https://orcid.org/0009-0006-7504-5934
Abstract
Integration of artificial intelligence (AI) in health and healthcare, especially for older adults, has significantly advanced healthcare delivery. AI technologies, with capabilities such as self-learning and pattern recognition, are employed to address social isolation and monitor older adults’ daily activities. However, rapid AI development often fails to consider the heterogeneous needs of older populations, which could exacerbate an existing digital divide and inequality. This scoping review examines older adults’ involvement in AI system design, implementation, and evaluation of AI systems in health and healthcare literature, emphasizing the necessity of their input for beneficial AI systems. We conducted a scoping review according to PRISMA-SCR. We reviewed 17 studies, finding that half of these studies (n = 8) engaged older adults during the design phase, a small number (n = 3) during the evaluation stage, and even fewer (n = 2) involved older adults in the implementation stage. Despite AI’s growing role, design processes often overlook older adults’ needs. Our findings emphasize the need for inclusive, participatory design approaches to address ethical and equity challenges, enhancing user engagement and relevance. We also highlight how these approaches address the needs of older adults and improve outcomes. Specifically, we integrated evidence showing the practical benefits of these approaches for better accessibility, usability, and engagement among older adults. Although AI has potential to improve healthcare delivery, these approaches must be part of broader efforts to ensure ethical, inclusive, and equitable AI practices, especially in gerontology.
Background
In recent years, the application of artificial intelligence (AI) in healthcare has seen great advances, introducing the potential for AI to redesign how health-related services are delivered and processed. The strength of AI lies in its ability to perform tasks that typically require human intelligence, including but not limited to self-learning, pattern recognition, prediction, decision-making, and argumentation (1). In the context of gerontology, AI is already being used for various purposes, such as addressing older adults’ social isolation and monitoring their activities of daily living to promote safety and well-being (2–4). Given the fast pace with which AI systems are being developed and deployed within healthcare and society in general, it is unclear how well these systems are designed to meet the needs of older adults. Older adults are often mistakenly stereotyped as a homogenous group that is frail and incompetent (5), and more often than not, this population has been excluded from the development and evaluation of AI. Excluding older adults from the design, implementation, or evaluation of AI systems can further exacerbate the already-existing societal ageism and digital divide (6).
Several approaches have been introduced to involve users in the design process, such as user-centered design, participatory design, and human-centered design. These frameworks commonly involve an iterative process of examining the interactions between the users and AI, resulting in frequent design experiments and prototyping (7,8). Although all design approaches aim to incorporate users’ feedback and input throughout the design process, there are slight differences in each approach. For instance, user-centered design focuses on incorporating the needs and preferences of the end-users, namely those who would ultimately be using the product, throughout the design process (7). Participatory approach views users more as equal partners who provide expertise from the conceptualization and ideation in the early stages of design and aims towards collective creativity between designers and users (9). Human-centered design goes beyond simply considering users as influential factors or resources for the AI system but rather positioning them at the center of the whole design process. This approach takes a more humanistic and holistic perspective, respecting the individuals’ prior experiences, interests, decision-making styles, and cultural context (8). These approaches facilitate collaboration and support older adults in actively shaping AI systems that address their needs and contexts in health and healthcare. Designers can also be more aware of the expectations older adults or stakeholders have toward the technology, and the situations in which they would be using it, allowing for more empathetic and creative designs (8). Furthermore, given that this population often feels ostracized and digitally excluded (10), acknowledging older adults as codesigners can lead to a sense of ownership and participation (11).
Despite the aforementioned benefits, it is unclear to what extent current AI systems are being developed while involving older adults in their design. Fischer et al. previously synthesized evidence to identify the purpose, nature, and consequences of older user involvement during the design of any technology systems. This review included 40 articles published between 2014 and 2018 that involved older adults in the development of various types of technology (eg, mobile applications, robots, assisted living technologies, and online platforms) (11). However, this systematic review did not specifically focus on AI and excluded articles that focused on persons with cognitive impairments, potentially overlooking technology targeting older adults with Alzheimer’s disease or related dementias (ADRD). Moreover, there is currently a lack of consensus on the best strategy for involving older adults when designing, implementing, or evaluating AI systems, resulting in inconsistent approaches and methods used across studies. These gaps in the literature highlight the need to comprehensively explore how older adults and stakeholders have currently been involved specifically when developing AI systems.
Research Question, Aims, and Objectives
This scoping review is conducted to answer the following overarching question: What is known from the current literature on engaging older adults and their caregivers, such as families, significant others, and clinicians, in AI design?
The objectives of this scoping review are as follows:
To describe the utility of engaging older adults and caregivers in the design, implementation, or evaluation of new health-related AI;
To explore the practices and strategies for stakeholder engagement for the design, implementation, or evaluation of health-related AI for older adults and caregivers;
To describe the challenges, barriers, and unintended consequences of stakeholder engagements in the design, implementation, and evaluation of health-related AI for older adults and caregivers.
Research Design and Methods
Scoping reviews are appropriate when research areas have not been extensively reviewed and when there is a need to identify gaps in existing literature (12). As there are not many studies discussing participatory technology design efforts for older adults, a scoping review is the most suitable approach to answer the overarching question and objectives. This review adheres to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Review (PRISMA-ScR), designed to ensure the transparency and completeness of synthesized evidence (13). The PRISMA flowchart describes this review’s identification process (see Figure 1).
Preferred reporting items for systematic reviews and meta-analyses (PRISMA). *Consider, if feasible to do so, reporting the number of records identified from each database or register searched (rather than the total number across all databases/registers). **If automation tools were used, indicate how many records were excluded by a human and how many were excluded by automation tools.
Study Eligibility
Population
To be included in the review, a study’s target population had to be older adults (ie, age 65 years old or older) and/or their family members, trusted others and clinicians involved in their care. The review excluded studies where the major target population was less than 65 years old and did not include and/or benefit older adults.
Intervention
The intervention of interest was a structured study that involved older adults or stakeholders during the design, implementation, or evaluation of AI systems (eg, participatory design, user-centered design, human-centered design, and usability testing).
Outcome
This review included studies where the ultimate goal of AI development pertained to older adult health, assessing their quality of life, well-being, and safety. It also required that the study be related to improvements in older adults’ quality of life and well-being, stakeholder input/feedback, and involvement in any of the stages of the development process: design, implementation, and evaluation. The review excluded studies where AI development was not intended to support older adults or improve their health outcomes.
Study Design
This review considered most study designs. Excluded were reviews, meta-analyses, and umbrella reviews. Nonoriginal articles were excluded, such as conference proceedings, abstracts, protocols, editorials, and book chapters (see Table 1 for further detail).
