This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR mHealth and uHealth, is properly cited. The complete bibliographic information, a link to the original publication on https://mhealth.jmir.org/, as well as this copyright and license information must be included.
As the world’s population rapidly ages, the number of older adults with cognitive impairment will also increase. Several studies have identified numerous complex needs of people with dementia, which assistive technologies still fail to support. Recent trends have led to an increasing focus on the use of embodied conversational agents (ECAs) as virtual entities able to interact with a person through natural and familiar verbal and nonverbal communication. The use of ECAs could improve the accessibility and acceptance of assistive technologies matching those high-level needs that are not well covered to date.
The aim of this thematic literature analysis was to map current studies in the field of designing ECAs for patients with dementia in order to identify the existing research trend and possible gaps that need to be covered in the near future. The review questions in this study were as follows: (1) what research frameworks are used to study the interaction between patients with dementia and ECAs? (2) what are the findings? and (3) what are the barriers reported in these studies?
Separate literature searches were conducted in PubMed, Web of Science, Scopus, and Embase databases by using specific umbrella phrases to target the population (patients with dementia) and the technology-based intervention (embodied conversational agent). Studies that met the inclusion criteria were appraised through the Mixed Methods Appraisal Tool and then discussed in a thematic analysis.
The search process identified 115 records from the databases and study references. After duplicates (n=45) were removed, 70 papers remained for the initial screening. A total of 7 studies were finally included in the qualitative synthesis. A thematic analysis of the reviewed studies identified major themes and subthemes: the research frameworks used to gather users’ perspectives on ECAs (theme 1), the insights shared by the 7 studies as well as the value of user involvement in the development phases and the challenge of matching the system functionalities with the users’ needs (theme 2), and the main methodological and technical problems faced by each study team (theme 3).
Our thematic literature analysis shows that the field of ECAs is novel and poorly discussed in the scientific community and that more sophisticated study designs and proofs of efficacy of the approach are required. Therefore, by analyzing the main topic of the narrative review, this study underscores the challenge of synchronizing and harmonizing knowledge, efforts, and challenges in the dementia care field and its person-centered paradigm through the user-centered design approach. Enabling strict collaboration between interdisciplinary research networks, medical scientists, technology developers, patients, and their formal and informal caregivers is still a great challenge in the field of technologies for older adults.
The world’s population is rapidly aging and approximately 47 million people are now experiencing dementia worldwide. This number could triple by 2050 with an incremental estimated cost that will range from US $818 billion in 2015 to US $2 trillion by 2030 [
Since the challenges of responding to the growing number of people with dementia and their complex needs are substantial for governments [
The use of ECAs could improve the accessibility and acceptance of computer-based assistive technologies when compared to graphical user and voice interfaces, especially for older adults with cognitive impairment [
The aim of this thematic literature analysis was to map current studies in the field of designing ECAs for patients with dementia in order to identify the existing research trends and possible gaps that need to be covered in the near future. The review questions were (1) what research frameworks are used to study the interaction between patients with dementia and ECAs? (2) what are the findings? and (3) what are the barriers reported in these studies?
Separate literature searches were conducted in PubMed, Web of Science, Scopus, and Embase databases by using the following umbrella phrases to target the population and the specific technology-based intervention: (“patient with dementia” OR “people with dementia” OR “person with dementia”) AND (“virtual agent” OR “personal virtual assistant” OR “virtual companion” OR “embodied conversational agent”). Inclusion criteria were published papers written in English with the aim of studying the use of ECAs (1) among older adults (≥65 years) with dementia living at home, in long-term care, or nursing homes and their formal and informal caregivers, (2) for coping in patients with dementia without any restriction in terms of service applications (ie, cognitive games, reminders, medicine intake, calendar, etc), (3) for presenting empirical findings about interactions between users and ECAs, and (4) in randomized controlled trials (qualitative, quantitative, and the mixed methods approach were included). There was no restriction on publication dates, and the searches were finalized in July 2020. Papers were excluded if reviews, theoretical or technical studies, and contributions were not original research papers that met the inclusion criteria or were not written in English. According to predefined criteria, the screening phase was based on the analysis of titles and then abstracts. Later, full texts of those titles/abstracts of screened publications were reviewed independently by the first and the corresponding author in August 2020. Another researcher was involved in reaching consensus in cases of disagreement. Studies that met the inclusion criteria were included, and the results of the searches were summarized. Then, we performed a manual thematic analysis of the findings. We used the Preferred Reporting Items for Systematic Reviews and Meta-Analysis [
Flow diagram of the studies included in the thematic review as well as the main reasons for rejection.
