1 Introduction

Mild Cognitive Impairment (MCI) is a disease between normal cognitive ageing and dementia [35]; people with MCI are capable of conducting self-care activities but they exhibit a slight impairment in the instrumental activities of daily living (IADL), whereas people with dementia show moderate to severe cognitive deficits and behavioural disturbances (i.e., depression and apathy) need assistance for the basic ADL and IADL [6]. A distinctive personality profile characterises people with cognitive deficits with high Neuroticism and low Openness and Extraversion [17] which have been considered risk factors for the increase in severity of cognitive deficits. Similar to patients with dementia, those with MCI can suffer from behavioural disorders such as apathy and depression. They need to be constantly stimulated to improve their condition and consequently their quality of life but with an increased burden for their caregivers.

For these users, care robots are believed to be helpful by assisting them and their caregivers in daily tasks, monitoring their behaviour and health and providing companionship [14]. Some studies evidenced that animal-shaped robots or animaloids can be used with people with dementia as emotional activators and cognitive stimulators, and they are usually accepted by the elderly [25,26,27, 34, 41, 43, 47]. However, a new generation of social humanoid robots that can move and act within human beings’ living space and interact with them, represents an unmissable opportunity to provide home care to the elderly with cognitive impairment.

Socially Assistive Robotics (SAR) [18, 30] describes a class of robots with social interaction and communication capabilities that assist users through social and nonphysical interaction. These robots are adopted in the health application domain. They generate social responses from users, communicate verbally and/or non-verbally, exhibit human and social characteristics, and have a humanoid appearance to facilitate interaction. More importantly, they can move and act within human beings’ living spaces and hence monitor and assist them in daily cognitive and physical activities at home. With the growing number of older people living alone in need of care, using these robots to support them is a tremendous societal challenge.

The use of SARs in health care settings suggests that robotics can be an important and cost-effective technology for the health care system. In this direction, studies report that robots enhance the mood and social relationships of patients with dementia, their perceived quality of life and cognitive ability [5]. Other studies report that robots able to perform assistive tasks [28] for providing life assistance have the potential to improve the daily life of patients with a mild level of dementia.

Acceptability of this new technology is crucial to address to increase its pervasive use.

1.1 Acceptability Evaluation

The adoption of robots in real home environments is still far to be reached since it presents several challenges to be addressed, mainly related to their acceptance in the everyday life of patients. In general, studying the acceptability of a technological device is complex and several aspects need to be analysed in different temporal phases. A proper evaluation of acceptance requires the evaluation of the “a priori” acceptability, acceptance after the use, and finally appropriation which is rarely found in the literature [8].

When evaluating the acceptability of a social humanoid robot used to assist a fragile class of users such as the elderly with MCI, it is not easy to perform all the phases contributing to understanding its acceptability. This is mainly due to users’ changing cognitive and physical conditions and to the difficulty of providing assistive robots in a real setting for a long time. So, researchers use the term “acceptance” to refer to pre- and post-interaction perceptions and judgements [8]. An analysis of judgements made before using a new device will predict the intention to use the device [7]. According to the Theory of Planned Behaviour (TBP) [1] intention to use can be considered a predictor of usage behaviour. Intention to use depends mainly on the perceived usefulness of the device and the perceived ease of use, as assessed by the Technology Acceptance Model (TAM) [9]. However, the perception of usefulness could be different before and after the interaction with the robot.

Moreover, to effectively evaluate possible differences due to expectations before using new technology and after its use, a testing procedure in ecological settings and for a sufficient amount of time is required. Observations of users’ reactions are usually gathered in controlled environments, and only a few studies report user experiences when using the robot in their everyday life in their own private homes.

Other models have been proposed to assess acceptability including many other dimensions, such as the Unified Theory of Acceptance and Use of Technology (UTAUT) [44], providing a holistic understanding of technology acceptance by integrating key constructs predicting behavioural intention and use that are determinants for the actual use of technology. UTAUT offers stronger predictive power than the rest of the models examining technology acceptance. The interactive effect of some constructs with personal and demographic factors demonstrates the complexity of the technology acceptance process, which depends on individuals’ age, gender and experience. Given the continuous advances and variety of information communication technologies, extensions of UTAUT have been used to adapt it to a specific context to improve its predictive power [45, 46].

Nevertheless, it is not assessed yet in what measure robot acceptance in older adults can be explained and predicted by different factors: utility, as reported in [22], enjoyments, as reported in [11], user characteristics in terms of psychosocial factors [2] (emotional loneliness, depressive mood, life satisfaction and social support), personality traits [40] (extraversion, agreeableness, openness, conscientiousness, neuroticism), socio-demographic factors [19] (age, gender, family status, and culture).

The evaluation of acceptance should also consider the emotions and attitudes of older adults that influence their reactions when interacting with robots. With humanoid robots able to recognise faces and emotions, and even to reproduce emotions [10], studies have shown how the interaction with humans is more effective when shaping robot behaviours considering also affective components of the user’s state of mind. The NAO humanoid robot was used to assist a memory-training program for people with MCI in a centre for cognitive disorders [36]. It is able to decode human emotions, simulate emotions, recognise faces and execute physical exercises. Experimental results report that patients experienced more attention and less depressive symptoms during the memory-training protocol when assisted by NAO.