Inclusion and Exclusion Criteria
| Inclusion Criteria | Exclusion Criteria | |
|---|---|---|
| Population | Older adults and caregivers | The major population was less than 65 years old and did not include and/or benefit older adults. |
| Years of publication | 2008-2023 (past 15 years) | Published before 2008 |
| Design | Case-control, randomized and nonrandomized controlled trials, cross-sectional studies, case studies, observational cohort studies, Written in English | Reviews, meta-analysis, and umbrella reviews. Nonoriginal articles were excluded, such as conference proceedings, abstracts, protocols, editorials, and book chapters, grey papers |
| Intervention | Involved older adults or stakeholders during the design, implementation, or evaluation of AI systems (eg, participatory design, user-centered design, human-centered design, and usability testing). | Did not exclusively explain uses of AI systems during the design, implementation, and/or evaluation of the research |
| Outcomes | This review included studies where the outcome or goal of AI development was to improve older adult health, including their quality of life, well-being, and safety. It also required that the study be related to improvements, stakeholder input/feedback, and involvement in stages of the development process: Design, Implementation, Evaluation. | AI development does not help or improve older adults’ outcomes. |
| Inclusion Criteria | Exclusion Criteria | |
|---|---|---|
| Population | Older adults and caregivers | The major population was less than 65 years old and did not include and/or benefit older adults. |
| Years of publication | 2008-2023 (past 15 years) | Published before 2008 |
| Design | Case-control, randomized and nonrandomized controlled trials, cross-sectional studies, case studies, observational cohort studies, Written in English | Reviews, meta-analysis, and umbrella reviews. Nonoriginal articles were excluded, such as conference proceedings, abstracts, protocols, editorials, and book chapters, grey papers |
| Intervention | Involved older adults or stakeholders during the design, implementation, or evaluation of AI systems (eg, participatory design, user-centered design, human-centered design, and usability testing). | Did not exclusively explain uses of AI systems during the design, implementation, and/or evaluation of the research |
| Outcomes | This review included studies where the outcome or goal of AI development was to improve older adult health, including their quality of life, well-being, and safety. It also required that the study be related to improvements, stakeholder input/feedback, and involvement in stages of the development process: Design, Implementation, Evaluation. | AI development does not help or improve older adults’ outcomes. |
Inclusion and Exclusion Criteria
| Inclusion Criteria | Exclusion Criteria | |
|---|---|---|
| Population | Older adults and caregivers | The major population was less than 65 years old and did not include and/or benefit older adults. |
| Years of publication | 2008-2023 (past 15 years) | Published before 2008 |
| Design | Case-control, randomized and nonrandomized controlled trials, cross-sectional studies, case studies, observational cohort studies, Written in English | Reviews, meta-analysis, and umbrella reviews. Nonoriginal articles were excluded, such as conference proceedings, abstracts, protocols, editorials, and book chapters, grey papers |
| Intervention | Involved older adults or stakeholders during the design, implementation, or evaluation of AI systems (eg, participatory design, user-centered design, human-centered design, and usability testing). | Did not exclusively explain uses of AI systems during the design, implementation, and/or evaluation of the research |
| Outcomes | This review included studies where the outcome or goal of AI development was to improve older adult health, including their quality of life, well-being, and safety. It also required that the study be related to improvements, stakeholder input/feedback, and involvement in stages of the development process: Design, Implementation, Evaluation. | AI development does not help or improve older adults’ outcomes. |
| Inclusion Criteria | Exclusion Criteria | |
|---|---|---|
| Population | Older adults and caregivers | The major population was less than 65 years old and did not include and/or benefit older adults. |
| Years of publication | 2008-2023 (past 15 years) | Published before 2008 |
| Design | Case-control, randomized and nonrandomized controlled trials, cross-sectional studies, case studies, observational cohort studies, Written in English | Reviews, meta-analysis, and umbrella reviews. Nonoriginal articles were excluded, such as conference proceedings, abstracts, protocols, editorials, and book chapters, grey papers |
| Intervention | Involved older adults or stakeholders during the design, implementation, or evaluation of AI systems (eg, participatory design, user-centered design, human-centered design, and usability testing). | Did not exclusively explain uses of AI systems during the design, implementation, and/or evaluation of the research |
| Outcomes | This review included studies where the outcome or goal of AI development was to improve older adult health, including their quality of life, well-being, and safety. It also required that the study be related to improvements, stakeholder input/feedback, and involvement in stages of the development process: Design, Implementation, Evaluation. | AI development does not help or improve older adults’ outcomes. |
Search Strategies and Databases
Two coauthors, both research librarians (L.G. and S.M.), developed the search strategies and conducted the searches. They selected search terms based on the three primary keywords: “artificial intelligence,” “older adults,” and “participatory design.” To assist in creating the terms list, the librarians referred to three systematic reviews on a related topic (11,14,15). We conducted a scoping review in November 2023 according to PRISMA-SCR and searched nine electronic databases, both medical and engineering, with a search that focused on three main concepts: “artificial intelligence,” “older adults” and “participatory design. A full list of search terms from PubMed can be found below (See PubMed Search Terms) (see Table 2 for further detail). Searches conducted in several of the engineering databases had to be modified or simplified due to limitations in their specific search interfaces (eg, number of truncated words allowed in a search). The search was filtered by language (English) and publication date (2008–2023). Citations retrieved in the search outputs of the different databases were then imported into Covidence, a web-based software for conducting systematic reviews, to be screened.
Search Terms
| Terms | |
| Key word 1 | Artificial Intelligence,” “Artificial Intelligence”[MeSH Terms] |
| Key word 2 | “Aged”[MeSH Terms], Aged, “Aging”[MeSH Terms], aging, “old age,” elder*, senior*, “older people,” “older person*,” “old people,” “old person*,” “older adult*,” geriatr*, gerontology* |
| Key word 3 | “participatory design,” “participatory action research,” “participatory research,” “design research,” “patient participation,” “collaborative design,” “cooperative design,” “user experience design,” “interaction design,” “participatory technology development,” “inclusive design,” user-centered, “user centered,” person-centered, “person centered,” human-centered, “human centered,” community-based, “community based,” community-centered, “community centered,” codesign*, “co-design*,” cocreate*, “co-create*,” coproduc*, co-produc*, “user involved,” “user participat*,” “user contribut*,” “user engag*,” “consumer involve*,” “consumer participat*,” “consumer contribut*,” “consumer engag*” |
| Terms | |
| Key word 1 | Artificial Intelligence,” “Artificial Intelligence”[MeSH Terms] |
| Key word 2 | “Aged”[MeSH Terms], Aged, “Aging”[MeSH Terms], aging, “old age,” elder*, senior*, “older people,” “older person*,” “old people,” “old person*,” “older adult*,” geriatr*, gerontology* |
| Key word 3 | “participatory design,” “participatory action research,” “participatory research,” “design research,” “patient participation,” “collaborative design,” “cooperative design,” “user experience design,” “interaction design,” “participatory technology development,” “inclusive design,” user-centered, “user centered,” person-centered, “person centered,” human-centered, “human centered,” community-based, “community based,” community-centered, “community centered,” codesign*, “co-design*,” cocreate*, “co-create*,” coproduc*, co-produc*, “user involved,” “user participat*,” “user contribut*,” “user engag*,” “consumer involve*,” “consumer participat*,” “consumer contribut*,” “consumer engag*” |
Search Terms
| Terms | |
| Key word 1 | Artificial Intelligence,” “Artificial Intelligence”[MeSH Terms] |
| Key word 2 | “Aged”[MeSH Terms], Aged, “Aging”[MeSH Terms], aging, “old age,” elder*, senior*, “older people,” “older person*,” “old people,” “old person*,” “older adult*,” geriatr*, gerontology* |
| Key word 3 | “participatory design,” “participatory action research,” “participatory research,” “design research,” “patient participation,” “collaborative design,” “cooperative design,” “user experience design,” “interaction design,” “participatory technology development,” “inclusive design,” user-centered, “user centered,” person-centered, “person centered,” human-centered, “human centered,” community-based, “community based,” community-centered, “community centered,” codesign*, “co-design*,” cocreate*, “co-create*,” coproduc*, co-produc*, “user involved,” “user participat*,” “user contribut*,” “user engag*,” “consumer involve*,” “consumer participat*,” “consumer contribut*,” “consumer engag*” |
| Terms | |
| Key word 1 | Artificial Intelligence,” “Artificial Intelligence”[MeSH Terms] |
| Key word 2 | “Aged”[MeSH Terms], Aged, “Aging”[MeSH Terms], aging, “old age,” elder*, senior*, “older people,” “older person*,” “old people,” “old person*,” “older adult*,” geriatr*, gerontology* |
| Key word 3 | “participatory design,” “participatory action research,” “participatory research,” “design research,” “patient participation,” “collaborative design,” “cooperative design,” “user experience design,” “interaction design,” “participatory technology development,” “inclusive design,” user-centered, “user centered,” person-centered, “person centered,” human-centered, “human centered,” community-based, “community based,” community-centered, “community centered,” codesign*, “co-design*,” cocreate*, “co-create*,” coproduc*, co-produc*, “user involved,” “user participat*,” “user contribut*,” “user engag*,” “consumer involve*,” “consumer participat*,” “consumer contribut*,” “consumer engag*” |
Study Selections
A total of 270 duplicates were removed, leaving 694 studies for the initial title and abstract screening based on predetermined inclusion and exclusion criteria. Screening was done by two out of three authors (H.C., O.O., N.G.), with conflicts resolved by a fourth author (G.D.). After the primary screening, 2 authors independently reviewed the full text of 60 articles to determine eligibility. When discrepancies among the authors occurred, the reviewers discussed and reconciled the discrepancies with the additional author. After the full-text review, 43 studies were excluded because they were the wrong study types, focused on the wrong populations, had indications or objectives that did not consider older adults as stakeholders or participants in developing AI technology for older adults, or involved the wrong patient population and outcomes. The final review incorporated a total of 17 studies (see Figure 1).
