Abstract
Worldwide, the number of people who are over 60 is growing more than any other age group. In the scenario of this research, a mediumsized city in Brazil, elderlies with Chronic Noncommunicable Diseases can have home care services. However, the healthcare professionals who assist these patients regularly face challenges in the visits, such as persistent fear of violence in certain areas. This research proposes the use of Intelligent Conversational Agents as a tool for supporting Telecare practices and tame these challenges. Our literature review, however, has demonstrated that previous works did not follow a methodology to promote a co-design of these agents in a scenario such as ours. There is also a lack of works that balance technological and social issues with non-technological experts in the design process (conceptualization, dialogues elaboration, intelligent agent goals and evaluation). We, thus, have performed a study following a qualitative methodology with the participation of health professionals. The first step aimed to raise the context, dialogues, and objectives of the agent. The second step, using observational and interviewing in a controlled environment, aimed to determine and evaluate the design goals, usefulness, compatibility with the reality, and find out additional requirements to aid the development of an Intelligent Conversational Agent. Results have led us to propose the agent MarIANA (Maria Intelligent conversAtioNal Agent). Therefore, it will act to prevent the clinical complications of these older adults suffering from Chronic Noncommunicable Diseases pointed as a critical matter in a recent work conducted in the city of the scenario of this research.
The authors want to thank the Brazilian funding agencies that support this project in different ways: CAPES, CNPq and FAPERJ. They would also like to express their gratitude to the volunteers who participated in the study.
You have full access to this open access chapter, Download conference paper PDF
Similar content being viewed by others
Keywords
1 Introduction
One of the main challenges facing society is the aging of the population [38]. All over the world the number of people who are over 60 is growing more than any other age group [30]. Brazil, by 2025, will be the sixth country in number of elderlies worldwide [38]. This increase is mainly due to the rise in life expectancy and the decline in birth rates [1, 18]. Accompanied by these demographic changes, there is a need to ensure the elderly with positive and healthy aging [29, 38].
In our research scenario, a medium-sized city in Brazil, elderlies with some mobility difficulties, or elderlies who need a multidisciplinary health team can have home care service. In a majority of cases, due to socioeconomic aspects, the patients have an informal familiar caregiver, which means a person with an uninterrupted task, which, in general, suffers from physical and psychological burnout that indirectly affects the patient and the caregiver him/herself [17].
However, the homecare is negatively affected, because the healthcare professionals who assist the patients regularly face challenges in carrying out the visits, such as lack of time, work overload, and persistent fear of violence in certain areas [23]. In this sense, this research assumes the use of a Telecare assistance to prevent clinical complications with intelligent agents, where the caregiver may act as a bridge between the professionals and those older adults’ health [25], during the interval between the healthcare professionals visits.
According to Denecke, Tschanz, Dorner, and May [15], intelligent agents can provide care in places where people would not have or have limited access. However, few studies are focused on the challenges and benefits of Intelligent Conversational Agents (ICAs) usage for elderlies [26]. Besides that, none of them focuses on the caregiver. Our literature review has also demonstrated that previous works did not follow a methodology to design an ICA in a challenging scenario like ours, that is, high illiteracy rates, poverty as well as high rates of motor disability. There is also a lack of works that balance technological and social and cultural issues with non-technological experts (the health professionals) in the design process (conceptualization, dialogues elaboration, ICA goals, and evaluation).
This research, thus, aims at investigating the design of ICAs as a tool for supporting Telecare practices and tame these challenges. In this sense, a research question emerges: “Which communicative strategies should be adopted in an Intelligent Assistant to support caregivers of elderly patients in a Telecare assistance context?”
We, thus, have performed a study following a qualitative methodology [10, 16, 28, 40] with the participation of health professionals. The first step aimed to raise the context, dialogues, and objectives of the agent. The second step, using observational and interviewing in a controlled environment, aimed to determine and evaluate the design goals, usefulness, compatibility with the reality, and to find out additional requirements to aid the development of an ICA for our Telecare context.
Results have led us to propose the agent MarIANA (Maria Intelligent communicAtioN Agent). It will act to prevent the clinical complications of these older adults suffering from Chronic Noncommunicable Diseases (NCDs), pointed as a critical matter in a recent work conducted in the city of the scenario of this research. This paper has been divided as follows. In Sect. 2, we present our theoretical foundation. In Sect. 3, we present an overview of some related works. Section 4 presents our methodology and the results from the studies, including MarIANA. Finally, we discuss and conclude our results in Sect. 5.
2 Conversational Agents in Healthcare
According to Jusoh [22], a conversational agent is one of the emerging applications of natural language processing. According to Montenegro and Costa da Righi [26], the definition of conversational agents consists of a computer program or artificial intelligence capable of holding a conversation with humans through natural language processing.
Recently, applications of ICAs include emergent fields such as healthcare or even online business. A significant benefit of deploying an ICA is that it is available 24 h a day, seven days a week, and has a high potential to reduce the cost of hiring human personnel [22]. According to Denecke, Tschanz, Dorner, and May [15], agents can provide care in places where people would not have access. Possibilities of use in the medical field include, for example, a virtual nutritionist for diabetic patients, in which chatbot asks questions and ultimately prepares a diet, or an electronic medication management assistant or even music therapy.