1.2 Motivations

In light of the above-mentioned considerations, we performed a study to evaluate if the acceptance of a social robot (i.e., Sanbot Elf of Qihan Technology Co. Ltd) used for assisting users in their daily activities, differs in elderly users with cognitive impairments after an interaction with the robot. In detail, we believe that an effective assistive robot system with a high degree of user acceptance should be based on the knowledge of the potential users to adapt its interaction modalities to the user’s personality traits [39, 40]. In this context, the robotic behaviour adapted and modulated according to the user’s personality traits and preferences allows for increasing the robot’s acceptance by the users [40].

Since the acceptance level is associated with personality traits and general cognitive functioning [39, 40], we also investigated if the level of acceptance of a social robot was influenced by these variables in the elderly with cognitive deficits who had an interaction with a social robot in their homes.

The robot was instructed to perform both tasks to help patients in their everyday activities and to entertain them. In particular, the robot provided monitoring tasks, such as checking the patient’s health state, and checking that the patient has lunch and dinner when planned. In addition, the robot was instructed to suggest cognitive entertainment activities. We also evaluated whether the acceptance level increased by modulating the robot’s behaviour according to the user’s cognitive and personality profile when performing the planned tasks.

Our hypotheses here are the following: H1. a robot is considered valuable when helping patients in their daily activities and robot acceptability increases when its interaction with the patients is modulated according to their personality profiles (confirmatory hypothesis); H2. robot acceptability increases when providing entertainment activities tailored to the patient’s preferences (confirmatory hypothesis); H3. the attitude towards the use of a robot before and after using it in a real environment could be influenced by personality and cognitive factors of the patients (exploratory hypothesis).

The evaluation was conducted on two sets of patients subject to an individual screening to collect information about their habits, education level, cognitive impairment, and personality profile. The first set interacted with the robot in a controlled environment, i.e., a University laboratory under the supervision of technical staff. In contrast, the second set interacted for a longer period with the robot in their own homes autonomously and without any external control and management.

2 Methods

Fig. 1
figure 1

The Sanbot Robot used for the experimentation

Full size image

2.1 Experimental Setup

A smart environment consisting of a robot platform and a set of technological devices was deployed both in the laboratory and individual homes [13]. The robot used for the experimentation is the Sanbot Elf equipped with infrared sensors, omnidirectional locomotion, two cameras (an RGB-D and a full-HD) and touch sensors, a microphone, speakers, a subwoofer, and a full-HD touchscreen (see Fig. 1). During the experiments, users interact with the robot through the touchscreen. Technical details about the functioning and performance of the robotic platform used in the experimentation at home were analysed in [16].

Fig. 2
figure 2

Overview of the study design

Full size image

2.2 Procedure

The experimental procedure consisted of two phases and an overview of the study design is reported in Fig. 2.

Phase-1 The first phase took place in a controlled environment at the University of Napoli Federico II. A dedicated area was arranged in the laboratory to look like a living room in a private house to increase participants’ comfort. The participants were asked to carry out some activities of daily life, such as working on a pc, watching tv, talking on the phone, making coffee, and ironing. The robot was instructed to ask YES/NO questions about performing predefined entertainment activities, such as: “do you want to watch a sports video?", “do you want to listen to classical music?", “do you want to read a newspaper?". The robot interacted with the user at fixed time frequencies, pre-defined for the duration of the experiment. Each session with the robot lasted two hours.

The participants of Phase-1 performed the experiments under the supervision of team members composed of technical staff, psychologists, and neurologists, and they were accompanied by a family member.

Phase-2 The second phase took place in the home of each participant. Before leaving the robot in the participant’s home, the investigator provided the participant and the caregiver with instructions on the robot’s functioning. The robot was instructed to be active for the entire day (except at the nighttime), and to perform two different types of tasks: to monitor 3 types of user activities, i.e., having lunch, having dinner, reminding medicine (if it was the case), and to suggest different entertainment activities, like the ones considered in the previous phase in the controlled environment [15].

The experiments for each participant lasted from 2 weeks to one month according to availability. The interaction with the robot occurred under two different modalities in a randomised order for each day:

  • Standard modality the robot performed the set of tasks by approaching the subject at a fixed pre-defined frequency of interactions, and at fixed pre-defined times. The pre-defined times were randomly generated without assuring that they were in accordance with the times when the subject was supposed to carry out the activities according to the daily routine, except for the medicine reminder task whose time was always set in accordance with the daily routine because of the specific nature of the task. In addition, regarding the suggestion of entertainment activities, the type of entertainment was randomly generated;

  • Modulated modality the robot performed the set of tasks by approaching the subject at different frequencies set according to the personality profile of the user, i.e., with high-frequency values for subjects with low neuroticism and high openness, and low-frequency values for subjects with high neuroticism and low openness. In addition, the times were set in accordance with the times when the subject was supposed to carry out the activities of daily living according to the daily routine. Moreover, regarding the suggestion of entertainment activities, the type of entertainment was chosen in accordance with the preferences expressed by the subjects when their daily routine was given.

It should be noted that the number of days spent with the robot of each patient was known in advance, so the number of days in standard and modulated modalities were equally distributed, but each modality was randomly assigned on each day to avoid the patient being aware of which one was dispensed.

2.3 Measurements

All participants underwent other screening tests to evaluate their cognitive and personality profiles using the Addenbrooke’s Cognitive Examination Revised (ACE-R) [32, 42], and the NEO Personality Inventory-3 (NEO-PI-3) [21, 38]. The cognitive and personality profiles were evaluated in a single session before the subject interacted with the robot.