Quality of studies
The Mixed Methods Appraisal Tool (MMAT) was used to assess the quality of the evidence. This tool appraises the quality of empirical studies effectively, providing a critical literature review (14). Three authors (H.C., O.O., N.G.) independently applied the MMAT criteria to each of the included studies and met for bias adjudication if there were disagreements in scoring. In the appraisal of multi-method articles, our approach prioritizes the predominant type of data presented by the authors. For instance, if 90% of the results focus on qualitative data, with only a brief paragraph dedicated to survey results, the article should be appraised primarily as a qualitative study rather than a mixed-method study. This methodological focus ensures a more accurate assessment of the study’s core contributions. In order to capture all current phenomena, all studies, including those of poor quality, were included because they still provided important value relevant to this review’s research questions and objectives. Any discrepancies between the authors were discussed until consensus was reached. This process helped to maintain the internal validity of studies, including selection bias and confounding bias.
Data Extraction
Data were extracted using an investigator-developed data extraction sheet in Microsoft Excel. The sheet included the following categories: citation, publication date, country, study location, study site, sample size, types of AI, stage of development (method), overall results, barriers, ethical considerations, cost to users, access issues and health disparity, training for user required, success and positive impact of the device, and limitations of the study.
Data Analysis and Synthesis of Studies
This review employed a narrative synthesis approach to systematically organize information by AI types, community participatory type, and caregivers’ outcomes. The narrative synthesis is appropriate for a scoping review to analyze diverse qualitative and quantitative information across disciplines (15).
Findings of the Review
Characteristics of Included Studies
After the process of title, abstract, and full-text screening, there were 17 papers included in this review (see Table 3 for further detail). The papers were published between 2012 and 2023, with the majority of the articles being published after 2019. The first authors’ affiliations by continent included Europe (n = 8; 18–25), North America (n = 5; 26–30), Asia (n = 2; 31,32) and multiple sites in different countries (n = 2; 33,34). The studies evaluated included qualitative methods (52.9%, n = 9; 18, 19, 22, 23, 24, 27, 28, 29, 33), quantitative methods (29.4%, n = 5; 20, 21, 26, 31, 34), and mixed methods (17.6%, n = 3; 30, 28, 32). The majority of the studies focused on older adults as their primary population of interest (52.9%, n = 9; 20, 21, 22, 25, 26, 27, 29, 31, 33), whereas one study only included geriatric experts (5.8%, n = 1; 19), and the remaining studies included both older adults and experts (41.2%, n = 7; 18, 23, 24, 28, 30, 32, 34). Only 2 studies (11.8%) specifically targeted AI development for individuals with AD/ADRD (30,31).
Summary of Studies Included in Review
| Authors (Year), Countrya | Type of AI | Aim of Study | Study Design | Sample / Stakeholders Characteristics | Stage of Development | Engagement Approach (Method) |
|---|---|---|---|---|---|---|
| Cantone et al., (2023), Italy (16) | Humanoid robot | To develop a comprehensive architecture that integrates robots, sensors, and AI, to provide care for elderly individuals at home | Qualitative | 10 doctors with experience in geriatrics and internal medicine (age range 35–68; female 40%), 100% Italian | Design | User-centered design (Interviews) |
| Cinini et al. (2021), Italy (17) | Caregiver robot | (1) to investigate older adults’ perception of acceptable technology and their acceptability of caregiver robots (2) To monitor linguistic behavior and spoken language productions of elderly subjects over the time (3) To automate the training and the monitoring of motor and cognitive functions of older adults | Quantitative non-RCT | (1) Planning stage: 100 older adults (age range 65–81; 64% male) for first experiment, 202 older adults (age range 65–87; 41% male) living independently for second experiment (2) Wellbeing assessment: 58 older adults (mean age 80; female 74%) (3) Motor and cognitive function assessment: 16 older adults (age range 65–78; female 56.3%) 100% Italian | Design and Implementation | Active user engagement / User-centered (Survey) |
| D’Onofrio et al. (2019), Italy and Japan (18) | Companion robot (Buddy) | To provide a pilot qualitative analysis of the needs of older adults and their caregivers when given conversational activities with robots | Qualitative | 17 older adults (10 from Italy (age range 65–81, male 70%); 7 from Japan (age range 63–92, male 14.3%)) 36 caregivers (30 from Italy (age range 33–68, male 43.3%; 6 from Japan (age unknown, male 66.6%)) 30 Italian, 6 Japanese | Design | Cocreation methodology participatory approach (Workshops) |
| Easton et al. (2019), UK (19) | Virtual agent system | To codesign the content, functionality, and interface modalities of an autonomous virtual agent to support self-management for patients with COPD and then to assess the acceptability and system content | Qualitative | Workshop #1: 5 individuals with COPD and 1caregiver (age range 69-86; female 60%) Workshop #2: 4 individuals with COPD (age range 66–80; female 50%) | Design | Participatory approach (Workshops) |
| Eun et al. (2023), South Korea (20) | Exercise game | To design a digital healthcare application aiming at improving the older adults’ physical and cognitive functions | Quantitative non-RCT | 25 older adults (mean age range 68–80) 100% Korean | Implementation | User-centered |
| Fraune et al. (2022), US and Japan (21) | Socially facilitative robot | To identify middle-aged and older adults’ challenges with current technology-mediated social interactions, and then brainstorm socially-facilitative robot concepts to address their stated needs and wants. | Qualitative | 19 middle-aged or older adults (7 from US (age range 51–68; female 57.1%); 12 from Japan (age range 50–63; female 50%) | Design | Participatory design (Workshops) |
| Garcia-Mendez et al. (2021), Spain | Entertainment chatbot (EBER) | To report the experimental tests of EBER chatbot with real users | Quantitative non-RCT | 31 older adults (mean age 75.5, female 64.5%) | Implementation and evaluation | User-centered approach |
| Gasteiger et al. (2022), New Zealand (22) | Daily care service robot (Bomy) | To evaluate the usefulness and older adults’ perception and experience of using daily care robot through unrestricted and unsupervised use at home | Qualitative | 6 older adults (age range 72–83; female 66.7%) | Implementation | Participatory approach: Co-design (Interviews) |
| Gyrard et al. (2023), Europe and Japan (23) | Social companion robots | To explore how internet of robotic things technology and cocreation methodologies help to design emotional-based robotic application | Quantitative non-RCT | 100 older adults and caregivers | Design | Participatory Approach—codesign (Workshops) |
| Kim et al. (2022), South Korea (24) | Robot (Dori) | To design the care robot’s services based on sensing movement during daily activities | Mixed methods | 25 nurses and medical workers (female 80%); 22 nursing care workers (90.9% female); caregivers? | Evaluation | Human-centered AI framework (Survey) |
| Louie et al. (2014), Canada (25) | Socially assistive robot (Brian 2.1) | To investigate the acceptance and attitudes of older adults toward the human-like expressive socially assistive robot Brian 2.1 | Quantitative non-RCT | 46 older adults (age range 62–91; female 80.4%) | Evaluation | User-centered (Demonstration) |
| Mahmoudi Asl et al. (2023), Netherlands and Spain | Social robot (MINI) | To investigate the attitude of stakeholders and potential facilitators and the barriers to implementing the social robot MINI in community-based meeting centers for people with dementia and carers | Qualitative | 12 people with dementia and 11 stakeholders | Design | User-centered (Demonstration) |
| Martin-Hammond et al. (2019), USA (26) | Intelligent assistant | To understand older adults’ perspectives regarding intelligent assistants for health information management and search | Qualitative | 18 older adults (age range 61–93; 83.3% female) | Design | Participatory approach: (Multiphases) |
| Murawski et al. (2024), USA (27) | Online AI-based caregiver negotiation program (NegotiAge) | To develop and pilot test an artificial-intelligence negotiation training program, NegotiAge, for family caregivers. | Mixed methods | 12 family caregivers of older adults who have cognitive loss | Design and Evaluation | Usability study/ participatory approach (Multiphases) |
| Muuraiskangas et al. (2012), Germany (28) | Virtual Coach (V2me) | To examine user experience and acceptance of a prototype of the virtual coach (V2me) system for deriving the requirements for a system with social networking and friendship functionality | Qualitative | Face-to-face interviews: 30 older adults (age range 69–90; female 63.3%) One-on-one usability testing: 6 participants (mean age 78.8, female 50%) Workshop: 7 participants (mean age 80.6, female 100%) | Design, implementation, evaluation | User-centered design (Interviews + Workshops) |
