relationships among patients, and emotional expression of individual
dementia sufferers [14, 15]. As part of a training program aimed to improve the cognitive status
of people with dementia, researchers investigated how patients relate to humanoid robots and
perceive serious games accessed through it [13]. In this study, it was observed that, along with
sessions, elders became more engaged with the Pepper robot showing a positive perception
about the interaction with it. In [16] it has been argued that Pepper is a suitable robot to employ
by patients with dementia, relatives, and caregivers and that its presence brings patients with
dementia in a more positive emotional state. In particular, music sessions stimulate patients
to recall memories and talking about their past. In [9, 17] researchers aimed at the design of
assistive functions for humanoid robots and the NAO has been proposed as support to the
training program. In particular, NAO has been employed in individual and group therapy
sessions to assist the trainer through speech, music, and movement.
Following these findings, we started a research project aiming at investigating the effective-
ness and acceptance of SAR in providing support to interventions addressed to people suffering
from cognitive impairments related to aging and dementia. Our research aimed at providing
evidence that large mutual influences between cognitive neuroscience and robotics enable a
better understanding which leads to an increased acceptance of future robotics in society and
health care services.
To this aim, we performed a few studies with the elderly affected by MCI and Mild Dementia
(MD), with the social robot NAO or with Pepper. In every investigation, we implemented
some cognitive stimulation tasks to be performed in a group with the robot and we measured,
besides the performance in the task itself, the patients’ engagement and attention and emotional
response during the training program. In both cases, results showed that using a social robot as
a cognitive coach is more effective to motivate and engage people, compared to what can be
offered by interactions supported by traditional computers [13, 18].
In the present paper, we report and discuss the main results of these studies by emphasizing
the effectiveness of social robots as a tool for cognitive stimulation therapy.
2. Experiences with Social Robots in Cognitive Rehabilitation
The ultimate goal of our research program is to create a user-friendly interface, for neuro-
cognitive treatment in the health care environment. This paragraph is giving a summarizing
overview of our progress in developing such an assistive robot and still ongoing functionality
testing and pending usability studies with the end-user target groups.
2.1. NAO H25
The humanoid NAO is a fully programmable and valuable robot for Human-Robot Interaction
(HRI) research. Our first experience with SARs for cognitive stimulation therapy in elderly
care employed the NAO as a social platform ideally fitted to monitor and promote cognitive
rehabilitation among the elderly population with neurocognitive disorders such as MCI and
Mild Dementia (MD). NAO was included as an experimental platform in an ecological setting
from a center of the Italian health service. Cognitive stimulation is a therapeutic activity that
plays a crucial role in the recovery of memory functions, or in enabling persons to adapt to
their problems by teaching them the use of different strategies to pay attention, and alternative
ways of encoding, storing and retrieving information. This may offer protection from cognitive
decline and mitigate dementia risk. For group-based interventions, the therapeutic effects of
being with others with similar problems may also help.
The participants were selected from the population of outpatients attending the Center for
Cognitive Disorders and Dementia of Parma (Italy). Twenty-one participants were enrolled
in the experiment (10 females and 11 males) with a mean age of 73.45 years (SD = 7.71). All
the participants were previously evaluated by memory-disorders specialists. Participants were
diagnosed with MCI according to Petersen guidelines and full marks in the two tests measuring
daily living activities (ADL and IADL). The study included data on objective measures of
cognitive functions for baseline and immediately after treatment which was supervised by
psychologists or neuropsychologists. The trial included a comparison between an experimental
condition in which individuals received the training assisted by NAO, and a control condition
in which subjects received the intervention without the robot. We introduced NAO in the
manualized, group-based memory training consisting of 8 sessions lasting eight weeks of the
standard program. Each session lasted about 70 minutes. The contents of the group treatment
programs included education and teaching of compensatory strategies (both internal/mental
strategies and external memory aids), education regarding memory function and deficit. Training
involved written and verbal practice of memory strategies such as visual imagery, association or
categorization and spaced retrieval, time orientation, spatial orientation, visual attention, logical
reasoning. It also included metacognitive strategies to improve awareness and self-regulation,
15 minutes to discuss problems, how to transfer strategies to everyday situations to encourage
practice and generalization of strategy use between sessions. The robot was programmed to
implement five tasks. In different training conditions, seniors participated in sessions with the
support of NAO or only from the psychologist while the interaction was recorded for subsequent
exploration by two cameras.
