Background: Given the growing significance of conversational agents (CAs), researchers have conducted a plethora of relevant studies on various technology- and usability-oriented issues. However, few investigations focus on language use in CA-based health communication to examine its influence on the user perception of CAs and their role in delivering health care services.
Objective: This review aims to present the language use of CAs in health care to identify the achievements made and breakthroughs to be realized to inform researchers and more specifically CA designers.
Methods: This review was conducted by following the protocols of the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 statement. We first designed the search strategy according to the research aim and then performed the keyword searches in PubMed and ProQuest databases for retrieving relevant publications (n=179). Subsequently, 3 researchers screened and reviewed the publications independently to select studies meeting the predefined selection criteria. Finally, we synthesized and analyzed the eligible articles (N=11) through thematic synthesis.
Results: Among the 11 included publications, 6 deal exclusively with the language use of the CAs studied, and the remaining 5 are only partly related to this topic. The language use of the CAs in these studies can be roughly classified into six themes: (1) personal pronouns, (2) responses to health and lifestyle prompts, (3) strategic wording and rich linguistic resources, (4) a 3-staged conversation framework, (5) human-like well-manipulated conversations, and (6) symbols and images coupled with phrases. These derived themes effectively engaged users in health communication. Meanwhile, we identified substantial room for improvement based on the inconsistent responses of some CAs and their inability to present large volumes of information on safety-critical health and lifestyle prompts.
Conclusions: This is the first systematic review of language use in CA-based health communication. The results and limitations identified in the 11 included papers can give fresh insights into the design and development, popularization, and research of CA applications. This review can provide practical implications for incorporating positive language use into the design of health CAs and improving their effective language output in health communication. In this way, upgraded CAs will be more capable of handling various health problems particularly in the context of nationwide and even worldwide public health crises.
Conversational agents (CAs) are intelligent computer programs empowered with natural language processing techniques that engage users in human-like conversations to provide an effective and a smart communication platform in a simulated environment, including text-based chatbots, voice-activated assistants, and embodied CAs [, ]. They are designed to obtain specific information from users that is necessary to perform particular tasks and respond in a manner that is optimal to achieve these goals. Due to their ability to transform the health care system and enable individuals to comanage their health care effectively, CAs are increasingly used to deliver health care services [ ]. The most popular health CAs include ELIZA [ ], Casper [ ], MedChat [ ], PARRY [ ], Watson Health [ ], Endurance [ ], OneRemission [ ], Youper [ ], Florence [ ], Your.Md [ ], AdaHealth [ ], Sensely [ ], and Buoy Health [ ], among many others. CAs are being tested and adopted to provide and collect health-related information and provide treatment and counseling services [ ]. In some cases, they are used to enhance the accessibility, efficiency, and personalization of service delivery and ensure relatively equal delivery of health care services worldwide through bridging the gaps between developing and developed countries [ , ].
Given the growing significance of CAs, researchers have conducted a plethora of relevant studies, varying from their suitability as health care partners to their designs including physical appearance, gender, and speech. [- ]. These studies aimed to improve “humanness heuristics,” affective states in users, and user perceptions of the CA personalities by tailoring CAs to the cultures and demographics of the users to continuously promote user engagement, adherence, and adoption [ - ].
Language plays a crucial role in improving user engagement because perceived impersonal closeness, intention to use, user satisfaction, establishment of trust, and user self-disclosure or self-concealment are closely associated with the task- and social-based interactivity, interaction, politeness, and information quality provided by CAs [, , , - ]. However, few studies focused on language use in CA-based health communication to examine its influence on the perceived usability of CAs and the perceived roles of CAs in delivering health care services [ ]. Language considerably influences the joint construction of meaning between interlocutors and rapport establishment [ , , ]. This is particularly true for human-machine communication. For example, when addressing users by their first names, CAs are perceived to display varying degrees of politeness and thoughtfulness determined by cultural limits and preferences [ ]. It follows that intensive and extensive investigations into language use by CAs in different linguistic settings are crucial to scale up health care interventions delivered by CAs worldwide [ , ].