| Petersen et al. (2020), USA (29) | Mobile app for monitoring the use of a bluetooth-connected resistance exercise band app. | To develop a mobile app with both older adults and clinicians while exploring whether data collected through this process can be used in NLP and sentiment analysis | Mixed methods | 6 clinicians (age range 37–51; male 33.3%) and 16 older adults (age range 66–85; male 18.8%) | Evaluation | User-centered design (Multiphases) |
| Tiersen et al. (2021), UK (30) | Smart home | To investigate the functional, psychosocial, and environmental needs of people living with dementia, their caregivers, clinicians, and health and social care service providers regarding smart home systems. | Qualitative | (1) Interview participants (3 rounds) –9 people with dementia, 9 caregivers –10 academic and clinical staff –10 caregivers and 2 people with dementia (2) Focus group participants (2 rounds) – 2 managers – 6 stakeholders (3) Workshop participants (3 rounds) –12 health and social care clinicians –35 pairs of people with dementia and caregivers –24 participants (14 occupational therapists; 4 National Health Service pathway directors; 6 researchers in occupational therapy, neuropsychiatry, and engineering) | Design | Participatory design (Multiphase: Interviews, Focus group, Workshops, Ethnographic observation) |
| Authors (Year), Countrya | Type of AI | Aim of Study | Study Design | Sample / Stakeholders Characteristics | Stage of Development | Engagement Approach (Method) |
|---|---|---|---|---|---|---|
| Cantone et al., (2023), Italy (16) | Humanoid robot | To develop a comprehensive architecture that integrates robots, sensors, and AI, to provide care for elderly individuals at home | Qualitative | 10 doctors with experience in geriatrics and internal medicine (age range 35–68; female 40%), 100% Italian | Design | User-centered design (Interviews) |
| Cinini et al. (2021), Italy (17) | Caregiver robot | (1) to investigate older adults’ perception of acceptable technology and their acceptability of caregiver robots (2) To monitor linguistic behavior and spoken language productions of elderly subjects over the time (3) To automate the training and the monitoring of motor and cognitive functions of older adults | Quantitative non-RCT | (1) Planning stage: 100 older adults (age range 65–81; 64% male) for first experiment, 202 older adults (age range 65–87; 41% male) living independently for second experiment (2) Wellbeing assessment: 58 older adults (mean age 80; female 74%) (3) Motor and cognitive function assessment: 16 older adults (age range 65–78; female 56.3%) 100% Italian | Design and Implementation | Active user engagement / User-centered (Survey) |
| D’Onofrio et al. (2019), Italy and Japan (18) | Companion robot (Buddy) | To provide a pilot qualitative analysis of the needs of older adults and their caregivers when given conversational activities with robots | Qualitative | 17 older adults (10 from Italy (age range 65–81, male 70%); 7 from Japan (age range 63–92, male 14.3%)) 36 caregivers (30 from Italy (age range 33–68, male 43.3%; 6 from Japan (age unknown, male 66.6%)) 30 Italian, 6 Japanese | Design | Cocreation methodology participatory approach (Workshops) |
| Easton et al. (2019), UK (19) | Virtual agent system | To codesign the content, functionality, and interface modalities of an autonomous virtual agent to support self-management for patients with COPD and then to assess the acceptability and system content | Qualitative | Workshop #1: 5 individuals with COPD and 1caregiver (age range 69-86; female 60%) Workshop #2: 4 individuals with COPD (age range 66–80; female 50%) | Design | Participatory approach (Workshops) |
| Eun et al. (2023), South Korea (20) | Exercise game | To design a digital healthcare application aiming at improving the older adults’ physical and cognitive functions | Quantitative non-RCT | 25 older adults (mean age range 68–80) 100% Korean | Implementation | User-centered |
| Fraune et al. (2022), US and Japan (21) | Socially facilitative robot | To identify middle-aged and older adults’ challenges with current technology-mediated social interactions, and then brainstorm socially-facilitative robot concepts to address their stated needs and wants. | Qualitative | 19 middle-aged or older adults (7 from US (age range 51–68; female 57.1%); 12 from Japan (age range 50–63; female 50%) | Design | Participatory design (Workshops) |
| Garcia-Mendez et al. (2021), Spain | Entertainment chatbot (EBER) | To report the experimental tests of EBER chatbot with real users | Quantitative non-RCT | 31 older adults (mean age 75.5, female 64.5%) | Implementation and evaluation | User-centered approach |
| Gasteiger et al. (2022), New Zealand (22) | Daily care service robot (Bomy) | To evaluate the usefulness and older adults’ perception and experience of using daily care robot through unrestricted and unsupervised use at home | Qualitative | 6 older adults (age range 72–83; female 66.7%) | Implementation | Participatory approach: Co-design (Interviews) |
| Gyrard et al. (2023), Europe and Japan (23) | Social companion robots | To explore how internet of robotic things technology and cocreation methodologies help to design emotional-based robotic application | Quantitative non-RCT | 100 older adults and caregivers | Design | Participatory Approach—codesign (Workshops) |
| Kim et al. (2022), South Korea (24) | Robot (Dori) | To design the care robot’s services based on sensing movement during daily activities | Mixed methods | 25 nurses and medical workers (female 80%); 22 nursing care workers (90.9% female); caregivers? | Evaluation | Human-centered AI framework (Survey) |
| Louie et al. (2014), Canada (25) | Socially assistive robot (Brian 2.1) | To investigate the acceptance and attitudes of older adults toward the human-like expressive socially assistive robot Brian 2.1 | Quantitative non-RCT | 46 older adults (age range 62–91; female 80.4%) | Evaluation | User-centered (Demonstration) |
| Mahmoudi Asl et al. (2023), Netherlands and Spain | Social robot (MINI) | To investigate the attitude of stakeholders and potential facilitators and the barriers to implementing the social robot MINI in community-based meeting centers for people with dementia and carers | Qualitative | 12 people with dementia and 11 stakeholders | Design | User-centered (Demonstration) |
| Martin-Hammond et al. (2019), USA (26) | Intelligent assistant | To understand older adults’ perspectives regarding intelligent assistants for health information management and search | Qualitative | 18 older adults (age range 61–93; 83.3% female) | Design | Participatory approach: (Multiphases) |
| Murawski et al. (2024), USA (27) | Online AI-based caregiver negotiation program (NegotiAge) | To develop and pilot test an artificial-intelligence negotiation training program, NegotiAge, for family caregivers. | Mixed methods | 12 family caregivers of older adults who have cognitive loss | Design and Evaluation | Usability study/ participatory approach (Multiphases) |
| Muuraiskangas et al. (2012), Germany (28) | Virtual Coach (V2me) | To examine user experience and acceptance of a prototype of the virtual coach (V2me) system for deriving the requirements for a system with social networking and friendship functionality | Qualitative | Face-to-face interviews: 30 older adults (age range 69–90; female 63.3%) One-on-one usability testing: 6 participants (mean age 78.8, female 50%) Workshop: 7 participants (mean age 80.6, female 100%) | Design, implementation, evaluation | User-centered design (Interviews + Workshops) |
| Petersen et al. (2020), USA (29) | Mobile app for monitoring the use of a bluetooth-connected resistance exercise band app. | To develop a mobile app with both older adults and clinicians while exploring whether data collected through this process can be used in NLP and sentiment analysis | Mixed methods | 6 clinicians (age range 37–51; male 33.3%) and 16 older adults (age range 66–85; male 18.8%) | Evaluation | User-centered design (Multiphases) |
| Tiersen et al. (2021), UK (30) | Smart home | To investigate the functional, psychosocial, and environmental needs of people living with dementia, their caregivers, clinicians, and health and social care service providers regarding smart home systems. | Qualitative | (1) Interview participants (3 rounds) –9 people with dementia, 9 caregivers –10 academic and clinical staff –10 caregivers and 2 people with dementia (2) Focus group participants (2 rounds) – 2 managers – 6 stakeholders (3) Workshop participants (3 rounds) –12 health and social care clinicians –35 pairs of people with dementia and caregivers –24 participants (14 occupational therapists; 4 National Health Service pathway directors; 6 researchers in occupational therapy, neuropsychiatry, and engineering) | Design | Participatory design (Multiphase: Interviews, Focus group, Workshops, Ethnographic observation) |
Notes:
a=country in which the study was conducted. AI = artificial intelligence, COPD = chronic obstructive pulmonary disease. NLP = natural language processing. RCT = randomized control trial. (MMAT Quailty: +++ strong, ++medium, + low).