The five tasks administered by NAO were: i) reading stories; ii) questions about the story; iii)
paired words learning; iv) paired words recall; v) song-singer matching. Equivalent exercises
were also performed by the psychologist in different conditions to obtain comparative data. For
example, the song-singer matching is a task in which the seniors have to recall the song’s title
as a response to the name of the singer. In the experimental condition, NAO sings the song
with the original singers’ voice, and wait for a spoken response from each group member before
to deliver a feedback about it. Conversely, the procedure with the psychologist is performed
matching the title with a written response. In both situations, sessions were held in a room
where the patients sat around a table. In order to evaluate the robot-elderly interaction, metrics
were automatically extracted through the analysis of the video-recorded sessions. The analysis
was made by a customized software that aimed to measure participants’ smiles and visual
attention (Figure 1). The automatically extracted metrics were:
• occurrence of visual attention (defined as the number of times each participant looks at
NAO or the psychologist);
• length of visual attention (defined as the time, expressed in seconds, in which each subject
turns to NAO or the trainer);
• frequency of positive expressiveness (described by the number of times each individual
smile towards NAO or the psychologist);
• length of positive expressiveness (defined by time, expressed in seconds, in which a user
smiles towards NAO or the clinician).
Results suggested a beneficial effect of training on objective reports across a number of domain-
specific, global cognitive, and mood measures immediately after treatment. This indicates larger
effects and clinical benefits following tasks assisted by NAO compared to the pen-and-paper
condition. Furthermore, it appears that tasks assisted by NAO had a positive effect on reports of
people who have had MCI. Significant comparisons emerged between the occurrence of smiles
addressed to NAO and those focused on the psychologist, between the amount of the visual
interaction towards NAO and that allocated to the trainer and, finally, that concerning the
length of the visual attention towards NAO contrasted to that towards the psychologist. So far
none of the studies reported the application of a humanoid robot in the health care setting for
individuals with MCI. Our approach determined that training assisted by a robot allows that the
elderly experienced more attention and less depressive symptoms during a memory-training
protocol.
Figure 1: An example of a frame captured during the memory training program and processed by the
automatic system.
In the previous study, we considered cognitive measures as the primary outcome of interest
and clinical measures as secondary outcomes. We also believe that the quality of the interaction
is perhaps another valuable concern because it can result in improvements in clinical gains. In
this respect, in a second study [19] we addressed the effectiveness of a system in automatically
decode facial expression from video-recorded sessions of a robot-assisted memory training.
Empirical evidence determined that humans interacting with a robot engaged the same social
conventions for eye-gaze and social distance as in human interaction. The study explored
the NAO potential to engage participants in the intervention and its effect on their emotional
state. Engagement in this field can be defined by the act of being involved with an external
stimulus. The main aim of the second investigation was to explore, through a new tool based on
computer vision, the engagement of participants in a cognitive stimulation program assisted by
NAO robot who administered specific tasks from the protocol giving instructions, suggestions
and consequences. The tool recognized six basic emotions detecting multiple faces in each
video frame. During each session, participants were seated around a table where was seated
NAO while the experimenter who operated the robot was in the same therapy room with the
computer visible to the participants.