Three authors independently appraised the final papers for their methodological quality by using the Mixed Methods Appraisal Tool version 2018 [
A thematic analysis was conducted to identify themes based on the 6 phases described by Braun and Clarke [
The search process identified 95 records from the databases and an additional 20 by manually searching those studies’ references. After duplicates (n=45) were removed, 70 papers remained for initial screening by title. This resulted in 37 potentially eligible abstracts. The abstracts retained were analyzed by authors according to the research questions in order to obtain the final list of full-text papers to be reviewed. After the analysis of the abstracts, 33 of them were excluded as they did not fit the settled criteria of the target population and the specific technology-based intervention. A second screening step was performed for those full-text papers that matched all the criteria (n=19). From this process, 12 papers were excluded because (1) studies recruited both patients with dementia and older adults in good health status (n=3), (2) no empirical feedback from users was reported (n=3), (3) the intervention was performed using both ECAs and other technologies (n=2), and (4) the intervention was not specific for patients with dementia (n=4). Finally, a total of 7 studies were included.
A summary of the studies and their findings are presented in
The different agents are shown in
Summary of the reviewed studies.
Studies | Purpose | Type of system | Method for data collection | Sample | Country, test setting | Findings |
Carrasco et al [ |
To validate a functional prototype that gives a measure on how natural the interaction between avatars and people with Alzheimer disease is. |
Avatar displayed on a standard television set | Yes/No questions and one-to-one observation | 21 persons had Alzheimer disease, with a Global Deterioration Scale [ |
Spain, day care center | All users engaged naturally with the avatar, understood the information conveyed by the avatar, and answered successfully by means of the television remote control |
Yasuda et al [ |
To investigate the effectiveness of a conversation system based on an animated face of a child | A computer screen that shows an animated face of a child agent | Qualitative interviews | 8 older adults (2 males and 6 females) had mild Alzheimer disease, with a Mini-Mental State Examination [ |
Japan, hospital | All users conversed with the conversational agent system and enjoyed the conversation. |
Wargnier et al [ |
To collect design guidelines to develop a semiautomated ECAa prototype | A semiautomated cartoon like ECA prototype that runs on a standard personal computer with Microsoft Windows | Semiautomated Wizard of Oz, video, observation, open interview, questionnaire | 14 specialists (4 males and 10 females) in assistive technologies for older adults or care professionals (medical doctors and neuropsychologists, mostly) | France, hospital | All participants interacted naturally with the ECA. Most displayed high levels of attention. Globally, the feedbacks turned out to be rather positive. |
Konig et al [ |
To identify affective identities in patients with dementia for the design of cognitive assistive technologies | An intelligent cognitive assistant in the form of a humanoid female character shown on a screen | Qualitative interview | 12 older adult care home residents (5 males and 7 females) with Alzheimer disease who showed cognitive and functional impairment to an extent that it affected their autonomy in performing certain complex activities of daily living and 9 associated caregivers (2 males and 7 females). The average age of the residents was 84.5 years. | Canada, University and Research Institute for Aging | Definition of user requirements for the design |
Wargnier et al [ |
To conduct a usability study to refine and validate the Louise ECA | A semiautomated cartoon like ECA prototype that runs on a standard personal computer under Microsoft Windows | Realistic assistive scenarios and semistructured interview | 14 participants (3 males and 11 females) with mild cognitive impairment (9/14) or Alzheimer disease (5/14), whose Mini-Mental State Examination [ |
France, hospital and University | Most of the participants were able to interact with the ECA, succeeded in completing the proposed tasks, and enjoyed the design |
de Jong et al [ |
Report the first iteration of a comprehensive user-centered development process of virtual agents for patients with dementia and their caregivers | A personal assistant called Anne that works on a Surface Pro tablet under the Microsoft Windows 10 operating system | Focus group | 16 caregivers: 10 in Luxembourg (6 qualified nursing assistants and 4 informal carers) and 6 in Italy (3 care professionals and 3 informal caregivers) | Luxembourg, Italy, hospital, day care center | Definition of user requirements for the design |
Stara et al [ |
How patients xperience a personal virtual assistant in the stage of moderate dementia; how a personal virtual assistant can be modified to the requirements of people in the stage of moderate dementia | A personal assistant called Anne that works on a Surface Pro tablet under the Microsoft Windows 10 operating system | Protected environment test scenarios with observation of the interactions between patients and the personal virtual assistant and interview to formal caregivers | 5 female patients with moderate dementia and 2 formal caregivers in Italy; 1 female patient with dementia and 2 formal caregivers in Luxembourg | Italy, Luxembourg, hospital, day care center | Definition of user requirements for the design |
aECA: embodied conversational agent.