Summary of Studies Included in Review
| Authors (Year), Countrya | Type of AI | Aim of Study | Study Design | Sample / Stakeholders Characteristics | Stage of Development | Engagement Approach (Method) |
|---|---|---|---|---|---|---|
| Cantone et al., (2023), Italy (16) | Humanoid robot | To develop a comprehensive architecture that integrates robots, sensors, and AI, to provide care for elderly individuals at home | Qualitative | 10 doctors with experience in geriatrics and internal medicine (age range 35–68; female 40%), 100% Italian | Design | User-centered design (Interviews) |
| Cinini et al. (2021), Italy (17) | Caregiver robot | (1) to investigate older adults’ perception of acceptable technology and their acceptability of caregiver robots (2) To monitor linguistic behavior and spoken language productions of elderly subjects over the time (3) To automate the training and the monitoring of motor and cognitive functions of older adults | Quantitative non-RCT | (1) Planning stage: 100 older adults (age range 65–81; 64% male) for first experiment, 202 older adults (age range 65–87; 41% male) living independently for second experiment (2) Wellbeing assessment: 58 older adults (mean age 80; female 74%) (3) Motor and cognitive function assessment: 16 older adults (age range 65–78; female 56.3%) 100% Italian | Design and Implementation | Active user engagement / User-centered (Survey) |
| D’Onofrio et al. (2019), Italy and Japan (18) | Companion robot (Buddy) | To provide a pilot qualitative analysis of the needs of older adults and their caregivers when given conversational activities with robots | Qualitative | 17 older adults (10 from Italy (age range 65–81, male 70%); 7 from Japan (age range 63–92, male 14.3%)) 36 caregivers (30 from Italy (age range 33–68, male 43.3%; 6 from Japan (age unknown, male 66.6%)) 30 Italian, 6 Japanese | Design | Cocreation methodology participatory approach (Workshops) |
| Easton et al. (2019), UK (19) | Virtual agent system | To codesign the content, functionality, and interface modalities of an autonomous virtual agent to support self-management for patients with COPD and then to assess the acceptability and system content | Qualitative | Workshop #1: 5 individuals with COPD and 1caregiver (age range 69-86; female 60%) Workshop #2: 4 individuals with COPD (age range 66–80; female 50%) | Design | Participatory approach (Workshops) |
| Eun et al. (2023), South Korea (20) | Exercise game | To design a digital healthcare application aiming at improving the older adults’ physical and cognitive functions | Quantitative non-RCT | 25 older adults (mean age range 68–80) 100% Korean | Implementation | User-centered |
| Fraune et al. (2022), US and Japan (21) | Socially facilitative robot | To identify middle-aged and older adults’ challenges with current technology-mediated social interactions, and then brainstorm socially-facilitative robot concepts to address their stated needs and wants. | Qualitative | 19 middle-aged or older adults (7 from US (age range 51–68; female 57.1%); 12 from Japan (age range 50–63; female 50%) | Design | Participatory design (Workshops) |
| Garcia-Mendez et al. (2021), Spain | Entertainment chatbot (EBER) | To report the experimental tests of EBER chatbot with real users | Quantitative non-RCT | 31 older adults (mean age 75.5, female 64.5%) | Implementation and evaluation | User-centered approach |
| Gasteiger et al. (2022), New Zealand (22) | Daily care service robot (Bomy) | To evaluate the usefulness and older adults’ perception and experience of using daily care robot through unrestricted and unsupervised use at home | Qualitative | 6 older adults (age range 72–83; female 66.7%) | Implementation | Participatory approach: Co-design (Interviews) |
| Gyrard et al. (2023), Europe and Japan (23) | Social companion robots | To explore how internet of robotic things technology and cocreation methodologies help to design emotional-based robotic application | Quantitative non-RCT | 100 older adults and caregivers | Design | Participatory Approach—codesign (Workshops) |
| Kim et al. (2022), South Korea (24) | Robot (Dori) | To design the care robot’s services based on sensing movement during daily activities | Mixed methods | 25 nurses and medical workers (female 80%); 22 nursing care workers (90.9% female); caregivers? | Evaluation | Human-centered AI framework (Survey) |
| Louie et al. (2014), Canada (25) | Socially assistive robot (Brian 2.1) | To investigate the acceptance and attitudes of older adults toward the human-like expressive socially assistive robot Brian 2.1 | Quantitative non-RCT | 46 older adults (age range 62–91; female 80.4%) | Evaluation | User-centered (Demonstration) |
| Mahmoudi Asl et al. (2023), Netherlands and Spain | Social robot (MINI) | To investigate the attitude of stakeholders and potential facilitators and the barriers to implementing the social robot MINI in community-based meeting centers for people with dementia and carers | Qualitative | 12 people with dementia and 11 stakeholders | Design | User-centered (Demonstration) |
| Martin-Hammond et al. (2019), USA (26) | Intelligent assistant | To understand older adults’ perspectives regarding intelligent assistants for health information management and search | Qualitative | 18 older adults (age range 61–93; 83.3% female) | Design | Participatory approach: (Multiphases) |
| Murawski et al. (2024), USA (27) | Online AI-based caregiver negotiation program (NegotiAge) | To develop and pilot test an artificial-intelligence negotiation training program, NegotiAge, for family caregivers. | Mixed methods | 12 family caregivers of older adults who have cognitive loss | Design and Evaluation | Usability study/ participatory approach (Multiphases) |
| Muuraiskangas et al. (2012), Germany (28) | Virtual Coach (V2me) | To examine user experience and acceptance of a prototype of the virtual coach (V2me) system for deriving the requirements for a system with social networking and friendship functionality | Qualitative | Face-to-face interviews: 30 older adults (age range 69–90; female 63.3%) One-on-one usability testing: 6 participants (mean age 78.8, female 50%) Workshop: 7 participants (mean age 80.6, female 100%) | Design, implementation, evaluation | User-centered design (Interviews + Workshops) |
| Petersen et al. (2020), USA (29) | Mobile app for monitoring the use of a bluetooth-connected resistance exercise band app. | To develop a mobile app with both older adults and clinicians while exploring whether data collected through this process can be used in NLP and sentiment analysis | Mixed methods | 6 clinicians (age range 37–51; male 33.3%) and 16 older adults (age range 66–85; male 18.8%) | Evaluation | User-centered design (Multiphases) |
| Tiersen et al. (2021), UK (30) | Smart home | To investigate the functional, psychosocial, and environmental needs of people living with dementia, their caregivers, clinicians, and health and social care service providers regarding smart home systems. | Qualitative | (1) Interview participants (3 rounds) –9 people with dementia, 9 caregivers –10 academic and clinical staff –10 caregivers and 2 people with dementia (2) Focus group participants (2 rounds) – 2 managers – 6 stakeholders (3) Workshop participants (3 rounds) –12 health and social care clinicians –35 pairs of people with dementia and caregivers –24 participants (14 occupational therapists; 4 National Health Service pathway directors; 6 researchers in occupational therapy, neuropsychiatry, and engineering) | Design | Participatory design (Multiphase: Interviews, Focus group, Workshops, Ethnographic observation) |
| Authors (Year), Countrya | Type of AI | Aim of Study | Study Design | Sample / Stakeholders Characteristics | Stage of Development | Engagement Approach (Method) |
|---|---|---|---|---|---|---|
| Cantone et al., (2023), Italy (16) | Humanoid robot | To develop a comprehensive architecture that integrates robots, sensors, and AI, to provide care for elderly individuals at home | Qualitative | 10 doctors with experience in geriatrics and internal medicine (age range 35–68; female 40%), 100% Italian | Design | User-centered design (Interviews) |
| Cinini et al. (2021), Italy (17) | Caregiver robot | (1) to investigate older adults’ perception of acceptable technology and their acceptability of caregiver robots (2) To monitor linguistic behavior and spoken language productions of elderly subjects over the time (3) To automate the training and the monitoring of motor and cognitive functions of older adults | Quantitative non-RCT | (1) Planning stage: 100 older adults (age range 65–81; 64% male) for first experiment, 202 older adults (age range 65–87; 41% male) living independently for second experiment (2) Wellbeing assessment: 58 older adults (mean age 80; female 74%) (3) Motor and cognitive function assessment: 16 older adults (age range 65–78; female 56.3%) 100% Italian | Design and Implementation | Active user engagement / User-centered (Survey) |