Data from this study concerned both the automatic recognition capabilities from the system
and the participants’ emotional expressiveness collapsed by tasks and gender. Findings revealed
that the system is able to recognize facial expressions from robot-assisted group therapy sessions
handling partially occluded faces. Statistical analysis revealed that emotional expressiveness
differed for tasks and gender with a different pattern across tasks that entangle different abilities
and personal preferences or interests, and females showing a larger number of emotions with
respect to males suggesting that detecting affective states is particularly relevant in HRI. The
more remarkable results were gender and task differences, with women showing mostly positive
emotions probably, which is fully understandable if we take into account the variety of the
nature of human emotions and the ways of expressing them.
Both studies [12, 19] indicated a reliable memory training program based on the NAO robot
Figure 2: Pepper and the group of elderly people during the cognitive stimulation program.
that adding new evidence base to factors involved in Human-Robot Interaction (HRI) for elderly
people. The use of a humanoid robot as a mediating tool appeared to promote the engagement
of participants in memory training programs. State anxiety level measured following a session
with the NAO exhibited an average value below the mild anxiety threshold [12]. The humanoid
robot provides engaging situations and, in some circumstances, enthusiastic behaviors were
detected in patients as a reaction to some reinforcement phrases after a task, rather than during
the task itself, as long as the reinforcement expressions were not repetitive, but casually chosen
from a list of general reinforcements. Participants thought the interaction with NAO stimulating
and many of them appreciated the reminders and prompts as the fact that the humanoid robot
called them by name and started the training sessions with an orientation to time and place.
An important insight of these studies is that a robot could become the right support for
therapist, in future work, particular attention has to be devoted to obtaining a robot able to
respond autonomously adapting in the context of interaction its behavior to users’ needs and
emotional states evaluating its effectiveness on health and well-being of care recipients.
2.2. Pepper
In a further study, we included the social robot Pepper in a cognitive stimulation program in
cooperation with the Alzheimer Bari ONLUS Association. Differently from the NAO used in the
previous study, Pepper is 1.20 mt. tall semi-humanoid robot that can move around using wheels.
He has got a tablet on its torso that can be used to show useful information (i.e., selection among
multiple options).
The experiment was conducted with a group of eight participants for 3 weeks, with weekly
meetings of about 35 minutes (Figure 2 illustrates the setting of the therapy with Pepper). The
participants were selected according to their MMSE score (Mini-Mental State Examination [20])
and their willingness to take part in the study. The MMSE is a 30-point test used to measure
cognitive functions (or “cognitive impairment”). It allows measuring a person’s level or stage of
dementia.
(a) (b)
Figure 3: Pepper showing (a) a physical exercise – (b) a visual-verbal associative memory task.
In order to form a group that could take advantage of the cognitive stimulation therapy, a
week before the experiment, the therapists carried out neuropsychological assessments with
the MMSE on potential participants in the experiments. We selected patients with an MMSE
score between 13 and 26.2, i.e., patients with the beginning of MCI to a mild stage of dementia,
since patients with these scores can make progress with the therapy. Mixed diagnoses studies
are beneficial in determining the potential for the generalizability of training across diagnostic
groups and reflect typical clinical practice in many centers.
The tasks to be performed during the training program with Pepper were selected by the
staff of specialized therapists of the center from the volumes of “A gym for the mind” [21], and
were adapted to Pepper communicative capabilities. Examples of exercises are:
• motor imitation, in which Pepper shows some physical exercises to the elderly (Figure 3a,
3b);
• word completion, in which the respondent is given the first few letters of a word (such as
VIK) and tries to complete the word as quickly as possible;
• visual-verbal associative memory, as shown in Figure 3b where Pepper shows on the
tablet the image of a famous person and asks for his/her name.
The interaction between the robot and the patients is vocal. More details about the experiment
can be found in [22]. Before running the CST with Pepper, participants and their relatives
received detailed information about the study and subsequently signed their consent to be
video-recorded during the experiments. These informed consents were also signed by their
legal representatives.