The ACE-R is a brief cognitive screening battery assessing five neuropsychological domains: orientation and attention, memory, verbal fluency, language, and visuospatial abilities. It incorporates the widely used Mini-Mental State Evaluation test but provides a more thorough assessment of cognitive functions. The results of each activity are scored to give a total score out of 100 (score range: 0–18 for attention, 0–26 for memory, 0–14 for fluency, 0–26 for language, and 0–16 for visuospatial processing). ACE-R is considered useful in discriminating cognitively normal subjects from patients with mild dementia.

The NEO-PI-3 is a personality questionnaire composed of 240 questions investigating the five main personality traits (i.e., Neuroticism, Extraversion, Openness, Agreeableness, and Conscientiousness) and, for each of them, it considers six sub-scales that define each domain (score range: 47–315 for Neuroticism, 46–230 for Extraversion, Openness, 48–240 for Agreeableness, and 48–240 for Conscientiousness). NEO-PI-3 evaluation software transforms the individual scores into percentiles and into ranges where 0–40 is assumed as low (\(\downarrow \)), 40–60 as average (\(\leftrightarrow \)), and 60–100 as high (\(\uparrow \)).

In our study, we particularly focused on Neuroticism, Openness, and Extraversion, which have been considered risk factors for the increase in severity of cognitive deficits [17] and resulted in affecting the perception of the robot and technology acceptance. Indeed, Neuroticism indicates the tendency to be anxious and to easily experience inadequacy, discouragement and dissatisfaction associated with difficulty in controlling and managing such experiences. Individuals who are high in neuroticism have a tendency to experience negative affects such as fear, sadness, embarrassment, anger and guilt. They are also prone to irrational ideas, are less able to control their impulses, and tend to cope poorly with stress. Openness indicates how the subjects are open to new experiences, have an active imagination, aesthetic sensitivity, attentiveness to inner feelings, preference for variety, intellectual curiosity, and independence of judgement. Open individuals are curious, willing to question authority, and prepared to entertain new ethical, social, and political ideas and unconventional values. Individuals who score low on openness to experience tend to be socially conservative and conventional in behaviour and conservative in outlook. Finally, extraversion indicates how the subject is sociable and how much propensity he has to stay in the company of others in a present and interactive way; they like people and prefer large groups and gatherings. Individuals who are low in extraversion are reserved, independent, and even-paced. Introverts prefer to be alone. Although they do not show the exuberant high spirit of extraverts, introverts are not unhappy or pessimistic.

The acceptance of the robot was evaluated according to the extended version of the Unified Theory of Acceptance and Use of Technology (UTAUT) questionnaire, the Almere model [23], adapted to the context of assistive robots for elderly users. The UTAUT model has been applied to evaluate the acceptance of different types of technology. It represents a sound basis to explore factors determining elderly users’ acceptance of social robots, by adapting questionnaire items to the specific technology it evaluates. The adopted questionnaire consists of 41 items and explores 12 constructs where the system refers to a robotic system: Anxiety (ANX) i.e., user’s feeling of unease when interacting with the system; Attitude (ATT) i.e., positive or negative feelings toward the use of the technology; Facilitating Conditions (FC) i.e., factors in the environment that facilitate the use of the system; Intention to Use (ITU) i.e., the intention to use the system over a period of time; Perceived Adaptability (PAD) i.e., the perceived ability of the system to adapt to the needs of the user; Perceived Enjoyment (PENJ) i.e., user’s feelings of pleasure associated with the use of the system; Perceived Ease of Use (PEOU) i.e., the degree to which the system is perceived to be used without effort; Perceived Sociability (PS) i.e., the perceived ability of the system to perform social behaviour; Perceived Usefulness (PU) i.e., the degree to which a user believes that the system would be assistive; Social Influence (SI) i.e., the users perception that their social network would want or not want them to use the system; Social Presence (SP) i.e., the experience of sensing a social entity when interacting with the system; and Trust (TR) i.e., the user’s belief that the system performs with integrity and reliability. The 41 items are scored with a Likert scale ranging from 1 to 5. The questionnaire has been translated into Italian and a version adapted to be also used before the interaction with the robot was released. The translation was examined at a consensus meeting, back-translated, and approved at a second consensus meeting. A comprehension test was carried out on a subgroup of 15 individuals with the aim of checking that there are no problems in understanding the items due to errors in form, content and grammar. Participants were asked to fill in the questionnaire before (UTAUT-1) and after the interaction (UTAUT-2) with the robot to evaluate if variations in the acceptability level of the robot could be revealed due to the lack of previous experience with robotic technology.

The participants of Phase-2 also underwent an emotional intelligence test, the Empathy Quotient (EQ) [3], and an Eyes Test [4] to assess the Theory of Mind (ToM). ToM describes the ability to attribute mental states (i.e., beliefs, goals, intentions, emotions) to oneself and others, and to understand that others have beliefs, desires, and intentions different from one’s own [37]. ToM is a fundamental aspect of social cognition and recent studies on patients with dementia showed that deficits in core social cognitive abilities, such as ToM, empathy, social perception, and social behaviour, are crucial as well as cognitive deficits [24]. Therefore, it is important to measure this dimension because the social cognitive deficit disrupts the ability to build and maintain supportive social relationships, thereby eliminating the benefits that social interactions have for people living with neurocognitive impairments. One recent meta-analysis shows that people with AD are significantly impaired in both their capacity for empathy and Theory of Mind [12]. The Emotional Intelligence test (Empathy Quotient, EQ) is a self-report questionnaire developed to measure empathy’s cognitive, affective, and behavioural aspects. The EQ comprises 40 questions; responses are given on a four-point Likert scale (score range 0 to 80). The “Reading the Mind in the Eye” test (Eyes test) evaluates the ability to recognise the emotional state of others. The test includes 36 photographs of male and female eyes depicting emotional states. For each picture, participants are asked to choose the emotional state that best describes the eyes, choosing between one of four possible emotions (score range: 0–36).