Embodied conversational agents described in the reported studies. A) Carrasco et al [
The design of the 7 research studies was assessed by using screening questions and the 5 criteria of the Mixed Methods Appraisal Tool [
S1. Are there clear research questions?
S2. Do the collected data allow to address the research questions?
Q1. Is the qualitative approach appropriate to answer the research question?
Q2. Are the qualitative data collection methods adequate to address the research question?
Q3. Are the findings adequately derived from the data?
Q4. Is the interpretation of results sufficiently substantiated by data?
Q5. Is there coherence between qualitative data sources, collection, analysis, and interpretation?
M1. Is there an adequate rationale for using a mixed methods design to address the research question?
M2. Are the different components of the study effectively integrated to answer the research question?
M3. Are the outputs of the integration of qualitative and quantitative components adequately interpreted?
M4. Are divergences and inconsistencies between quantitative and qualitative results adequately addressed?
M5. Do the different components of the study adhere to the quality criteria of each tradition of the methods involved?
All the studies used appropriate research design, taking into account the research questions and problems related to the use of a specific innovation technology among persons with dementia. The prevalent data collection methods were open or semistructured interviews [
Quality scores of the selected studies.
Studies | Screening question score | Qualitative studies score | Mixed methods studies score | Total score | Appropriate criteria (n) | Quantity score (%) | Score category |
Carrasco et al [ |
2 | N/Aa | 5 | 7 | 7 | 100 | High |
Yasuda et al [ |
2 | 5 | N/A | 7 | 7 | 100 | High |
Wargnier et al [ |
2 | N/A | 5 | 7 | 7 | 100 | High |
Konig et al [ |
2 | 5 | N/A | 7 | 7 | 100 | High |
Wargnier et al [ |
2 | N/A | 5 | 7 | 7 | 100 | High |
de Jong et al [ |
2 | 5 | N/A | 7 | 7 | 100 | High |
Stara et al [ |
2 | 5 | N/A | 7 | 7 | 100 | High |
aN/A: not applicable.
Following the analysis of reviewed studies, 3 major themes and subthemes within each theme were identified: (1) research frameworks, (2) efficacy of ECAs, (3) limitations of the studies and problems faced.
All the studies dealt with 2 research questions: (1) could virtual agent be a technology that patients with dementia can really use? and (2) which are the design features that can facilitate or hinder this usage? Methodologically, both qualitative and quantitative designs were applied to answer these questions focusing on meanings and understanding of experiences of people with dementia and their carers. In all cases, participants took part in the researches in participatory sessions or observations to avoid the discomfort of being the subjects of an experimental study. Overall, the 7 studies came to the general assumption that the use of ECAs is suitable for people with dementia. This common evidence based its foundation on the use of voice as an interaction modality between the systems. The voice as input/output modality is the natural and familiar way to engage people with dementia in such researches. Therefore, from 2008 to 2019, even though the readiness level of the virtual agents was changed considerably, especially in the human-looking appearance, none of the studies shared skepticism or disadvantages in the use of ECAs by the enrolled patients. Additionally, the use of common screen devices as the presentation platform on the television [
As previously mentioned, all the studies reported positive feedback on the use of agents by users: the majority of patients with dementia naturally interacted and responded to the virtual character fulfilling the assigned tasks [
User-centered design as “a philosophy based on the needs and interests of the user, with an emphasis on making products usable and understandable” [
People with dementia have many changing needs during the progression of their disease, varying from memory support to almost all aspects of daily functioning. In these studies, specific ECA functions matched the physiological, comfort, and attachment needs. ECAs described in 3 of the included studies [
Common limitations and problems are reported in the selected studies. The challenge of small samples is clearly mentioned even if they underwent the preliminary evaluation of the systems [
This review surveyed the literature on the usage of ECAs by patients with dementia with the aim to identify the current research trends and possible gaps to cover in the future. Three main questions piloted this study: (1) what research frameworks are used to study the interaction between persons with dementia and ECAs? (2) what are the findings? and (3) what are the barriers? Only 7 papers were returned from the search. Overall, the main findings of this narrative review demonstrate that research on ECAs as an innovative way to cope with dementia is little covered in the state of the art even though interesting topics emerged from the mapped studies: the research design used to perform such studies (theme 1), major findings, the value of a user-driven approach, the importance of a well-balanced matching between the system functionalities and the users’ needs (theme 2), as well as the reported problems faced by each study team (theme 3). Therefore, this section will discuss the implications raised through the lens of each theme reported in this study.