| D’Onofrio et al. (2019), Italy and Japan (18) | Companion robot (Buddy) | To provide a pilot qualitative analysis of the needs of older adults and their caregivers when given conversational activities with robots | Qualitative | 17 older adults (10 from Italy (age range 65–81, male 70%); 7 from Japan (age range 63–92, male 14.3%)) 36 caregivers (30 from Italy (age range 33–68, male 43.3%; 6 from Japan (age unknown, male 66.6%)) 30 Italian, 6 Japanese | Design | Cocreation methodology participatory approach (Workshops) |
| Easton et al. (2019), UK (19) | Virtual agent system | To codesign the content, functionality, and interface modalities of an autonomous virtual agent to support self-management for patients with COPD and then to assess the acceptability and system content | Qualitative | Workshop #1: 5 individuals with COPD and 1caregiver (age range 69-86; female 60%) Workshop #2: 4 individuals with COPD (age range 66–80; female 50%) | Design | Participatory approach (Workshops) |
| Eun et al. (2023), South Korea (20) | Exercise game | To design a digital healthcare application aiming at improving the older adults’ physical and cognitive functions | Quantitative non-RCT | 25 older adults (mean age range 68–80) 100% Korean | Implementation | User-centered |
| Fraune et al. (2022), US and Japan (21) | Socially facilitative robot | To identify middle-aged and older adults’ challenges with current technology-mediated social interactions, and then brainstorm socially-facilitative robot concepts to address their stated needs and wants. | Qualitative | 19 middle-aged or older adults (7 from US (age range 51–68; female 57.1%); 12 from Japan (age range 50–63; female 50%) | Design | Participatory design (Workshops) |
| Garcia-Mendez et al. (2021), Spain | Entertainment chatbot (EBER) | To report the experimental tests of EBER chatbot with real users | Quantitative non-RCT | 31 older adults (mean age 75.5, female 64.5%) | Implementation and evaluation | User-centered approach |
| Gasteiger et al. (2022), New Zealand (22) | Daily care service robot (Bomy) | To evaluate the usefulness and older adults’ perception and experience of using daily care robot through unrestricted and unsupervised use at home | Qualitative | 6 older adults (age range 72–83; female 66.7%) | Implementation | Participatory approach: Co-design (Interviews) |
| Gyrard et al. (2023), Europe and Japan (23) | Social companion robots | To explore how internet of robotic things technology and cocreation methodologies help to design emotional-based robotic application | Quantitative non-RCT | 100 older adults and caregivers | Design | Participatory Approach—codesign (Workshops) |
| Kim et al. (2022), South Korea (24) | Robot (Dori) | To design the care robot’s services based on sensing movement during daily activities | Mixed methods | 25 nurses and medical workers (female 80%); 22 nursing care workers (90.9% female); caregivers? | Evaluation | Human-centered AI framework (Survey) |
| Louie et al. (2014), Canada (25) | Socially assistive robot (Brian 2.1) | To investigate the acceptance and attitudes of older adults toward the human-like expressive socially assistive robot Brian 2.1 | Quantitative non-RCT | 46 older adults (age range 62–91; female 80.4%) | Evaluation | User-centered (Demonstration) |
| Mahmoudi Asl et al. (2023), Netherlands and Spain | Social robot (MINI) | To investigate the attitude of stakeholders and potential facilitators and the barriers to implementing the social robot MINI in community-based meeting centers for people with dementia and carers | Qualitative | 12 people with dementia and 11 stakeholders | Design | User-centered (Demonstration) |
| Martin-Hammond et al. (2019), USA (26) | Intelligent assistant | To understand older adults’ perspectives regarding intelligent assistants for health information management and search | Qualitative | 18 older adults (age range 61–93; 83.3% female) | Design | Participatory approach: (Multiphases) |
| Murawski et al. (2024), USA (27) | Online AI-based caregiver negotiation program (NegotiAge) | To develop and pilot test an artificial-intelligence negotiation training program, NegotiAge, for family caregivers. | Mixed methods | 12 family caregivers of older adults who have cognitive loss | Design and Evaluation | Usability study/ participatory approach (Multiphases) |
| Muuraiskangas et al. (2012), Germany (28) | Virtual Coach (V2me) | To examine user experience and acceptance of a prototype of the virtual coach (V2me) system for deriving the requirements for a system with social networking and friendship functionality | Qualitative | Face-to-face interviews: 30 older adults (age range 69–90; female 63.3%) One-on-one usability testing: 6 participants (mean age 78.8, female 50%) Workshop: 7 participants (mean age 80.6, female 100%) | Design, implementation, evaluation | User-centered design (Interviews + Workshops) |
| Petersen et al. (2020), USA (29) | Mobile app for monitoring the use of a bluetooth-connected resistance exercise band app. | To develop a mobile app with both older adults and clinicians while exploring whether data collected through this process can be used in NLP and sentiment analysis | Mixed methods | 6 clinicians (age range 37–51; male 33.3%) and 16 older adults (age range 66–85; male 18.8%) | Evaluation | User-centered design (Multiphases) |
| Tiersen et al. (2021), UK (30) | Smart home | To investigate the functional, psychosocial, and environmental needs of people living with dementia, their caregivers, clinicians, and health and social care service providers regarding smart home systems. | Qualitative | (1) Interview participants (3 rounds) –9 people with dementia, 9 caregivers –10 academic and clinical staff –10 caregivers and 2 people with dementia (2) Focus group participants (2 rounds) – 2 managers – 6 stakeholders (3) Workshop participants (3 rounds) –12 health and social care clinicians –35 pairs of people with dementia and caregivers –24 participants (14 occupational therapists; 4 National Health Service pathway directors; 6 researchers in occupational therapy, neuropsychiatry, and engineering) | Design | Participatory design (Multiphase: Interviews, Focus group, Workshops, Ethnographic observation) |
Notes:
a=country in which the study was conducted. AI = artificial intelligence, COPD = chronic obstructive pulmonary disease. NLP = natural language processing. RCT = randomized control trial. (MMAT Quailty: +++ strong, ++medium, + low).
Characteristics of the Samples
Studies recruited participants from community settings (70.6%, n = 12; 18, 20, 21, 22, 24, 25, 26, 27, 29, 31, 33, 34) clinical/medical settings (5.9%; n = 1; 19) or both (23.5%; n = 4; 23, 28, 30, 32). Sample sizes across studies ranged from 6 to 202 participants (21.33). Regarding participant demographics, gender distribution was reported in about half of the studies (16,19,21,22,24,26,27,29,32), where females were generally overrepresented, comprising 60%–80% of participants. The remaining studies did not explicitly report gender distribution.
Types of AI / Technology
Nine out of 17 articles focused on robots (16–18,21–25,31). More specifically, roughly half of these robots aimed to address the social needs of older adults and provide companionship (18,21,23,25,31). Four robots were developed to monitor various aspects of older adults’ daily activities and overall functional status. For instance, Kim et al. monitored older adults’ movement during daily activities (24), whereas Cinini et al. assessed linguistic behaviors along with motor and cognitive functions (17). Additionally, Cantone et al. developed a humanoid robot that assessed older adults’ emotional state and measured vital signs (16). Gasteiger et al. developed a robot designed to remind older adults of their daily activities and engage them in games that stimulate their cognitive abilities (22).
A total of 5 out of 17 articles focusing on virtual assistants or chatbots (19,26–28,32). Most of these virtual assistants/chatbots were developed for educational purposes. For example, Murawski et al. designed computer agents that teach negotiation skills to caregivers when encountering conflicts with the care-recipient (27), whereas Martin-Hammond et al. focused on intelligent assistants in the context of helping older adults manage and search health information at home (26). Furthermore, Easton et al. created Avachat, which is a virtual agent designed to support individuals with Chronic Obstructive Pulmonary Diesease (COPD), and mental health issues through self-management principles. Avachat utilizes advanced natural language processing (NLP) algorithms for seamless communication, whereas its persona, Ava, is designed to foster trust and engagement (19).