Nonetheless, the authors Tsiourti et al. [34] affirm that real-world applications of conversational agents require in-depth exploration and a more scenario-based approach. Besides, it increases the chances of success in terms of acceptance, perceived usability, and usefulness. The authors King et al. [24] presented the efficiency of a culturally and linguistically adapted virtual agent. Results indicate a meaningful increase in physical activity relative of some of the older adults from the study.
In the work of Amato et al. [2], the authors investigate the effectiveness of a chatbot in a clinical context. In that scenario, the chatbot is used as a medical decision support system, having the goal of providing useful recommendations concerning to several disease prevention pathways. More in detail, the chatbot has been developed to help patients in choosing the most proper disease prevention pathway. The intention is to support the related prevention check-up and the final diagnosis. Preliminary experiments report the possibility of scalability of such system.
In the mental healthcare for elderlies scenario, there is a need to consider that the feeling of loneliness may come up when a patient lacks company or when they do not receive attention, which is common [21]. According to Eschweiler and Wanner [21], in the case of moderate dementia, patients tend to talk again and again about the same past events, which may lead to the adverse reaction of their conversation partners.
In this context, the role of an ICA could consist of dedicated affirmative responses that encourage the patients to continue their story [21]. Other applications for company could comprehend the support for locating objects, offering reminders, and guidance with household activities [34]. However, this implies the consideration of data privacy and data protection issues, both the patients themselves and the individuals possibly mentioned or commented upon in the conversation.
Another application for ICA in healthcare is the early identification of some diseases or the diagnosis of age-related diseases, affirms Eschweiler and Wanner [21]. As an example, Parkinson includes reduced perception and distinction of odors and disturbed sleep. Through the instruction of the caregivers or identification calls to the patient, some questions could be made to identify such issues. Still, according to Eschweiler and Wanner [21, pp. 8], the ICA could be programmed to apply geriatric assessment tools and questions of different aspects, such as Geriatric Depression Scale (GDS); Quality of life in Alzheimer’s disease; Mini-Mental State Examination (MMSE); ADL/IADL; the revised memory and behavior problem checklist (RMBPC); among others.
In a work presented by the authors Montenegro, da Costa, and Righi [26], five categories concerning to the challenges of using conversational agents emerged: dialogue generation, integration with other technologies, adaptation and use of conversational agents by the older adults, user experience, and new approaches to methodology and architecture. Still, according to the authors, all papers that sought to solve the challenges of ICA for the older adult [24, 27, 31, 33], focused on the patient scenario.
Among these works are three major fronts: prevention [24], training [27], and assistance with daily activities [33]. In the work of Nikitina, Callaioli, and Baez [27], the authors describe the requirements, models, and designs of inexperienced professional training for reminiscence, which is a reminder of important aspects of the life of older adults. Concerning prevention, ICAs can contribute in choosing the best path for treating the disease, preventing suicide, and even dealing with depression [26].
Regarding the initial and exploratory topic, which is the use of ICA with older adults, many applications, benefits, and challenges are upon ahead. Among the benefits, according to Jusoh [22], ICAs are available 24/7, can provide care in places where people would not have access, reduce the cost of hiring human personnel. About the applications, it is possible to mention, in the clinical context, to support physicians with the appropriate treatment path [2], mental healthcare regarding companionship [21], or even the support of locating objects, reminders, and guidance with daily activities [33, 34]. In nursing, the ICA application includes the diagnosis of age-related issues using geriatric assessment tools and questions of different aspects [21]. Other usages cover prevention [24], training of inexperienced healthcare professionals [27], preventing suicide, dealing with depression [26], and many others. As seen, there are many challenges regarding to older adults’ ICA application [24, 27, 31, 33]. Real-world applications require in-depth exploration, and, in some cases, a longitudinal study [34]. According to Tsiourti et al. [34], this increases the chance of success in terms of acceptance, perceived usability, and usefulness. Moreover, including the agent in culturally and linguistically adapted developing, indicates a meaningful increase in behavior change of its users [24].
3 Related Works
Many studies have addressed some challenges of ICA for older adults by promoting mental health, such as reminiscence [27], training for cognitive [20, 37], companions for the older adults who live in isolation [6, 36, 39], suicide prevention, how to deal with depression [26], and even palliative care [35]. Literature regarding agents for older adults with NCDs, in turn, is not extensive. Efforts are placed in the self-care promotion in the domain of endocrinology [7], cardiology [3, 41], or pneumology [19]. The lack of research of NCDs is, perhaps, related to a concentration of the efforts in first-world countries, such as the United States [4,5,6, 32], Ireland [3], France [37], United Kingdom [19] and Portugal [7, 34].
Moreover, similar to agents that focus on improving mental health (see [11]), agents for physical health should not address the communicative strategies as a secondary issue. For example, Tsiourti and colleagues [34] argue that the agent for isolated older adults should not be intrusive, should avoid proactivity when monitoring and be informal when remembering the daily activities. So, anticipating communicative strategies for an ICA to support caregivers of older adults with NCDs is not only related to improve the social abilities of agents but also to avoid risks.