In addition, they were asked to provide a description of their daily routine, collected through additional interviews, reporting the activities carried out with the corresponding time ranges with the help of family members or caregivers, and to fill out a questionnaire with questions (rated in the range from 1 to 5) about their familiarity with the technology in everyday life (see the Appendix A for the questionnaire).

Finally, only participants of Phase-2 were asked to complete, every day in the evening, a visual analogue mood scale assessing the daily pleasantness (with pleasantness scale in the range [1,3]) experienced by the subject after interacting with the robot (see Fig. 3). The patients were provided with a number of sheets, one for each day, with the questionnaire to be filled. Such a questionnaire was designed as simpler as possible (with simple and few choices) to let the patients understand and fill it on their own. The aim is to evaluate if the level of pleasantness increased when the robot was set in a “modulated modality” according to the user’s cognitive and personality profile rather than in standard modality after an interaction with the robot at home.

Fig. 3
figure 3

Robot pleasantness questionnaire

Full size image

2.4 Patients Enrolment

The neurologists enrolled the patients at the University Hospital Federico II. They were selected from 50 patients that were contacted among the ones registered at the Cognitive Disorders and Dementia Center of the Neurology Unit of the University Hospital Federico II with the following demographic characteristics: 29 males, 21 females, average age \(73\pm 8\) in the range 55–99, and years of education \(10\pm 3\) in the range 5–17 (primary school - university).

Exclusion criteria to be eligible for the present study were: no mobility impairments; no history of psychiatric disorders, like schizophrenia, bipolar disorders and neurological diseases; no history of relevant head injury or cerebrovascular diseases and major medical diseases (e.g., neoplasms, clinically relevant renal or hepatic insufficiency); no secondary dementia.

Their cognitive state was evaluated according to the Clinical Dementia Rating Scale (CDR) [29, 33] and the Mini-Mental State Examination (MMSE) [20, 31] with a diagnosis of Subjective Memory Disorder (SCM) corresponding to a CDR=0, a Mild Cognitive Impairment (MCI) according to the criteria of the Diagnostic and Statistic Manual V (DSM-V) corresponding to a CDR=0.5, and a Mild or Moderate Dementia (AD) corresponding respectively to a CDR=1 and CDR=2. The clinical characteristics of the patients were the following: illness duration in years \(3.9\pm 7\) (in the range 1–8); MMSE \(25.8 \pm 3\) (in the range 16–30); 8 patients with CDR=0, 16 patients with CDR=0.5, 20 patients with CDR=1, 6 patients with CDR=2.

Of the patients selected for the experiments, 20 refused the trial because they did not want to experiment with another support different from their family members, 10 had pets in the house so they could not host the robot at home, 5 lived in a house not suitable to host the robot for the presence of stairs, and 1 was illiterate to be able to interact with the robot through textual messages. These requirements for the patients’ house were taken into account to limit possible damages to the robot. Patients’ safety was assured by the implementation of appropriate navigation and avoiding strategies for the robot that stopped in case of closer distances and by excluding patients with motion impairments.

2.4.1 Participants: Phase-1

Out of the 14 enrolled participants, a small subgroup was identified for Phase-1. The 4 patients enrolled were selected because they could be accompanied in the laboratory by a family member that was required to be present during the experimentation. The demographic and cognitive characteristics of the participants of Phase-1 are summarised in Table 1.

Table 1 Demographic and cognitive characteristics of 4 participants at Phase-1
Full size table

2.4.2 Participants: Phase-2

The remaining 10 patients were enrolled for Phase-2. However, 3 had to withdraw from the experimentation because 1 had a dramatic worsening of the cognitive state, 1 could not host the robot for changing family conditions, and 1 had too many carpets in the house to allow for robot movements.

The demographic and cognitive characteristics of the participants of Phase-2 are reported aggregated in Table 2. In Table 9, the details for each patient are reported.

Table 2 Demographic and cognitive characteristics of 7 participants at Phase-2
Full size table

2.5 Ethical Approval

All procedures performed in the studies involving human participants were in accordance with the ethical standards of the institutional and national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The experiments were approved by the ethical committee of the University of Naples Federico II.

2.6 Informed Consent

Informed consent was taken from all the participants. In the latter case, the consent was signed in the presence of a next of kin and the physician responsible for the experimentation. The agreement to host the robot for the set period was signed for the experiments conducted at home.

2.7 Statistical Analysis

The demographic and cognitive domains were evaluated for the patients of Phase-1 and Phase-2 as mean and standard deviation. The scores of personality traits are classified for each patient according to 3 levels: low (score under 40), average (score between 40 and 60), and high (score above 60).

The Cronbach’s Alpha for the items of each construct of the UTAUT was calculated to establish its reliability considering acceptable an \(\alpha >0.7\). When a construct is composed of more than two items and it presents an alpha value less than 0.7, if by removing one or more items the alpha value is greater than 0.7, the corresponding items are removed.

Comparison between UTAUT-1 and UTUAT-2 was evaluated by Wilcoxon-Mann–Whitney for matched measurements within participants of Phase-1 and Phase-2. A correlation analysis with Spearman’s non-parametric test was performed between scores on cognitive variables, personality traits, EQ, Eye Test, and one of the subscales of UTAUT-1 and UTAUT-2 of Phase-1 and Phase-2.