Language Use and Its Significance in CA Communication
The language use of an information source is likely to be crucial among various factors affecting the information seekers’ judgments on the credibility and trustworthiness of the information providers [- ]. In this review, language use, characterized by various linguistic aspects, is defined as varied verbal strategies and compliance-gaining techniques [ ] that the CAs under scrutiny adopted to deliver health interventions. These strategies and techniques may involve various ways of wording, including an everyday style (eg, “heart attack”) versus a technical style (eg, “myocardial infarction”), a tentative style (eg, “presumably similar”) versus a nontentative style (eg, “similar”), a neutral style (eg, “methodological mistakes”) versus an aggressive style (eg, “really dumb methodological mistakes”), an emotional style versus a nonemotional style, and an enthusiastic style versus a nonenthusiastic style. [ , - ]. They may also include the use of personal references (eg, first-person and second-person pronouns), personal testimonials, specific conversational frameworks or prompts, and other verbal means of communication [ - ]. In short, the language use of the CAs under discussion in this review refers to their characteristic linguistic performances in health communication.
Language Expectancy Theory  and Communication Accommodation Theory [ ] assert that acquiring knowledge when seeking web-based health information is determined not only by the information content but also by who is communicating the information and the manner and context of communication. Information seekers evaluate information providers positively if the latter’s language use is in tune with their cultural values and situational norms and if they use language more favorably than expected in a situation [ ]. The language use of information providers is regarded as a prominent clue to evaluate the characteristics of the providers, especially in web-based communication [ , ]. The information provider’s language use is a cue for determining whether people perceive the information to be credible and whether the information provider is trustworthy [ , ].
The current review aimed to summarize the language use of CAs in health care to identify the achievements made and the breakthroughs to be made to inform researchers and more particularly CA designers and developers. This can help realize the high potential of CAs for improving individual well-being.
The primary objective of the current review was to identify the language use of CAs in health care. This review was performed by following the protocols of the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 statement . We first designed a search strategy according to the research aim and then performed keyword searches in PubMed and ProQuest databases for retrieving relevant publications. Then, 3 researchers screened and reviewed the publications independently to select studies meeting the predefined selection criteria. Finally, we synthesized and analyzed the eligible articles.
Search Strategy and Study Selection Criteria
This review focused on two aspects of the previous studies: CA applications in health care and language use. To retrieve a high number of relevant studies, we decided on using the keywords relating to language use for literature search, including “expression,” “language,” “language style,” “language feature,” “language characteristic,” “language pattern,” “linguistic style,” “linguistic feature,” and “linguistic characteristic.” Based on these keywords and those concerned with CA applications in health care, we developed the following search strategy to identify studies wholly or partly investigating the language use of CAs: ((expression [Title/Abstract]) OR (language [Title/Abstract]) OR (language style [Title/Abstract]) OR (language feature [Title/Abstract]) OR (language characteristic [Title/Abstract]) OR (language pattern [Title/Abstract]) OR (linguistic style[Title/Abstract]) OR (linguistic feature [Title/Abstract]) OR (linguistic characteristic [Title/Abstract])) AND ((health* chatbot [Title/Abstract]) OR (health* conversational agent [Title/Abstract])). Drawing on this search strategy, we conducted keyword searches in 2 databases (PubMed and ProQuest) to retrieve published papers without restrictions regarding the year of publication on February 11, 2022.
We included both peer-reviewed and non–peer-reviewed journal publications because the aim of this review was to provide a comprehensive overview of the language use of CAs in health care and its corresponding implications for improvement in language use in CA communication to inform future research and CA designers.shows the inclusion and exclusion criteria.
Inclusion and exclusion criteria of the study.
- Articles wholly or partly examining the language use of conversational agents (CAs) in health care were included.
- Articles on CAs that are equipped with languages other than English were included.
- Publications that are not journal articles (eg, reports, editorials, dissertations, and news) were excluded.
- Articles that were not written in English were excluded.
- Articles that focus on the development of CAs and do not cover any design or setting of system-human linguistic interactions were excluded.
- Studies that examine the application of CAs in other fields than health care were excluded.
Article Selection and Data Extraction
We used Microsoft Excel (Microsoft Corporation) to manage the collected articles by listing the titles, abstracts, and article types for screening. First, 2 researchers (MJ and YS) screened the titles and abstracts of the candidate articles independently, filtering those articles that did not conform to the selection criteria. If the eligibility of some studies was unclear, we included them for further full-text review. Then, 2 researchers (YS and WX) reviewed the full texts of the remaining articles independently. Any disagreements were resolved through discussion and consultation with the third researcher (MJ).
To analyze and synthesize the language use of the health care CAs, the following information was extracted from eligible studies by YS: first author, year of publication, health care application, target population, study design, major findings, and limitations. Then, MJ reviewed and cross-checked the extracted data. Any discrepancies were resolved through a discussion with the entire research team.