Each session was video-recorded. Besides the Pepper’s internal video camera located inside
its mouth (which allowed to better capture the faces of the patients), another video camera was
positioned in the room in order to have a front view of patients’ faces and to be able to analyze
the entire group behavior. We collected the video-recording of the 3 sessions. Recordings were
segmented to have one video for each task. Differently from the previous investigations with
NAO, in this experiment, the same measures (the number of correct answers, eye contact, and
emotions experienced by each participant during each session) were assessed by three expert
observers (two women and one man, of average age 37.67 years old). They showed an almost
perfect agreement index (0.83), calculated through the Fleiss’ kappa [23].
We decided to adopt this approach because, in our future work, we want to compare the
performance of our software for the automatic analysis of elderly behavior with human annota-
tion. The three observers calculated the number of correct answers and the number and total
time each senior looked at Pepper during each exercise of the session. In order to observe all
basic emotions (angry, disgust, fear, happy, sad, surprise, and neutral) on each elderly face, the
assessors were first trained on the Facial Action Coding System (FACS) [24].
From the analyses of the collected video, we can report that:
• Based on the number of correct answers, we can conclude that the patients participated
actively in the experiment.
• The engagement was measured, according to [25], by recording the eye gaze of each
participant towards Pepper. On average, seniors were engaged in the 70% of the time
by Pepper and, in particular, they showed more engagement in the task about motor
imitations, in which they paid attention to Pepper for the 76.53% of the duration of the
exercise. The tasks on visual-verbal associative memory were also especially successful
(74% on average).
• During the experiment, seniors experienced, besides the “neutral” state (on average 79.44%
per session), more positive emotions (on average 19.33%) than negative ones (1.25% for
Session 1, 2.02% for Session 2 and 1.08% for Session 3, respectively). Considering the
videos, it has been noticed that these emotions emerged when subjects disagreed with
the statements made by the other participants and not towards Pepper.
• There were some interesting correlations, calculated using the Pearson coefficient, be-
tween behavioral observations and the neuropsychological evaluations’ scores. In detail,
seniors with a higher MCI tended to experience mostly neutral emotions (r=0.70) and were
less happy (r=-0.80) than the most seriously compromised seniors; positive correlations
emerged between the eye gaze engagement estimation and the MMSE scores (r=0.42).
Despite the experiment design anticipated in the study protocol a comparison with a control
group, because of the COVID-19 emergency, it was not possible to achieve this study arm.
3. Discussion and Outcomes
The reported studies aimed at investigating how social assistive robots can be used to support
therapists in training programs for improving the cognitive function of elderly people suffering
from MCI and MD. Results obtained so far are encouraging since they show that this technology
is engaging and allows seniors to perform cognitive stimulation tasks, but we must recognize
Figure 4: Robot-senior interaction architecture.
some methodological limitations that should be taken into consideration when interpreting the
above-reported findings.
The first limitation concerns the sample size of groups in the experiments, even if the
number of involved subjects is in line with the typical number of group members during
cognitive stimulation sessions, these kinds of investigations should involve a larger sample also
considering a greater number of trials extended over more sessions. Moreover, while in the
experiments with NAO the presence of the control groups or conditions allowed to compare
results in different situations, in the experiences with Pepper this was not possible because of the
COVID emergency. Therefore, we cannot make quantitative but only qualitative comparisons
between the results of the two experimental situations. Pepper appears more effective and
engaging thanks to the tablet placed on the torso that can be used to play tasks including visual
stimulation; it can also move towards the group and also towards specific group members. The
sample groups in the studies with Pepper were heterogeneous concerning their clinical profiles,
an aspect that limited the possibility of identifying a specific target. This issue would be a
newsworthy dimension to scrutinize in future work.
In every investigation, evaluation and feedback from participants showed that the technology
is easy to use and suggested future research challenges the community should address. The
older adults approached the humanoid robot as a human, and a stimulus to go to the care-center
to do the rehabilitation program. For example, participants talked to the robot as an entity
having its own personality, they asked the robot to sing a song with them or to play with them.