Correlations among UTAUT constructs and demographic, cognitive, and personality factors were carried out to provide an indication of possible relationships since correlation scores can be established with any number of participants.

Linear regression analysis with multiple independent and dependent variables was not carried out because the sample number was less than 10 participants.

The comparison between the level of the pleasantness of each participant that was recorded when the participants interacted with the robot in “standard modality” (pleasantness scale-standard) and the one recorded when the robot was in “modulated modality” (pleasantness scale-modulated) was performed by the Mann–Whitney’s U test.

3 Phase-1: In-Lab Experiments

The Phase-1 of the experimentation was carried out with a twofold objective: to test the robotic platform in a controlled environment and to tune the system for its deployment in the home environments; to collect information on the interaction of patients with the robot for a short time under the supervision of the team of technicians, psychologists, and neurologists.

3.1 Results

The personality profile, UTAUT-1 and UTAUT-2 scores for the 4 participants of Phase-1 are shown in Table 3.

The final \(\alpha \) values of the UTAUT constructs are reported in Table 4. For PEOU, question 4 was removed, for PU question 2 was removed, and for SP question 2 was removed.

Table 3 Personality traits profile, UTAUT-1 and UTAUT-2 scores of the 4 participants at Phase-1
Full size table
Table 4 Cronbach’s alpha for UTAUT values of the 4 patients at the laboratory
Full size table

We did not find any difference between the score on any sub-scale of UTAUT-1 and the score on any subscale of UTAUT-2, as reported in Table 5.

A general positive perception of the robotic application (it is assumed when the construct score is greater than 3) was identified for almost all the constructs before and after the interaction with the robot. The only construct resulting in a negative perception (it is assumed a negative perception when the average score is lower than 3) is the SI both before and after the interaction. Indeed, a low value for SI is expected in the case of such assistive technology since its use could be perceived as a recognition of a health condition requiring special support. This perception can possibly generate uncomfortable feelings and shame toward their networks and peers. The highest score is obtained for the PENJ construct. Hence, these results were encouraging for the following phase of the experimentation.

Table 5 Comparison between score on “UTAUT-1” and score on “UTAUT-2” in 4 patients interacting with the robot at laboratory
Full size table
Table 6 Personality traits profile, UTAUT-1 and UTAUT-2 scores of the 7 participants at Phase-2
Full size table

Regarding personality traits, we found that Neuroticism (\(\rho \)=\(-\)0.986, p=0.014), Extraversion (\(\rho \)=0.998, p=0.002), and Conscientiousness (\(\rho \)=0.988, p=0.012) correlate with ANX of UTAUT-1. Note that for ANX higher values mean low anxiety and vice-versa. Hence, higher Neuroticism values negatively correlate with ANX, while for Extraversion and Conscientiousness, there is a positive strong correlation. No correlation with personality traits was found for UTAUT-2.

Neuroticism negatively correlates with PENJ (\(\rho \)=\(-\)0.988, p=0.002) and PEOU (\(\rho \)=\(-\)0.956, p=0.044). Extraversion correlates positively with PENJ (\(\rho \)=0.972, p=0.028) and PEOU (\(\rho \)=0.993, p=0.007). Conscientiousness correlates positively with PENJ (\(\rho \)=0.998, p=0.002) and PEOU (\(\rho \)=0.960, p=0.040).

Finally, we did not find any significant statistical correlation between the cognitive data (MMSE and ACE-R) and the sub-scale of the UTAUT-1 and UTAUT-2.

4 Phase 2: At-Home Experiments

The participants of Phase-2 performed the experiments at home without any supervision of external staff, with complete autonomy, and without having any previous experience with the robot in a controlled environment.

4.1 Results

The personality profile and UTAUT-1 and UTAUT-2 scores for the 7 participants of Phase-2 are shown in Table 6. The final \(\alpha \) values of the UTAUT constructs are reported in Table 7. FC was removed as composed only of two questions with a low resulting value for \(\alpha \). For PAD the first question was removed and for PEOU, questions 2, 4, and 5 were removed.

Table 7 Cronbach’s alpha for UTAUT values of the 7 patients at home
Full size table

The Wilcoxon-Mann–Whitney for paired measurements did not show a significant difference between the UTAUT-1 and the UTAUT-2 scores (see Table 8).

Table 8 Comparison between score on “UTAUT-1” and score on “UTAUT-2” in 7 patients interacting with the robot at home
Full size table

A general positive perception of the robotic application was also identified in the at-home experiments for almost all the constructs before and after interacting with the robot. Unlike the previous case, the construct resulting in a negative perception before and after the interaction is the Social Perception (SP). Moreover, ITU and PEOU received a lower score in UTAUT-2 but the difference with UTAUT-1 is insignificant. The highest score, again, is obtained for the PENJ construct.

4.1.1 User Profile and UTAUT

A correlation analysis between personality traits scores and UTAUT-1 and UTAUT-2 showed: i. a significant correlation between a high level of Neuroticism and a high score in the sub-scales of Attitude (\(\rho \)=0.775, p=0.041), Perceived Enjoyment (\(\rho \)=0.804, p=0.029), Perceived Usefulness (\(\rho \)=0.789, p=0.035) of the UTAUT-1, while the dimensions did not correlate with any sub-scale of the UTUAT-2; ii. a significant correlation between a high level of Openness and a low score at the sub-scale of Attitude (\(\rho \)=\(-\)0.821, p=0.023), Perceived Enjoyment (\(\rho \)=\(-\)0.927, p=0.003), Perceived Usefulness (\(\rho \)=\(-\)0.837, p=0.019), and Perceived Adaptability (\(\rho \)=\(-\)0.756, p=0.049) of the UTAUT-1, while the same dimension did not correlate with any sub-scale of the UTUAT-2; iii. no significant correlation between Agreeableness, Extraversion or Conscientiousness and the sub-scale of the UTAUT-1 and UTAUT-2.