Even in the same domain as Tsiourti and colleagues [34], in the study of Vardoulakis et al. [36], one of the participants commented she was feeling worse after the study because the interactions with the agent has made the participant realize the lack of communication with humans and friends that she desired in her life. In a longitudinal study with elderlies reported in Bott et al. [6], participants emphasized discontent with the repetitive nature of the agent and inappropriate dialogues - “I love you (the agent’s dialog) [...] it meant nothing to me, there was no reason why it would love me”. These reports highlight the need for prior co-creation and evaluation process with health professionals before the evaluation stage with the older adults. Zhang and colleagues [41] claim that previous interaction with professionals during the development phase allows a more assertive opinion and anticipates possible embedded risks.
Therefore, we can conclude that there is a lack of research about how to develop and ICA focusing on communicative strategies and attributes, especially for older adults with NCDs. Even more, the development of ICAs should consider socioeconomic and cultural factors of older adults. The complexity of variables from our research scenario urges for anticipating risks and avoiding more barriers to the older adults’ health.
4 Co-designing Intelligent Conversational Agents in Healthcare
4.1 Goal and Methodology
As already mentioned in the introduction, the goal of these studies was to investigate which communicative strategies should be adopted in an Intelligent Assistant for support caregivers of elderly patients in our scenario. This is the first step of a larger research project to create a framework to inform the design of such agents and to allow the design team to anticipate and deal with known problems in a specific context.
To this purpose, we have designed a qualitative study with two steps to investigate who the health professionals think the users are, the context and scenario details, the clinical expectations and conditions, the objectives to be achieved in Telecare assistance (engage, monitor, for instance). The option for a qualitative methodology derives from the fact that we want to learn in-depth and, in detail, the particularities of a specific scenario.
Given our goal, we had to choose how to characterize and frame our phenomenon of study. So, we chose to anchor our study in Semiotic Engineering (SE) theory [12,13,14], because it puts together the two themes of interest in our research, the communicability of ICAs and Human-Centered Computing (HCC) [13]. The procedures and analysis presented below have thus been carried out with constant reference to dimensions proposed by SE authors.
In the first step, we performed an exploratory study to discover context issues, dialogues and strategies that could provide prevention of clinical complications to the elderly in the context. The second step, using observational and interviewing in a controlled environment, aimed to determine and evaluate the perception of the agent design goals, usefulness, compatibility with the reality, and to find out additional requirements to aid the development of the ICA.
The recruitment process has selected academic and professional community, with no distinction of gender or age. The participants should be health professionals (nurses, doctors, psychologists, therapists, among others), preferably nurses, based on the level of technical/scientific/practical knowledge.
4.2 Step 1
Through the process of recruitment of health professionals who have experience in the home care setting, two health professionals were selected based on their previous experiences with older adults and knowledge about our research scenario. The first participant (P1) is a nursing student who is in the ninth semester. Her final essay project is related to monitoring postoperative seniors through a messaging application and she also has experience in a local project of Telecare. The second participant (P2) is a nurse, graduated since 2012, specialist in health care for the older adults who is currently a master student. Besides, the participant also has experience in working in several critical home care scenarios, as manager of a Telecare company.
Concerning the data collection, before the workshop, an initial survey was applied to select the participants, based on the level of technical/scientific/practical knowledge regarding the context of the research. This survey consisted basically in demographic information and a Likert scale for the subjects in which the study aims in the workshop. During the workshop, an empirical collection took place through audio, video, and artifacts from the Design Thinking stages.
Design Thinking has three steps (Immersion, Ideation, and Prototyping), in which we intend to:
-
Immersion - build a Persona with demographic information (name, age, education, technology skills, bio, frustrations, limitations) and two points from the Empathy Map (“Need to change” and “Works out”), i.e., improvements and solutions, techniques that are not working;
-
Ideation - the technique “How Might We” is applied. The participants must raise questions, for example, “How Might We - reduce the clinical complications in the scenario?” and seek answers for the resulting issues, regarding the ICA;
-
Conversational flow prototyping - co-create the dialog between the ICA and the Persona from the Immersion step. In general terms, the participants are thinking in creative solutions that could occur with a human-human interaction intermediated by the telephone or any communication source, thinking in solutions that meet the particular needs of the Persona. To do so, the participants should elaborate on the conversation, in a tree structure, with the materials (post its, styrofoam, pins, among others).
Data analysis of the transcribed info followed a non-predictive and interpretive paradigm with the coding approach of collected and transcribed empirical data [8]. The analysis started with an open coding naming sentence by sentence and an axial coding that used the most significant codings to organize, synthesize, integrate, and organize large amounts of data.
The main products of this step were a: Persona containing demographic information (e.g., name, age, among others), frustrations (e.g., lack of technical knowledge, informality, among others), limitations (e.g., social, economic, and cultural aspects), what the participants consider as pain points in the context and should be improved, and what the participants already use as solution for the pain points; “How Might We” answers, that is, possible solutions for the Persona, as well as the respective positive and negative impacts; and a “hands-on” conversational flow prototype of interactions between the agent and the Persona. Besides, with the analysis of the products of step 1 and the transcribed data, preliminary communicative strategies were elaborated.
To conclude, having the results from step 1, we developed on DialogflowFootnote 1 an initial interactive prototype. This version was deployed on a FacebookFootnote 2 page and had a programmed dialogue to verify and guide the caregiver of older adults with hypertension, as shown in Fig. 1.
4.3 Step 2
The second step, using observational and interviewing in a controlled environment, aimed to determine and evaluate the perception of the agent design goals, usefulness, compatibility with the reality, and to find out additional requirements to aid the development of the ICA.