Other objectives included developing a virtual coach-assisted system addressing older adults’ social networking and friendship functionality (28) and a chatbot providing entertainment (32). Other AI systems developed were mobile apps for monitoring older adults’ use of an exercise band (29), AI-based personalized exercise game (20), and smart homes (30). Among the two articles that aimed to support older adults with ADRD, the AI features included social robots and remote monitoring devices for living environments that support the users’ safety (30,31).
Stages of Development
The stages of development (ie, design, implementation, and evaluation) of the AI system were determined based on the phase when older adults or other stakeholders were involved in the study. Articles that involved older adults in multiple stages were categorized separately.
Design (n = 8)
Half of the studies (8 out of 17 studies) involved older adults or stakeholders during the design process (16,18,19,21,23,26,30,31). The needs of older adults, related to the AI system’s purpose, were commonly assessed, including socialization needs (18,21), self-management challenges (19), and the functional and psychosocial needs of people living with dementia (30). Older adults’ current use and perception towards the AI system, as well as their needs pertaining to the technology, were also commonly explored throughout this phase (18,21,26,30). Mahmoudi et al. specifically aimed to identify the barriers and facilitators to deploying social robots to people with dementia and their caregivers (31). Feedback and insights regarding codesign and system refinement were also requested during this phase (16,23).
Implementation (n = 2)
Two articles involved older adults in the implementation stage (20,22). More specifically, Eun et al. conducted empirical monitoring of exercise to analyze older adults’ task performance on their AI-based exercise game (20). The study conducted by Gasteiger et al. examined the feasibility of their daily care service robot, considering the environment in which the robot will be used and relevant factors that may influence its implementation (22).
Evaluation (n = 3)
Three articles engaged older adults during the evaluation stage, each focusing on different aspects without consistency (24,25,29). Petersen et al. focused on receiving clinicians and older adults’ feedback about the design and functionality of the mobile app (29), whereas Louie et al. assessed older adults’ attitudes and acceptance towards the robots’ human-like assistance after conducting a live robot demonstration (25). Kim et al. presented the care robots’ services to the participants through photos and videos, asking them to choose desired services and respond to ethical dilemmas concerning robots’ monitoring and management via mobile phones (24).
Multistage (n = 4)
A total of 4 articles involved older adults and related stakeholders across multiple stages (17,27,28,32). Cinini et al. engaged with older adults during the design and implementation stage, assessing their acceptability of caregiver robots and later their wellbeing and motor-cognitive functions through real-world experiments (17). Murawski and et al. involved stakeholders in both design and evaluation, mainly focusing on pilot-testing the AI-based negotiation training program, while also having collaborated with family caregivers and geriatricians on the program development beforehand (27). Garcia-Mendez et al. had older adults interact with the chatbot, assessing their emotions during this implementation stage, and their satisfaction, amazement, and chatbot-human likeliness postexperiment (32). Muuraiskangas et al. engaged older adults throughout the design, implementation, and evaluation stages via four iterative processes: analyzing their communication technology needs, integrating these requirements into the design, prototyping, and evaluating the system with older adults (28).
Methods of Engagement
The methods of engagement were identified based on how the authors described the involvement of stakeholders (older adults) in their studies. Multiple methods were listed and then grouped based on different approaches. For example, the methods included interviews, workshops based on participatory design and cocreation/codesign, demonstration sessions, surveys, and several phases that involved different approaches.
User-Centered Approach (n = 8)
A total of 8 studies adopted a user-centered design approach, actively involving a variety of stakeholders in the technological design process (16,17,20,25,28,29,31,32). This approach aims to bridge the digital divide for older adults, enhancing their access to digital content (32), and to foster an environment that encourages collaboration, sharing, and interaction among users (20).
Interviews and surveys are the most prevalent methods within the user-centered approach. Specifically, the System Usability Scale (SUS) and the Usefulness, Satisfaction, and Ease of Use (USE) questionnaires were utilized in one round of study (29), whereas the Post-Study System Usability Questionnaire (PSSUQ) was implemented in another study (17). These tools are critical for gathering user feedback on the usability and satisfaction with the technology, ensuring that the end products align with user needs and expectations. Cinini et al. utilized multiple phases that prioritized active user engagement from the planning stage to assess technology acceptability (17). Participants engaged in monitoring language changes through natural language processing for well-being assessments. Cognitive and motor functions were assessed using automated systems that delivered tests and exergames through user-friendly interfaces.
Participatory Design Approach (n = 8)
Eight out of 17 studies employed a participatory design approach (18,19,21–23,26,27,30). Participatory design in technology development emphasizes involving everyone, not just viewing older adults and caregivers as subjects or data sources (8). Codesign is a team effort where older adults work with designers or researchers to create solutions, leading to user friendly and innovative products. For example, a study by Easton et al. quickly produced a protype for an autonomous agent to help older adults with COPD manage their symptoms, resulting in a design that matches their needs (19). Similarly, cocreation broadly involves any act of collective creativity, which was employed by D’Onofrio et al. to develop companion robots for older adults (18). Gyrard et al. specifically integrated codesign, involving multidisciplinary partners as active participants in development, and cocreation, generating tools and techniques based on explicit needs and observations of user behavior (23).
Workshops and interviews are the most common methods in participatory design. Typically, participatory design workshops are structured into three main thematic sessions: an initial discussion on the types of technology that participants use, a review session, and a concluding design session (19,21,23). The format for interviews usually involves open-ended questions, often probing users’ satisfaction (27), as well as identifying challenges and needs related to daily activities. This approach also helps gauge initial reactions to a robot’s appeal and its key features (18). Tiersen et al. used a mixed methods approach including semistructured interviews, focus groups, workshops, and ethnographic observation to understand the needs of patients, caregivers, and clinicians, leading to the creation of patient-centered interventions. The research also investigated potential innovations in care technologies and public health strategies by engaging stakeholders (30).
Human centered design (n = 1)
Kim et al. used the Human-Centered Artificial Intelligence (HCAI) framework to guide the development of the care robot Dori (24). The framework prioritizes dignity, autonomy, controllability, and privacy for older adults. The study used a human-centered design approach by selecting caregivers, nurses, and clinicians as the target audience and conducting focus group interviews with 47 participants. Their feedback on the robot’s appearance, materials, and services informed its development, emphasizing dignity, control, and autonomy. Preferences and ethical judgments on medication management and sensing processes were also incorporated into the robot’s design.
Challenges or Barriers in Engaging Stakeholders
Digital literacy/access issues (n = 4)
Four studies (21,22,26,31) reported challenges in engaging stakeholders due to a lack of digital literacy and technical issues such as connectivity and usability problems (22). Meanwhile, Martin-Hammond et al. noted that some participants did not have internet access (26). Fraune et al. found that the participants had to have access to the internet and exhibited some degree of technical competency, as they were able to participate in a group online video call; however, this prevented them from studying populations of older adults who lack either perceived technical competency or access to the internet altogether (21). Another study by Mahmoudi et al. outlined several challenges in introducing new technology to persons with dementia (PwD), highlighting a generational divide with some PwD perceiving these technologies as more suited for younger individuals or those in early stages of dementia (31). A significant barrier was the lack of familiarity with technology, as many participants were inexperienced with basic devices like smartphones.
Individual health problems (n = 2)
Two studies highlight the participants’ health issues (30,32), such as hearing loss (32) and cognitive impairment (30). One major concern is the diverse and individualized needs of older adult users, particularly those with cognitive impairments. For instance, 8 people with moderate and advanced dementia were excluded to ensure participants could directly express their experiences (30). Designing a system that accommodates the broad spectrum of functional, psychosocial, and environmental needs can be challenging, as each persons’ health condition and technological comfort levels vary widely.
Other challenges (n = 1)
Fraune et al. observed that participants did not discuss or build upon each other’s ideas as much as expected (21). Using participatory design methods, this study explored how robots could serve as social facilitators for adults aged 50+ in the United States and Japan. Participatory design workshops revealed that U.S. participants viewed robots as tools to facilitate social connections, whereas Japanese participants saw them as surrogate companions to combat loneliness. They found that U.S. participants spoke more than the Japanese participants, which could be attributed to a number of factors including the participants’ lack of confidence, insufficient rapport among participatory design groups, or unintended changes in facilitation style when culturally adapted for use in Japan.