Education correlates with Attitude (\(\rho \)=\(-\)0.879, p=0.009), Perceived Enjoyment (\(\rho \)=\(-\)0.816, p=0.025), Perceived Sociability (\(\rho \)=\(-\)0.850, p=0.015), Perceived Usefulness (\(\rho \)=\(-\)0.791, p=0.034), and Trust (\(\rho \)=\(-\)0.810, p=0.027) of UTAUT-1. Moreover, Education correlates with Social Influence (\(\rho \)=\(-\)0.757, p=0.049) of UTAUT-2. Familiarity with the technology does not correlate with UTAUT-1. In the case of UTAUT-2, familiarity with the technology correlates with PS (\(\rho \)=\(-\)0.788, p=0.035), PU (\(\rho \)=\(-\)0.781, p=0.038), and the total score (\(\rho \)=\(-\)0.776, p=0.040).

A correlation analysis between Eye test scores and UTAUT-1 and UTAUT-2 did not show significant results. However, correlation analysis between EQ scores and UTAUT-1 showed a significant correlation between a high level of EQ and a low score on the Intention of Use sub-scale (\(\rho \)=\(-\)0.855, p=0.014);. At the same time, the same dimension did not correlate with any sub-scale of the UTUAT-2.

Finally, the correlational analyses between the cognitive data (MMSE and ACE-R) and the sub-scale of the UTAUT-1 and UTAUT-2 showed: i. a significant correlation between the score on Visuospatial and the sub-scale of Attitude (\(\rho \)= \(-\)0.852, p=0.015), Perceived Enjoyment (\(\rho \)= \(-\)0.923, p=0.003), Perceived Usefulness (\(\rho \)= \(-\)0.811, p=0.027), and Trust (\(\rho \)= \(-\)0.849, p=0.016) of the UTAUT-1 and a significant correlation with Intention to Use (\(\rho \)= \(-\)0.854, p=0.014) of UTAUT-2; ii. a significant correlation of Memory with the Perceived Enjoyment (\(\rho \)= \(-\)0.766, p=0.044), Perceived Usefulness (\(\rho \)= \(-\)0.759, p=0.048) of UTAUT-2; iii. a significant correlation between the score on Verbal fluency and the sub-scale of Social Presence (\(\rho \)= 0.791, p=0.034) of UTAUT-2; iv. a significant correlation between the score on Language and the sub-scale of Perceived Sociability (\(\rho \)= \(-\)0.774, p=0.041) of UTAUT-2.

4.1.2 Robot Behavior and UTAUT

During the experimentation, the robot’s behaviour was modulated considering each patient’s needs. We instructed the robot to execute each day exactly 8 activities chosen among the 3 types available, i.e., medicine reminders, monitor and entertainment. According to the patient profile, doctors indicated the 8 activities more suited for each patient by specifying how many times to perform each of them in a day. Hence, we evaluated whether the number/type of activity impacts user acceptance by correlating them with UTAUT-2.

Results showed: i. a significant correlation between a high number of entertaining activities and a high score in the Social Presence (\(\rho \)=0.852, p=0.015); ii. a significant correlation between a high number of reminders of medicines and a high score in the Social Presence (\(\rho \)=0.841, p=0.018); iii. no correlations were found between monitoring activities and the number of experimentation days.

4.1.3 Adaptivity and Pleasantness

For participants of Phase-2, the level of pleasantness after the participant interacted with the robot was recorded every day and the average values are reported in Table 9. The score on pleasantness was significantly different when the robot was in “standard modality” and in “modulated modality”; in detail, the score on the pleasantness scale was higher when the interaction was with the robot set in “modulated modality” (mean=2.44, standard deviation=0.812) rather than “standard modality” (mean=1.93, standard deviation=0.842; U test= 487.00, p= 0.008).

Table 9 Overview of the Phase-2 participants
Full size table

To support the results on pleasantness experienced by the participants, we also collected information about their personal experiences after interacting with the robot for the whole experimentation time without any specific questions but just asking them to express their feelings freely. Most subjects commented positively on their direct experience with the robot and its usefulness as a tool that helps and keeps them company during the day.

Of course, this information is difficult to analyse. Still, it provided useful insights that only realistic experimentation carried out in their environments and during their usual daily living allows gathering.

Patient 1 84 years old woman with 13 years of education and with Alzheimer’s disease corresponding to a CDR value of 2. She does not have any familiarity with technology. She lives in wealthy conditions with her husband and an experienced caregiver. She receives regular visits from relatives. In the first days, the robot’s presence was seen with reluctance (once she also said “if it comes close I’ll kick it”) while at the end of the experiment, the participant became attached to its presence. The entertainments were initially not properly calibrated to the patient’s taste. The caregiver pointed out that the participant was very happy when she realised that the robot was always looking for her among the family members. The participant felt important to the robot as a human being. The most pleasant moments were during the musical entertainment activities when she started singing involving her husband playing the piano, as they were used to doing in the past (we were told by the caregiver).