There were three participants, and each interview took place at the participant’s preferred location. In general, participants presented a good knowledge concerning older adults, self-care, clinical complications, informal/formal caregivers, hypertension, and Telecare. The first participant (P1) is a nurse, a Ph.D. Candidate, and works in a government secretary of health, where our research is being carried out. The second participant (P2) is a professor and researcher in nursing and a doctor in cardiovascular sciences. Finally, the third participant (P3) is a nurse, postgraduate in intensive care, and has a master degree in nursing. Both second and third participants have worked for years on projects to assist the elderly with socioeconomic challenges.
Regarding data collection, before the interviews, pursuing the same goal as in Step 1, demographic information and the level of knowledge of the participants with the topics of our objective were collected. During the interviews, we collected the audio recorded, video recordings of the interaction with the initial prototype, and participants’ notes on the “hands-on” conversational flow prototype (printed in paper).
Hence, sessions of observational and interviews determined the perception of other specialists on the design goals, compatibility with the reality, and additional requirements to aid the development of the ICA. The research followed the semi structured steps below:
-
Present the research;
-
Ask the participant about more details related to the context and Persona;
-
Ask the participant to use the initial interactive prototype with the Thinkaloud protocol [9];
-
Show the participant the “hands-on” conversational flow (printed in paper) and ask for corrections, annotations or observations of any mistakes;
-
Ask the participant about the communicative strategies and attributes.
For the analysis step of the collected data, we combined discourse analysis [8] with semiotic analysis [12, 14], while examining records of what participants said, how they behaved in the workshop, kinds of comments they build, and so on. The semiotic analysis was conducted with the SigniFYIng Message tool [13]. In this tool, the following SE-based meaning categories are used to identify and describe: [13, pp.64]:
-
(i) Designers’ (in our context, the health professionals) beliefs about
-
(a) The logic of the user’s context?
-
(b) User’s needs;
-
(c) User’s preferences;
-
(d) User’s profile; and/or
-
(e) User’s goals;
-
-
(ii) the Developer’s (in our context, the health professionals) intentions and expectations with respect to
-
(a) The Logic of the system’s design?
-
(b) The System’s mode of use;
-
(c) The System’s description; and/or
-
(d) The System’s functionality;
-
-
(iii) the Developer’s (in our context, the health professionals) provisions and support for alternative modes/purposes of use that are compatible with the system’s design? of the system.
The main products of this step were: additional information (profile, context, user preferences, among others) for the Persona; critical paths which the health professionals anticipate in light of the scenario, on the “hands-on” conversational flow (through notes), and the interactive prototype (through interaction). To conclude, with the analysis of the products of step 2 and the transcribed data, additional communicative strategies were elaborated, and previous strategies from step 1 were refined.
4.4 Findings
Step 1. Regarding the first step (resulting artifacts in Fig. 2), it has showed that male elderlies between 70–75 years old are the most common profile that faces high hospitalization rates due to lack of self-care, causing clinical complications (injuries, mutilations, and even in extreme cases, death). This profile generally proves to be dependent on specific activities. Also, they do not properly maintain the treatment until the next service. Due to the neglect of their health, relatives such as their wives, aged about 60–65, end up taking care of their husbands for themselves. However, this can often happen without the relative having the proper technical preparation to be a caregiver.
Regarding the communicative aspects, the interaction human-human of the participants’ daily routine allowed the elicitation of communication models within the current Telecare context. Textual interaction enables professionals to receive short responses and with more agility. On the other hand, the use of voice interaction enables the reduction of subjectivity, allows illiterate older adults to interact with the application. One of the participants in the first step affirms that “The older adults [...] in applications such as Facebook or Whatsapp [...] guide themselves through figures and icons [...] and even with low literacy, they manage to send voice notes”.
Another result is the social skills of the agent: An incisive posture is noted in the dialogues with messages such as “Measure now”; Simpler dialogues denote the need for dialogues without subjectivity and reformulation of phrases. According to the report of one of the participants -“I saw an older adult who once said to me - Young man, do not go around. Speak exactly what I need to hear”; Repetition is the need to repeat the dialog. According to the participants, the negligence of actions happens intentionally or unintentional, “You can ask him to do it [...], but he may end up not remembering or not doing it at all. The ideal would be for you to get in touch again”.
As a result of the last stage from step 1, conversational flow prototyping, the participants developed 17 dialogues interaction between the agent and Persona, including questions as if the caregiver measured the older adult’s blood pressure. According to the immersion phase, our Persona’s husband does not apply the self-care.
Step 2. Regarding the (i) first category of the SigniFYIng Message, the discourse of the participants about (a) the context was related to dealing with actions promoted by educational institutions and movements of the public power about permanent education on the health of the elderly. Another part of the discourse (i.c) is associated with issues of the elderly and health professionals. It is interesting to notice the depth of the participants’ discussion when they stressed the use of digital meters and disconnection with part of the dialogues presented in the prototype (e.g., 120 \(\times \) 80). According to the participants, this would create a barrier. This barrier is necessarily related to the caregivers’ when they see a different sign in the dialogue other than they are used to in the digital blood pressure monitor (e.g., 12 \(\times \) 8), causing a communicational rupture.