Ethical Considerations
Ethical considerations pertaining to AI systems involving older adults or stakeholders were not discussed in the current literature in depth. Most commonly reported concerns were related to continuous monitoring of the older adults and privacy issues (21,24,30,31). Muuraiskangas et al. also pointed out the ethical concerns around establishing a long-term socioemotional relationship between a virtual coach and end users, questioning whether AI systems could replace humanistic contact and its associated impact (28).
Outcomes
We reviewed studies to determine if they addressed cost to users (monetary or otherwise), access issues/health disparities, whether training was required to utilize the AI, and if the positive impact of the AI was being reported. Only three studies address possible associated cost to user by noting financial resources are needed for many of the AI methods outlined, this including both monetary means to purchase a device and an internet connection to utilize a device (26,30,31). Four studies commented on factors that encompassed access issues/health disparities primarily mentioning participants characteristics including access to internet, technical proficiency, experience with using devices and high cost in obtaining and maintaining different devices (21,26,30,31). Tiersen et al. explicitly acknowledged limitation in their recruitment strategy and the lack of representation of individuals with limited experience with technology (30). Four studies considered end-user training, with two studies having a training phase in their procedures (20,29), and the other 2 studies highlighting the importance of training the facilitators of the AI devices/intervention(s) and ensuring the devices have simple and easy to use features to minimize training barriers (26,31). Eleven articles mentioned the success and positive impact of AI for improving overall health outcomes and the satisfaction and usefulness of AI devices (16,17,20,22,24,26–29,31,32).
In the studies exploring technology design for older adults, user-centered design approaches enhanced digital accessibility and satisfaction by incorporating extensive user feedback through tools like the SUS and PSSUQ, effectively tailoring products to user needs (17,20,29,32). All four studies collectively highlighted the potential and challenges of integrating digital technologies in healthcare. Eun et al. reported the effectiveness of a serious game in enhancing cognitive and physical abilities among older adults, noting substantial improvements in game performance and task success rates over a three-month period (20). Petersen and colleagues employed the system usability scale (SUS) and the usefulness, satisfaction, and ease of use (USE) questionnaires in the initial rounds, and subsequent rounds of feedback via think-aloud and verbal prompting methods, revealed critical insights into the usability and practical application of a prototype exercise app (29). The iterative feedback and adjustments based on clinician and patient interactions suggest a methodical approach to refining digital health applications to better meet user needs.
Participatory design approaches fostered innovation and strong stakeholder relationships through codesign sessions, resulting in personalized and user-friendly solutions like companion robots and chronic condition management tools (18,19,23). The human-centered design approach, specifically through the development of the care robot Dori, aligned technology with core human values such as dignity and autonomy, ensuring ethical use of technology in sensitive applications (24). However, challenges such as unfamiliarity with digital tools, functional limitations, and cultural differences impacted stakeholder engagement and necessitated adaptive strategies to ensure inclusivity and effective participation (21,22,32).
Results of Quality Rating
Based on the MMAT, 9 studies were rated as high quality (18,21,22,24,26,27,29–31), 6 as moderate quality (16,17,19,20,25,28), and another 2 as low quality (23,32). Any discrepancies encountered during the assessment process were thoroughly discussed by the review team until consensus was achieved, ensuring a robust and comprehensive evaluation of the study quality.
Discussion and Implications
The review aimed to assess the involvement of older adults in the design, implementation, and evaluation of health-related AI technologies. It explored effective practices and strategies for engaging stakeholders and identified the challenges and barriers that arise in this process. The findings emphasize the importance of designing AI tools that meet the specific needs of older adults to ensure their accessibility and effectiveness. This review highlights the need to address barriers to enhance stakeholder participation, which is crucial for the successful adoption of AI in healthcare for older populations.
Many researchers and developers concentrated on assessing older adults’ perceptions and needs related to the AI systems, often tailoring the development to these insights. The primary methodologies employed included interviews and surveys for user-centered approaches, and workshops in conjunction with interviews for participatory designs. However, despite these efforts, the terms “codesign” and “participatory design” are frequently used interchangeably without standardized definitions, a finding also noted by Sumner et al. who emphasized the need for clearer definitions and standardized processes in stakeholder engagement (33).
Our review also revealed that AI technology such as robots and virtual assistants predominantly aimed at addressing social needs, daily activity monitoring, and educational support reflect a somewhat narrow view of older adults as isolated or in constant need of supervision and education. This perspective emphasizes a critical need for broader and more inclusive engagement strategies that encompass the diverse capabilities and aspirations of older adults.
A significant gap identified in the studies was the limited involvement of older adults with ADRD and their caregivers. Similar to Fischer et al. findings (11), our review revealed that older adults with ADRD were often excluded from recruitment due to potential communication barriers and regulatory challenges, despite the significant impact these technologies could have on their lives. This exclusion raises ethical concerns about the equitable development of AI technologies, an area also identified as a gap in the current literature. Discussions on ethics were mostly confined to privacy and user interaction, rather than the inclusive involvement of vulnerable populations.
This review has several implications for future research and practice in the field of AI development for healthcare, especially for older adults and their caregivers. Future studies should involve older adults throughout all phases of AI development—design, implementation, and evaluation—to provide deeper insights into how these technologies operate in real-world settings and how they are perceived by end users. Additionally, the lack of detailed reporting on how older adult feedback influenced final AI designs suggests that these engagements could be superficial. Therefore, it is crucial for future research to ensure that older adults are not merely consulted but are integral to the design process, with their feedback genuinely shaping the outcomes.
Limitations
This review has a few limitations. The term “Artificial Intelligence” is used inconsistently in literature due to its broad and even evolving definition, which often leads to the conflation of diverse technological aspects that may not constitute true AI. This inconsistency arises from the varying interpretations of what constitutes intelligence, and the capabilities required to be classified as AI. For instance, some literature equates AI with any advanced computational processes, such as machine learning or data analytics, which, while related, do not fully encompass the autonomous reasoning or learning capabilities central to AI. Furthermore, more recent recognition of the potential of AI exacerbates this issue, as products and systems are frequently labeled as AI to enhance their perceived innovation and appeal, regardless of whether they genuinely embody the principles of AI. In our study we did not question the use of the term if the authors chose to label their system as an AI application and often there was not sufficient information about the technical or computational aspects of the tools developed or evaluated to determine if the term was used consistently.
Similarly, the term codesign or participatory design is not used consistently in literature (34). To mitigate this, we used very broad terms and keywords but may still have excluded studies that did not label their work as participatory but still actively engaged end users in all phases of the work.
Our review focused only on studies published in English potentially excluding work that was disseminated in other languages. Finally, by excluding gray literature we may have failed to include work that aligns with our review focus and questions but was not disseminated in peer-reviewed academic outlets.
Conclusion
In sum, our review revealed the critical need for standardized methodologies in the engagement of older adults in the design of AI healthcare technologies. The review highlighted the importance of incorporating diverse and often underrepresented populations, such as those with ADRD, to ensure that AI solutions are both ethical and equitable. Moreover, the interchangeable use of terms like “codesign” and “participatory design” without clear definitions underscores the need for more precise and universally accepted guidelines in AI development practices. By addressing these gaps, future research can better tailor AI to meet the specific needs of older adults, enhancing their quality of life and healthcare outcomes.
Conflict of Interest
None.
Acknowledgments
This project is supported in part by the Penn Artificial Intelligence and Technology (PennAITech) Collaboratory for Healthy Aging funded by the National Institute on Aging Grant Nr. P30AG073105. Content and views expressed in this work are solely the responsibility of the authors and do not necessarily represent the official views of the NIA or NIH.
Author Contributions
H.C.: formal analysis, investigation, methodology (lead), writing- original draft (lead), review & editing (lead); O.O.: formal analysis, investigation, methodology, writing- original draft (supporting), review & editing (supporting); N.G.: formal analysis, investigation, project administration (lead), writing- original draft (supporting), review & editing (supporting); L.G.: data curation (lead), writing- review & editing (supporting); S.M.: data curation (lead), writing- review & editing (supporting); L.W.: conceptualization, funding acquisition, supervision, validation, writing- review & editing (supporting); G.D.: conceptualization (lead), funding acquisition (lead), supervision (lead), validation (lead), writing- review & editing (supporting)