Patient 2 61 years old men with 5 years of education and mild cognitive impairment corresponding to a CDR value from 0.5 to 1. He does not have any familiarity with technology. He lives in wealthy conditions with his wife and without any caregivers. He receives regular visits from relatives and his nephew. He showed particular care for the robot paying attention to a possibly dangerous situation for the robot’s safety. When the robot was taken away, he asked when the robot would be brought back. He was a bit upset that the robot could not follow him from room to room since he would have liked to have the robot close to him all the time.

Patient 3 55 years old woman with 8 years of education and with subjective memory disorder corresponding to a CDR value of 0. She has an average familiarity with technology. She lives in wealthy conditions with her husband and is without any caregivers. She receives regular visits from relatives. When the robot was taken away, she also asked when the robot will be brought back. The participant reports that the robot has been a very pleasant company, it has been useful in reminding the intake of medication, and in proposing entertainment activities. The participant would like to have the robot back in the future and said “This is my friend! Now it is leaving I feel more lonely as a person is missing at home”.

Patient 4 68 years old men with 8 years of education and mild cognitive impairment corresponding to a CDR value from 0.5 to 1. He states to have a high familiarity with technology. He lives with his wife without any caregivers. He receives regular visits from relatives. The participant states that the robot does not seem to be complete, it is not adapted to his needs, and he would like it to be more active; however, he would still use it in case the robot is designed exactly for his needs and he provided an average daily evaluation of 1.5.

Patient 5 83 years old woman with 18 years of education and with subjective memory disorder corresponding to a CDR value of 0. She has little familiarity with technology. She lives with her daughter working all day, so she is alone most of the day and without any caregiver. She lives in wealthy conditions. The participant states that the experience with the robot has been very positive, she would still use it but he would like the robot to do many more things with her. It reminded her correctly of the medications and interacted with her with 15-minute activities. She said “I feel like I have a child to look after and it’s nice, I feel occupied”. She also perceived the robot as a real person attributing behaviours that were not occurring like performing a specific dance while it was only moving, and assuming it was offering a flower to her when a flower appeared on the touchscreen as a screensaver.

Patient 6 78 years old man with 8 years of education and mild cognitive impairment corresponding to a CDR value from 0.5 to 1. He does not have any familiarity with technology. He lives with his wife and son. The participant said: “I am thrilled to have used the robot. At times, I felt it was not ok, and I was very worried. I was very happy to have it at home and the time seemed too short. Since it left, I have felt more lonely. If I have the same opportunity again, I’d be glad to have him at home again. I was very excited about it. It was a virtual company that kept me active and busy, and we seemed like friends. It filled a part of my day, reminding me of the medication schedule and distracting me with songs and documentaries. Proud to have joined this new experiment that was successful for me. Thank you.”

Patient 7 77 years old men with 13 years of education and mild cognitive impairment corresponding to a CDR value from 0.5 to 1. He is highly familiar with technology, especially with computers he was able to program. He lives with his wife. The participant claims that the robot has been boring. He complained that he received a limited explanation of its functionality. The robot interacts only at certain intervals during the day and spends a lot of time doing nothing between its activities. He would not use it again and provided an average daily evaluation of 1.

5 Discussion

The study aimed to evaluate the degree of acceptability of a social humanoid robot in elderly people with cognitive impairments after direct interaction for a long-time. Moreover, we aimed to explore if the social humanoid robot’s degree of acceptability and pleasantness was related to the fact that the robot was customised according to the cognitive and personality profile. The participants’ personality was assessed through a specific personality test that identifies five personality traits: Neuroticism, Extraversion, Openness, Agreeableness, and Conscientiousness.

The results obtained on a small sample of participants who interacted with a humanoid robot at the laboratory did not show an increase nor a worsening of the level of acceptability after a direct interaction with the robot. We obtained the same results when we evaluated the acceptability of the humanoid robot in the elderly with cognitive deficits at home. The absence of a change between UTAUT-1 and UTAUT-2 might depend on some methodological factors: cognitive deficits may have prevented the recall of being helped in everyday life activities by the robot. We found that participants with worse performance in global cognitive batteries (MMSE and ACE-R) tended to not recognise the robot’s usefulness because they cannot perceive the ability of the robot to be adaptive to their needs even after a direct interaction. Further studies would be necessary to assess the acceptability and usefulness of robots for assisting the elderly by caregiver-rated assessments. However, the results showed greater pleasantness when the robot was set in modulated modality, according to the user’s personality and cognitive profile, compared to when the robot was set in standard modality. So while with UTAUT we could not find any significant differences, results on a daily base support our hypothesis H1. In future work, randomised experimentation considering patients interacting with either “modulated” or “standard” modality has to be conducted to evaluate the effects of UTAUT before and after the interaction and their relationships with the robot’s adaptation.

Regarding H2, results showed that the robot acceptability is influenced by the type of activities provided by the robot and, in particular, interactive tasks. For example, Social Presence increases when providing more entertainment activities tailored to the patient’s preferences. The same is in the case of Medicine Reminders.