Besides that, according to the participants, 13 \(\times \) 9 is not considered hypertension, even in a generic context without considering the health history of the older adult. In a scenario in which the older adult presented this pressure and was referred to the emergency room, it would lead to a denial by the professionals, indicating the absence of hypertension and possible disbelief with technology, consequently affecting the care process. Still, on breaks in the dialogues, the participants also stressed the importance of promoting physical exercise and not only changing eating habits.
Regarding the profile of the actors (i.d) in the scenario, the participants stressed the difficulty of changing the habit of the elderly, the need to engage in the care process and with technology, such as using festive dates in the dialogues, and finally, the real possibility of wrong medicalisation by caregivers. Finally, about the item, (i.e.), the emotional stress of the caregivers(s), as already stated above, affects not only the older adult but the wife caregiver who is often also elderly. This statement conducts the goal of technology to something more than just the health of the husband, but a relational agent in an indirect way to the caregiver.
About the (ii) second category of the SigniFYIng Message, the Developer’s intentions and expectations discourse was related to the design logic (ii.a), mode of use (ii.b), and functionality (ii.d). The sub-category (ii.a) refers to factors necessary for the interaction to occur correctly. Participants raised questions regarding strategies before and during use. Before use, promotion of interviews and training with caregivers were suggested by the participants.
Furthermore, during the use, for example, clarity in the way information should be entered by the user. Firstly, the use of simple agent dialogue for simple responses from the caregivers is evident, specially when considering that there is the possibility of answers in disconnected phrases. Limiting text input to predefined responses would be an obvious answer to the challenge, using, for example, buttons or checkboxes. However, according to one of the participants, the ideal would be to map tracks and common words. In that sense, it is continuously detecting words or phrases that may cause a warning and helping the health professionals in the decision-making process about the next medical appointment or acting immediately.
This alert to professionals is deeply associated with the “symbiotic” discourse of the participants in the second step. This discourse is a variation compared to the works of literature. Agent strategies are often very much associated with a context, when, in fact, the agent will be placed in a setting with several actors, such as the health professionals. Strategies for both the caregiver and the professionals allow a closer relationship and value the opinion of professionals about the technology, for example encouraging responsible nurses, from time to time, to analyze caregivers with good practices and make contact to congratulate them. According to one of the participants, the call makes much difference in the care of the elderly, because, on certain occasions, the call from the responsible nurse is the only call that the older adult receives over a long period.
Regarding the ways of Use (ii.b), the participants warned about the distance between the participants with long-term care and technology. The participants stressed how caregivers can fail to notice essential parts of the dialogue, needing something that promotes contrast or Use of visual aids. Besides, playful strategies are necessary to get him to use the technology for a longer time, thus avoiding the standardization of the care process. One of the participants suggested the Use of festive dialogues according to the date of the year. However, by the end of the study, the participant himself anticipated a possible barrier, saying that each patient has their religion (or not) and that not everyone celebrates specific dates.
Regarding the resulting functionalities and strategies (ii.d), new strategies point to improvements in the dialogues, promoting retention through the mentioned festive approach and personalized dialogues, promoting more assertive care through the mapping of the so-called target values and keywords. The discussion about the trade-off of the limitation or not of the user input enabled two points: limiting the target ranges when questioning the pressure at first. So, if the user responds with a critical option such as “Higher than 16 \(\times \) 9”, ask them how much higher. Thus, acknowledge if it is necessary to refer the older adult to the health center. In the case of target words, keep open and allow the agent to be continually listening to the user’s speeches, and if they enter a term that may be dangerous to the life or well-being of the elderly, notify the nurse linked to the elderly – finally, the Use of contrast in the parts that need the caregiver’s attention.
Also, call to congratulate when the caregiver is getting good results with the care – in this way, further strengthening the institutional bond between the elderly and caregiver with public health. Finally, personalizing the dialogues for the profile of the elderly is essential for the assertive care process, as each patient has his or her depth of context, culture, health, and beliefs.
Regarding the last item of the SignFYing Message, (iii) provisions and support for alternative modes/purposes, the participants’ discourse brought the possibility of using the agent with independent older adults – these, lonelier, also a profile with difficulties such as the caregiver profile, previously anticipated. Also, carelessness in these older adults’ care can cause a regression in the health condition, generating a further increase in the statistics of clinical complications. Specially because in some cases, they refuse the need of a caregiver, living alone without the help of a third party in self-care of their health.
MarIANA. The resulting agent MarIANA, it is essentially conceptualized to orient a caregiver of an older male adult with hypertension. Also, the agent promotes emotional support to the caregiver through valuing the care with pieces of dialog and by fostering the contact between health professionals and the caregiver. There are mainly two actors in the design space of MarIANA, the informal caregiver and the health professionals of the primary care. The informal caregiver, usually the wife of the patient, is susceptible to some conditions that can directly influence the care of the spouse. These conditions are low education, lack of technical knowledge (of the care procedures), pressure and lack of family support, among many others. On the other hand, professionals in primary care are vulnerable to external conditions of access to care, that is, difficulties in access and excessive activities and responsibilities.
For this reason, the benefits of the intelligent agent divide into promoting for professionals through MarIANA: the possibility of noting caregivers’ behavioral changes; receive requests for help from patients or caregivers; view in real-time or the records of the dialogues between agent and caregiver; promote the recognition of good practices; have inputs for the next visits; have general information about care for micro or macro-regions, and receive notifications that alert the patient is in risky situations. For the caregivers, it is: guide on some activities of the care, receiving monitoring by the agent for eventual emergencies; receive emotional support through dialogues and contacts from health professionals.