By examining the research frameworks of studies included in this review, it clearly emerges that the use of ECAs deserves (1) a more sophisticated study design and (2) proof of the efficacy of the approach, as in any other technology designed for people with dementia. The key to managing both demands is to directly involve older adults with dementia in the design of services dedicated to them. The early engagement of users from the outset and across all stages of the development cycle is relevant for people affected by dementia, since they progressively lose the ability to generalize between past and present experiences or to modify cognitive representations. For this reason, familiarity with the technology to be used is to be firmly considered and needs to be carefully planned. Indeed, while healthy adults are easily able to manage routine changes such as introducing a new device into their home environment, for older adults with dementia, these novelties can become extremely distressing and disorientating. Moreover, different limitations can overburden users: limitations in knowledge and understanding of the technology and limitations in communication between the user and the technology. As discussed in themes 1 and 2, the use of a common television or computer screen and the possibility of using verbal response as an input/output mode enables a more natural way of interaction. This is a valuable benefit for people with mild and moderate dementia. According to Kaplan and Kaplan [
The concept of familiarity is not the only principle to follow in the field of designing ECAs for persons with dementia. In the last decade, we conceived the important shift to a model of care centered on the person, which broke the traditional disease-focused approach. Thanks to this new paradigm, care and support are seen as ways to prevent functional decline, frailty, and disability [
This new paradigm emphasizes the power of self-determination over decisions that affect the individual’s body and mind. Therefore, the individual dignity and autonomy, which are the primary values and the fundamental rights of every human being, are restored. In this new vision, patients actively participate in clinical decisions outside the old schema of only being a sick person who needs to be treated. Nowadays, well-being is the goal of dementia care that offers individualized interventions and considers the person as a whole, thus considering individuals’ medical, cognitive, psychological, environmental, cultural, and social needs [
Another instance of harmonization is to focus on personhood and needs by clarifying what functionality and attributes are important in the new products for target users, what motivates them to use a product, what factors would hinder a positive user experience with a proposed product, and to conceptualize how parts of their lives could be improved by technology. Across the 3 themes analyzed in this review, the user experience of ECAs could be improved firstly by responding to the changing needs of people with dementia. This matching will enable technologies that better support the quality of life of people with dementia. This is particularly highlighted in the studies with more advanced ECAs such as Louise [
To the best of our knowledge, no other narrative reviews are reported in the literature regarding the research frameworks used to study the interaction between persons with dementia and ECAs and between the mapped outcomes and barriers. Despite this positive aspect, there are some limitations to this review. Data sources were drawn from only 4 databases (ie, Scopus, Web of Science, PubMed, and Embase) and accessed only during a specific period of time (July 2020). The choice of using specific phrases to target the population (“patient with dementia” OR “people with dementia” OR “person with dementia”) and the specific technology-based intervention (“virtual agent” OR “personal virtual assistant” OR “virtual companion” OR “embodied conversational agent”) could have omitted some results from the search. It is possible that other literary sources were available in other unselected databases. However, well-known and broad-spectrum databases were used in this review. Moreover, we collected a relatively small sample of studies and excluded non-English language studies. Therefore, even if the 7 studies included in this paper were homogeneous in terms of their qualitative research design and their meeting our inclusion criteria, this may have created some biases. In addition, some authors of this review are co-authors in 2 of the reported studies [
This review mapped the actual use of ECAs in the research field of dementia. The readiness level of this specific technology-based intervention grew across the years, shifting from to be initially displayed on a standard television set [
This narrative review summarizes the current research on ECAs for patients with dementia. Technologically, these artificial characters are very interesting and the mapped studies shared promising results in terms of engagement of patients. Unfortunately, until now, it has been difficult to prove that ECAs are effective and more efforts need to be spent to achieve to this evidence. Therefore, our thematic analysis reported on 3 main themes, namely, the research frameworks used to gather users’ perspectives on ECAs (theme 1), the valuable insights shared by the 7 studies as well as the value of user involvement in the development phases and the challenge of matching the system functionalities with users’ needs (theme 2), and the main methodological and technical problems faced by each study team (theme 3). It emerged that this specific field of research is novel and poorly discussed in the scientific community, but possible steps forward for the use of such systems in health care delivery are predictable. Moreover, analyzing the main metaphors across the studies, our work underscored the challenge to synchronize and harmonize knowledge, efforts, and challenges within the dementia care field and its person-centered paradigm. This can be effectively possible by adopting the well-known but still little used user-centered design [
PRISMA checklist.
embodied conversational agent
This study is cofunded by the EU Active and Assisted Living Program (reference AAL-call-2016-102) and partially supported by Ricerca Corrente funding from the Italian Ministry of Health.
VS and MR conceptualized and designed the study and drafted the manuscript. LR, EF, and SP analyzed the data and reviewed the manuscript. All authors provided intellectual contributions and critical feedback and reviewed the final manuscript.
None declared.