Robots applied to real-world applications should perform their activities in a reactive but flexible
manner. Thus, a cognitive robot architecture capable to adapt to human interaction is very
suitable.
Although the current paper concerns specific tasks, other abilities can be surely included for
rehabilitation purposes. Given the great importance of the emotions and their expression, the
investigation of new features and less restrictive forms of choosing robot behaviors may lead
to relevant contributions. The analysis of expressiveness must be performed in real-time to
consent the robot the ability to adapt its behavior to the sensed user reaction.
With this aim and taking into account what we learned from the findings of our studies, we
are developing the architecture illustrated in Figure 4 that includes several modules that require
artificial intelligence.
First of all, there is a clear need to include a module for automatic emotion and engagement
recognition in our system for giving the robot the capability to understand in real-time seniors’
affective reactions, motives on them, and activate the most suitable behavior for the situation.
Then, in order to establish and enhancing social relations, tasks like the mental status check
and the cognitive tasks will be proposed through multimodal dialog. The Dialog Manager
module will understand the meaning of the user sentence by appropriately recognizing its
intent and the involved entities generating the appropriate answer message (i.e., reminds, quiz
exercises, jokes, object recognition, questions, etc.). The user’s dialog move will be analyzed
and the results will be forwarded to other components, if necessary. Thereby, the robot could
personalize the interaction by detecting emotions and engagement. The Cognitive Exercises
module provides cognitive stimulation or training. The user performance in the practice will be
stored in a database for evaluation by careers. The database will keep data for all the modules.
The caregivers need to be actively involved in this process using a web interface that will allow
them to create a program of a therapeutic session and personalize the activities according to
the cognitive evaluation of the members of the group. Caregivers will also able to monitor the
progress of the therapeutic intervention.
Acknowledgments
The authors thank Ilaria Grimaldi, Rosalinda Trevino, the “Alzheimer Bari” ONLUS, the thera-
pists C. Lograno and C. Chiapparino for their support, and all the seniors who participated in
the study.
References
[1] N. Simaan, R. M. Yasin, L. Wang, ”medical technologies and challenges of robot-assisted
minimally invasive intervention and diagnostics”, Annual Review of Control, Robotics,
and Autonomous Systems 1 (2018) 465–490.
[2] R. Colombo, V. Sanguineti, Rehabilitation robotics: Technology and application, Academic
Press, 2018.
[3] J. Fasola, M. J. Matarić, A socially assistive robot exercise coach for the elderly, Journal of
Human-Robot Interaction 2 (2013) 3–32.
[4] J. Broekens, M. Heerink, H. Rosendal, et al., Assistive social robots in elderly care: a review,
Gerontechnology 8 (2009) 94–103.
[5] S. Kriglstein, G. Wallner, Homie: an artificial companion for elderly people, in: CHI’05
extended abstracts on Human factors in computing systems, 2005, pp. 2094–2098.
[6] A. V. Libin, E. V. Libin, Person-robot interactions from the robopsychologists’ point of
view: The robotic psychology and robotherapy approach, Proceedings of the IEEE 92
(2004) 1789–1803.
[7] M. Kanamori, M. Suzuki, M. Tanaka, Maintenance and improvement of quality of life
among elderly patients using a pet-type robot, Nihon Ronen Igakkai zasshi. Japanese
journal of geriatrics 39 (2002) 214.
[8] T. Tamura, S. Yonemitsu, A. Itoh, D. Oikawa, A. Kawakami, Y. Higashi, T. Fujimooto,
K. Nakajima, Is an entertainment robot useful in the care of elderly people with severe
dementia?, The Journals of Gerontology Series A: Biological Sciences and Medical Sciences
59 (2004) M83–M85.
[9] M. Valentí Soler, L. Agüera-Ortiz, J. Olazarán Rodríguez, C. Mendoza Rebolledo,
A. Pérez Muñoz, I. Rodríguez Pérez, E. Osa Ruiz, A. Barrios Sánchez, V. Herrero Cano,
L. Carrasco Chillón, et al., Social robots in advanced dementia, Frontiers in aging neuro-
science 7 (2015) 133.