It is noteworthy that those benefits are derived from communication aspects. The first strategy (S1) is related to the dialog definition, which is determined by the objective (i.e., greeting, health orders, collection, guidance, reminders) of the message. That objective determines the (a) temporality (i.e., instant, daily, weekly), (b) means of transmitting (i.e., textual, videos, visual resources), and the (d) social ability (i.e., non-subjective, imperative, repetition, contextual language). For example, when the message is about health orders, the health professionals in the prototype instantiated an imperative posture - “Measure now!”.
The second strategy (S2) is related to personalizing the resulting hypertension dialog individually. In this sense, in order to make the dialog elements individual, some data are required from the caregiver (name, age, region) and from the older adult (name, age, health history, current medicines usage, family beliefs, interests, associated health professional).
The third strategy (S3) it is about clarity in the way the user should enter the information. Certain dialogues require different approaches since the “freedom” that a natural language conversation allows can be a barrier or a path for intervention. See the following example of the interaction between the agent and the caregiver:
-
(Agent) Did you measure John’s blood pressure today?
-
(Caregiver) yes
-
(Agent) How much was it? Select (or write) one of the options, Maria.
-
(1) Up to 12 \(\times \) 8
-
(2) Between 12 \(\times \) 8 and 14 \(\times \) 9
-
(3) Higher than 16 \(\times \) 9
-
-
(Caregiver) 3
-
(Agent) How much higher, Maria?
-
(Caregiver) 19
The first time that the agent asks for BP, then predefined BP is used. However, if the third option is selected, then the agent asks, “How much higher?”. This happens because a critical condition is relative to each individual. If this number is considered critical for the patient, then the patient is guided to the emergency. Since there is the possibility of “he goes to the hospital [...], and then the health professional himself will underestimate it [...] (the elderly person) leaves home [...] and (will have to) go back”. Not only putting the elderly, the caregiver, and the family in an alert situation but can also make the technology as untrustworthy as if it was placed by one of the participants.
The fourth strategy (S4) is related to another way to concede intervention. MarIANA will be capable of identifying words that can characterize a danger. For example, if, through audio or text, the caregiver sends the word “sleepy” then the health professional associated with the family should be warned. Besides, as in the case of the “critical words”, when the blood pressure is in dangerous levels, the team is also notified. One of the participants from step 1 said - “[...] maybe notify the team that he (older adult) was instructed to go to the hospital [...] because then at the end of the month they (health professionals) [...] could call to ask if he went to the hospital [...] and it is no longer the problem of the agent, it is the problem of the team that will take the necessary provisions”.
The fifth strategy (S5) is related to communication models. So that both the agent can process the information and the health professional make trusting decisions, the dialogues need to allow direct and straightforward interactions, agility in the interaction, detect sings and avoid questions that allow subjective answers.
Other strategies with great potential in care are the (S6) Engagement and (S7) Symbiosis. Regarding (S6), the participants argued the need to keep them engaged with technology and care:
-
a. Dialogues of interest - dialogues embedded during the flowchart, to comment on something of interest to the participant. For example festive dialogues, a Merry Christmas message.
-
b. Visual Resources - these visual resources depend on the means of transmitting. In the case of MarIANA, the use of emojis, figures, and even GIFs was recommended. Also, presenting contrast through stronger colors in alert messages, colorful or cheerful emojis, such as, for example, the second participant of the second step said, “Something that would show an OK here [...], a thumbs up”.
-
c. Dynamism - another concept inherited from human-human practice. Protocol dialogues like Mariana’s tend to become nauseating. Therefore, the use of (S5.a) or (S6.b) may be essential in this process of patient improvement, changing, for example, different visual resources, colors or even adding a dialog related to the caregiver interest.
Regarding (S7) Symbiosis, the participation of professionals in this context is essential. The cooperative participation between agent and scenario allows caregivers (and adjacent actors) to realize that it is not just a technology; there is a team behind it. This symbiosis occurs through:
-
Mapping of executions - how many times the agent needed to ask the caregiver to perform a certain task. For example, if it is needed many times, the agent should, in this case, contact the team. The Mapping of executions can also happen with a non-punitive intention. One example may be that, after a specified time, contact the health team. Then the associated professional get in touch to congratulate that the care is being adequate;
-
Mapping of target words and blood pressure values - once the agent maps this data, when it presents any risk, send an alert to the responsible professional or team.
5 Conclusions
The in-depth co-design approach of the ICA, with the participation of specialists, allowed the design of MarIANA. It is noteworthy that the use of agents in this scenario does not propose the replacement of human contact, essential for overcoming loneliness, cognitive, and emotional decline of the older adults [21]. In step 1, participants have cooperated by defining the profile and needs of the Persona, proposing solutions for the scenario, and prototyping a conversational flow for the hypertensive Persona’s husband. Additionally, challenging socioeconomic and cultural characteristics reflected directly in the intentions and expectations of the co-designers concerning the agent. In step 2, participants have brought additional information for the design space and additional communicative strategies; evaluated design decisions from step 1, such as the conversational flow; and also anticipated intricacies that were not compatible with the reality and needs of the health professionals.