The present study evaluated the association between personality traits and acceptability in the elderly with cognitive deficits (exploratory hypothesis H3). The results of the correlational analysis showed that personality traits such as Neuroticism, Extroversion, Openness, and Conscientiousness, distinctive of people with cognitive deficits [17], can influence the acceptability of a robot in the elderly only before a direct interaction, while these personalities traits seemed not to have any influence on the acceptability of the new technology after a direct interaction. In particular, the association between high levels of Neuroticism and a high score in the subscales of Attitude, Enjoyment, and Usefulness perceived of the UTAUT-1 would indicate that a tendency towards emotional instability (indicated by high levels of Neuroticism) may be associated with increased acceptance and predisposition to use the robot considered as a useful and advantageous support. PENJ, PU and PAD are also positively correlated with Openness. Moreover, a significant correlation between high levels of Openness and a low score in the sub-scale of PENJ, PU, and PAD of the UTAUT-1 would indicate that the tendency to be more involved in the social world and to be more nonconformist and open to culture would affect some aspects of the acceptability of the robot by reducing the perception of the robot as a useful support in everyday life. However, such personality traits seem not to affect any aspect of the acceptability of the robot after a long and direct interaction with the robot at home. Indeed, in the literature, these two traits (Neuroticism and Openness) are often associated with different perceptions and evaluations of robotic technologies. Once confirmed with further experimentation, the lack of correlations of personality traits with UTAUT-2 is indeed a positive indicator of the possible wider acceptability of this technology, meaning that long-term evaluation could mitigate initial (but also short-term) bias due to personality.

The same cannot be sustained in the case of cognitive profiles. Indeed, our results showed that cognitive impairments correlate with UTAUT-2, as in the case of Visuospatial, Memory, Verbal fluency and Language abilities. This sustains the need for a stronger adaptation process that considers cognitive profiles.

Considering demographic traits, results showed that education negatively correlates with ATT, PENJ, PS, PU, and Trust of UTAUT-1, while, after the interaction, it negatively correlates only with SI of UTAUT-2. Again, this can be interpreted, once confirmed with additional experiments, that the level of education may not have a strong impact on the use of such technology in the long term. This is confirmed by the individual comments provided by the patients of Phase-2, for example, P2. Different is the case of familiarity with technology where no correlation occurs with UTAUT-1 while it negatively correlates with PS, PU, and UTAUT-2 total score. Hence, higher expectations on technology availability and readiness impact the case of experimentation with simple robotic applications. In particular, P4 and P7 were the patients with a greater familiarity with technologies, with higher expectations, and provided the lowest average daily scores (1.5 and 1).

In the present study, we also evaluated the association between emotional intelligence measured by EQ and the acceptability of robots. In detail, we found that people with a high level of emotional intelligence tend to be less willing to use the robot, before the real interaction, and feel pleasure in using it; this result might be interpreted as a support to the idea that the tendency to be more involved in the social world would affect the acceptability of the robot.

5.1 Limitations

The present study has strengths but also limitations. Specifically, using a social robot in a real context (participant’s home) is the first strength of the study. The evaluation of the personality and cognitive profile of the participants is the second strength since it allowed modulation of the robot’s behaviour according to the needs and personality profile of people with cognitive deficits to improve acceptability and pleasantness.

However, the present study has the following limitations: the sample size was small, but we had great difficulty in recruiting a larger number of people because their home was incompatible with the movements of the robot due to specific physical characteristics of the house (presence of steps, carpets, pets, absence of Wi-Fi network). Therefore, confirming our results with a larger sample of participants will be relevant.

Another limitation could be the design of the experiments that provided both the modulated and the standard modality to the same participants. This might have hindered participants’ perception of the robot’s adaptation. Studies allowing a more prolonged interaction between robots and subjects with cognitive deficits and with a different design protocol could deepen the knowledge about the level of acceptability and its relationship with adaptation. This is why the experimental evaluation with patients lasted for a variable time according to the availability of the patients to collect information for as long time as possible, and to highlight potential differences due to the time spent with the robot. Even though the variability in time could significantly impact the statistical analysis, with only 7 participants the aim of the study is more on evaluating the feasibility of the approach than on generalising it.

Another limitation could be the failure to distinguish between subjects with major cognitive impairment and minor cognitive impairment (according to DSM 5) who may have a different level of acceptability of the robot in daily life. This issue deserves to be investigated in further studies.

Finally, the robot’s lack of natural language dialogue capabilities is also a limitation. Nowadays the availability of vocal personal assistants puts unfulfilled expectations on the interaction modalities of the robot.

6 Conclusions

The use of technological assistance systems for people with neurodegenerative diseases could represent an innovative care strategy. The advantage of this type of care could have two objectives: on the one hand, it could make the patient autonomous for a longer period of time, delaying entry into nursing homes or home care h24, reducing social isolation and slowing cognitive impairment; on the other hand, it could support and reduce the physical and psychological distress of caregivers in managing an increasingly widespread disease. Although social humanoid robots can provide a solution for the ageing population challenge, they may generate discomfort in elderly people usually reluctant to use a technology perceived as dangerous. To study the determinant factors that make a social robot acceptable or not to the potential user, an evaluation of the subjective evaluation and perception of the technology before its actual use is carried out. To study the acceptance of the technology, objective measures of acceptability and subjective measures concerning perceptions after use and satisfaction with the use of the device are collected. Finally, an a posteriori analysis of use is carried out to evaluate the appropriation of the technology when integrated into everyday life.

Our study showed that the acceptability of a humanoid robot in the elderly with cognitive deficits is related to the customisation of the robot according to the user’s personality traits and cognitive status. This evidence is also supported by the positive comments of the participants at the end of the study. Therefore, identifying the personality traits and the cognitive status seems useful to modulate the type and frequency of interaction of the robot with the user to increase the acceptability of the instrument and pleasantness in daily life. Indeed, our results confirmed that personalization of the robot behaviour is a key factor for effective deployments of such assistive technologies since it has an impact on acceptance.

As personalization is directly linked to cognitive impairments, in future work, we plan to investigate the impact of the different grades of such impairments on the robot’s acceptability. This study requires a different set of patients since it was not possible with the available number of patients and their uneven distribution in each considered impairment degree.