[10] W. M. van der Flier, P. Scheltens, Epidemiology and risk factors of dementia, Journal of
Neurology, Neurosurgery & Psychiatry 76 (2005) v2–v7.
[11] M. Law, C. Sutherland, H. S. Ahn, B. A. MacDonald, K. Peri, D. L. Johanson, D.-S. Vajsakovic,
N. Kerse, E. Broadbent, Developing assistive robots for people with mild cognitive impair-
ment and mild dementia: a qualitative study with older adults and experts in aged care,
BMJ open 9 (2019) e031937.
[12] O. Pino, G. Palestra, R. Trevino, B. De Carolis, The humanoid robot nao as trainer in a
memory program for elderly people with mild cognitive impairment, International Journal
of Social Robotics 12 (2020) 21–33.
[13] M. Manca, F. Paternò, C. Santoro, E. Zedda, C. Braschi, R. Franco, A. Sale, The impact of seri-
ous games with humanoid robots on mild cognitive impairment older adults, International
Journal of Human-Computer Studies 145 (2020) 102509.
[14] A. A. Vogan, F. Alnajjar, M. Gochoo, S. Khalid, Robots, ai, and cognitive training in an era of
mass age-related cognitive decline: a systematic review, IEEE Access 8 (2020) 18284–18304.
[15] D. Koutentakis, A. Pilozzi, X. Huang, Designing socially assistive robots for alzheimer’s
disease and related dementia patients and their caregivers: Where we are and where we
are headed, in: Healthcare, volume 8, Multidisciplinary Digital Publishing Institute, 2020,
p. 73.
[16] R. De Kok, J. Rothweiler, L. Scholten, M. van Zoest, R. Boumans, M. Neerincx, Combining
social robotics and music as a non-medical treatment for people with dementia, in: 2018
27th IEEE International Symposium on Robot and Human Interactive Communication
(RO-MAN), IEEE, 2018, pp. 465–467.
[17] F. Martín, C. E. Agüero, J. M. Cañas, M. Valenti, P. Martínez-Martín, Robotherapy with
dementia patients, International Journal of Advanced Robotic Systems 10 (2013) 10.
[18] A. Tapus, C. Tapus, M. J. Mataric, The use of socially assistive robots in the design of
intelligent cognitive therapies for people with dementia, in: 2009 IEEE international
conference on rehabilitation robotics, IEEE, 2009, pp. 924–929.
[19] G. Palestra, O. Pino, Detecting emotions during a memory training assisted by a social robot
for individuals with mild cognitive impairment (mci), Multimedia Tools and Applications
79 (2020) 35829–35844.
[20] M. F. Folstein, S. E. Folstein, P. R. McHugh, Mini-mental state: a practical method for
grading the cognitive state of patients for the clinician, Journal of psychiatric research 12
(1975) 189–198.
[21] D. Gollin, A. Ferrari, A. Peruzzi, Una palestra per la mente. Stimolazione cognitiva per
l’invecchiamento cerebrale e le demenze, Edizioni Erickson, 2007.
[22] B. De Carolis, V. Carofiglio, I. Grimandli, N. Macchiarulo, G. Palestra, O. Pino, Using the
pepper robot in cognitive stimulation therapy for people with mild cognitive impairment
and mild dementia, in: Proceedings of ACHI - The Thirteenth International Conference
on Advances in Computer-Human Interactions, 2020, pp. 452–457.
[23] J. L. Fleiss, Measuring nominal scale agreement among many raters., Psychological bulletin
76 (1971) 378.
[24] P. Ekman, Facial action coding system (1977).
[25] Y. I. Nakano, R. Ishii, Estimating user’s engagement from eye-gaze behaviors in human-
agent conversations, in: Proceedings of the 15th international conference on Intelligent
user interfaces, 2010, pp. 139–148.