Therefore, this research brings three main contributions. First, our research presented an initial exploration of a methodology for co-designing an ICA to such a socioeconomic and culturally challenging scenario. That may serve as a detailed approach to future investigation and possibly providing a starting point. Second, communicative aspects of a conversational agent, which we did not have the goal to offer definitive answers but plausible strategies and attributes to guide further development. Finally, our third contribution relates to the computational artifact for our research scenario, which may act as a tool for older adults with NCDs, specifically hypertension, in the scenario. Future work includes another co-design and evaluation cycle of the actual prototype and the inclusion of the caregivers in the design process. Also, more research is needed to reveal the potentials as well as the limitations of using such technology and the resulting behavior changes from such technology.
References
de Aguiar Sá Azeredo, Z., Neto Afonso, M.A.: Solitude from the perspective of the elderly. Revista Brasileira de Geriatria e Gerontologia 19(2), 313–324 (2016)
Amato, F., Marrone, S., Moscato, V., Piantadosi, G., Picariello, A., Sansone, C.: Chatbots meet eHealth: automatizing healthcare. In: WAIAH, pp. 40–49 (2017)
Barrett, M., et al.: Artificial intelligence supported patient self-care in chronic heart failure: a paradigm shift from reactive to predictive, preventive and personalised care. EPMA J. 10(4), 445–464 (2019). https://doi.org/10.1007/s13167-019-00188-9
Bickmore, T., Gruber, A., Picard, R.: Establishing the computer-patient working alliance in automated health behavior change interventions. Patient Educ. Couns. 59(1), 21–30 (2005)
Bickmore, T.W., et al.: Patient and consumer safety risks when using conversational assistants for medical information: an observational study of Siri, Alexa, and Google assistant. J. Med. Internet Res. 20(9), e11510 (2018)
Bott, N., et al.: A protocol-driven, bedside digital conversational agent to support nurse teams and mitigate risks of hospitalization in older adults: case control pre-post study. J. Med. Internet Res. 21(10), e13440 (2019)
Buinhas, S., et al.: Virtual assistant to improve self-care of older people with type 2 diabetes: first prototype. In: García-Alonso, J., Fonseca, C. (eds.) IWoG 2018. CCIS, vol. 1016, pp. 236–248. Springer, Cham (2019). https://doi.org/10.1007/978-3-030-16028-9_21
Charmaz, K.: Constructing Grounded Theory. SAGE, London (2014)
Charters, E.: The use of think-aloud methods in qualitative research an introduction to think-aloud methods. Brock Educ. J. Educ. Res. Pract. 12(2), 68–82 (2003)
Creswell, J.W.: Qualitative Inquiry and Research Design. SAGE Publications, Thousand Oaks (2007)
Cruz-Sandoval, D., Favela, J.: Incorporating conversational strategies in a social robot to interact with people with dementia. Dement. Geriatr. Cogn. Disord. 47(3), 140–148 (2019)
De Souza, C.S.: The Semiotic Engineering of Human-Computer Interaction. MIT Press, Cambridge (2005)
Sieckenius de Souza, C., Fontoura de Gusmão Cerqueira, R., Marques Afonso, L., Rossi de Mello Brandão, R., Soares Jansen Ferreira, J.: Software Developers as Users. Springer, Cham (2016). https://doi.org/10.1007/978-3-319-42831-4
De Souza, C.S., Leitão, C.F.: Semiotic engineering methods for scientific research in HCI. Synth. Lect. Hum. Center. Inf. 2(1), 1–122 (2009)
Denecke, K., Tschanz, M., Dorner, T., May, R.: Intelligent conversational agents in healthcare: hype or hope? Stud. Health Tech. Inf. 259, 77–84 (2019)
Denzin, N.K., Lincoln, Y.S.: Collecting and interpreting qualitative materials, vol. 3. SAGE Publications, Thousand Oaks (2008)
Diniz, M.A.A., Monteiro, D.Q., Gratão, A.C.M.: Education for informal elderly caregivers (educação em saúde para cuidadores informais de idosos). Soc. Health Soc. Change (Saúde & Transformação) 7(1), 028–040 (2016)
Dohr, A., Modre-Opsrian, R., Drobics, M., Hayn, D., Schreier, G.: The Internet of things for ambient assisted living. In: Seventh International Conference on Information Technology: New Generations (ITNG), Las Vegas, NV, USA, pp. 804–809. IEEE (2010)
Easton, K., et al.: A virtual agent to support individuals living with physical and mental comorbidities: co-design and acceptability testing. J. Med. Internet Res. 21(5), e12996 (2019)
El Kamali, M., Angelini, L., Caon, M., Andreoni, G., Khaled, O.A., Mugellini, E.: Towards the nestore e-coach: a tangible and embodied conversational agent for older adults. In: Proceedings of the 2018 ACM International Joint Conference and 2018 International Symposium on Pervasive and Ubiquitous Computing and Wearable Computers, pp. 1656–1663. ACM (2018)
Eschweiler, G.W., Wanner, L.: How can intelligent conversational agents help? The needs of geriatric patients and their caregivers. In: Intelligent Conversation Agents in Home and Geriatric Care Applications, Stockholm, Sweden, pp. 1–9. CEUR-WS (2018)
Jusoh, S.: Intelligent conversational agent for online sales. In: International Conference on Electronics. Computers and Artificial Intelligence (ECAI), Iasi, Romania, pp. 1–4. IEEE (2018)
Kebian, L.V.A., Acioli, S.: The home visit of nurses and community health agents of the family health strategy. Revista Eletrônica de Enfermagem 16(1), 1–9 (2014)
King, A.C., Bickmore, T.W., Campero, M., Pruitt, L.A., Yin, J.L.: Employing virtual advisors in preventive care for underserved communities: results from the compass study. J. Health Commun. 18(12), 1449–1464 (2013). https://doi.org/10.1080/10810730.2013.798374. pMID: 23941610
Ministry of Welfare and Social Assistance of Brazil (Ministério da Previdência e Assistência Social do Brasil): Publicity of the role of caregiver (Publicização do papel do cuidador domiciliar) (1998). Accessed 17 Jan 2020
Montenegro, J.L.Z., da Costa, C.A., da Rosa Righi, R.: Survey of conversational agents in health. Expert Syst. Appl. 129, 56–67 (2019)
Nikitina, S., Callaioli, S., Baez, M.: Smart conversational agents for reminiscence. CoRR abs/1804.06550 (2018). http://arxiv.org/abs/1804.06550
Olson, J.S., Kellogg, W.A. (eds.): Ways of Knowing in HCI. Springer, New York (2014). https://doi.org/10.1007/978-1-4939-0378-8
World Health Organization: Global action plan on the public health response to dementia 2017–2025 (2017)
Schneider, R.H., Irigaray, T.Q.: Aging today: chronological, biological, psychological and social aspects. Estudos de Psicologia 25(4), 585–593 (2008)
Shaked, N.A.: Avatars and virtual agents-relationship interfaces for the elderly. Healthcare Technol. Lett. 4(3), 83–87 (2017)
Sidner, C.L., et al.: Creating new technologies for companionable agents to support isolated older adults. ACM Trans. Interac. Intell. Syst. (TiiS) 8(3), 17 (2018)
Tokunaga, S., Horiuchi, H., Tamamizu, K., Saiki, S., Nakamura, M., Yasuda, K.: Deploying service integration agent for personalized smart elderly care. In: 2016 IEEE/ACIS 15th International Conference on Computer and Information Science (ICIS), pp. 1–6, June 2016. https://doi.org/10.1109/ICIS.2016.7550873
Tsiourti, C., et al.: A virtual assistive companion for older adults: design implications for a real-world application. In: Bi, Y., Kapoor, S., Bhatia, R. (eds.) IntelliSys 2016. LNNS, vol. 15, pp. 1014–1033. Springer, Cham (2018). https://doi.org/10.1007/978-3-319-56994-9_69
Utami, D., Bickmore, T., Nikolopoulou, A., Paasche-Orlow, M.: Talk about death: end of life planning with a virtual agent. IVA 2017. LNCS (LNAI), vol. 10498, pp. 441–450. Springer, Cham (2017). https://doi.org/10.1007/978-3-319-67401-8_55
Vardoulakis, L.P., Ring, L., Barry, B., Sidner, C.L., Bickmore, T.: Designing relational agents as long term social companions for older adults. In: Nakano, Y., Neff, M., Paiva, A., Walker, M. (eds.) IVA 2012. LNCS (LNAI), vol. 7502, pp. 289–302. Springer, Heidelberg (2012). https://doi.org/10.1007/978-3-642-33197-8_30
Wargnier, P., Carletti, G., Laurent-Corniquet, Y., Benveniste, S., Jouvelot, P., Rigaud, A.S.: Field evaluation with cognitively-impaired older adults of attention management in the embodied conversational agent louise. In: IEEE International Conference on Serious Games and Applications for Health (SeGAH), pp. 1–8. IEEE (2016)
World Health Organization (WHO): Global Health and Aging (2011)
Yaghoubzadeh, R., Kramer, M., Pitsch, K., Kopp, S.: Virtual agents as daily assistants for elderly or cognitively impaired people. In: Aylett, R., Krenn, B., Pelachaud, C., Shimodaira, H. (eds.) IVA 2013. LNCS (LNAI), vol. 8108, pp. 79–91. Springer, Heidelberg (2013). https://doi.org/10.1007/978-3-642-40415-3_7
Yin, R.K.: Case Study Research. SAGE Publications, Thousand Oaks (2008)
Zhang, L., Babu, S.V., Jindal, M., Williams, J.E., Gimbel, R.W.: A patient-centered mobile phone app (iHeartU) with a virtual human assistant for self-management of heart failure: protocol for a usability assessment study. JMIR Res. Protoc. 8(5), e13502 (2019)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2020 Springer Nature Switzerland AG
About this paper
Cite this paper
Monteiro, M., Salgado, L., Seixas, F., Santana, R. (2020). Co-designing Strategies to Provide Telecare Through an Intelligent Assistant for Caregivers of Elderly Individuals. In: Gao, Q., Zhou, J. (eds) Human Aspects of IT for the Aged Population. Healthy and Active Aging. HCII 2020. Lecture Notes in Computer Science(), vol 12208. Springer, Cham. https://doi.org/10.1007/978-3-030-50249-2_12
Download citation
DOI: https://doi.org/10.1007/978-3-030-50249-2_12
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-50248-5
Online ISBN: 978-3-030-50249-2
eBook Packages: Computer ScienceComputer Science (R0)