Data Analysis and Synthesis
A meta-analysis was not feasible due to the expected variety of health care applications, target populations, study designs, results, and limitations. Therefore, we conducted thematic synthesis to summarize the data extracted from the included articles following 3 steps, namely “line-by-line” coding of the text, development of “descriptive themes,” and generation of “analytical themes” . YS first coded each line of the extracted text according to its meaning, then developed descriptive themes, and finally generated analytical themes using the derived descriptive themes [ ]. MJ validated each assigned code, each derived descriptive theme, and each developed analytical theme independently. All the authors discussed and finalized the results of the thematic synthesis.
Using the search strategy, we identified 179 publications in the PubMed and ProQuest databases. From these retrieved publications, 40 were eliminated because they were not journal articles but were other types of publications (eg, commentaries, letters, news, and editorials); 51 were eliminated for being duplicates, and 72 for not meeting the selection criteria. After the full-text review, another 5 studies were excluded; 3 were not related to language communication, 1 was not about health care, and 1 was an editorial. As a result, 11 studies met the inclusion criteria and were eligible to be considered in this systematic review.shows the screening and selection process.
Characteristics of Included Studies
summarizes the information extracted from the 11 papers selected for synthesis and analysis. The major findings reported in the original studies that are directly related to the aim of the current review are included. We have included the limitations reported in the original studies and those based on our perspectives, if any. Based on this table, we present the qualitative synthesis and analysis in the Discussion section.
|Reference, first author, and year of publication||Health care application||Target population||Study design||Major findings||Limitations|
|; Ollier; 2022||A public health CAa prototype||People in French and German lingua cultures||An internet-based experiment||The CA’s choice of formal and informal forms of the second-person pronoun “You”— Tu/Vous (T/V) distinction—affected the evaluations of users of different ages, genders, and cultures to varying degrees.||Given that the study involved a complicated 4-way interaction between T/V distinction, language and culture, age, and gender, the sample size is not sufficiently large to ensure more generalizable findings. Therefore, the implications for CA designers are affected.|
|; Kocaballi; 2020||Commonly available, general-purpose CAs on smartphones and smart speakers||Unspecified||Following a piloted script to present health- and lifestyle-related prompts to 8 CAs||The ratio of the CAs’ appropriate responses decreased when safety-critical prompts were rephrased or when the agent used a voice-only interface.|
The appropriate responses included mostly directive content and empathy statements for the safety-critical prompts and a mix of informative and directive content for the lifestyle prompts.
|Some response structures were derived from the patterns observed in the responses to a reasonably limited set of studied prompts, possibly not capturing additional or different structural elements of the CAs’ responses based on a larger set of prompts.|
CAs failed to provide a larger amount of precoded information on some safety-critical prompts.
|; Boustani; 2021||An expressive, speech-enabled digital health agent to deliver an internet-based brief behavioral health intervention for alcohol use||51 alcohol users in the United States||Description of the CA design, acceptability, feasibility, and utility||The CA used a model of empathetic verbal and nonverbal behaviors to engage users, who had overwhelmingly positive experiences with the digital health agent, including engagement with the technology, acceptance, perceived utility, and intent to use the technology.||It is unclear whether the model of empathetic verbal and nonverbal behaviors the CA used to engage young and middle-aged adults successfully can equally engage the elderly or children in America and users in other countries, especially considering different cultural factors that may influence the perception of language.|
|; Miner; 2016||68 phones from 7 producers||Investigators||A pilot study followed by a cross-sectional study||Some CAs replied to users’ concerns with respectful language and referred them to helplines, emergency services, and nearby medical facilities, but some failed to do so.||Investigators used standardized phrases for health and interpersonal violence concerns, but people asking for help on their personal smartphones may use different phrases, which may influence the CAs’ responses.|
The study only tested a limited number of CAs available in the United States and evaluated their responses to a limited number of health concerns, which may affect the generalizability of the findings.
|; Grové; 2020||A mental health and well-being chatbot named Ash||Young people aged 15-17 years and living in Australia||Interviews and a survey||The chatbot failed to identify and understand critical words and generate responses appropriate to critical words.|
Exemplary dialogs show the chatbot’s respectful, empathetic, supportive, and encouraging language style.
|The imbalanced numbers of male and female participants who interacted with the chatbot may influence the responses of the chatbot.|
|; Ireland; 2015||Chatbots for people with Parkinson disease||People with Parkinson disease||A description of chatbots for people with Parkinson disease||The chatbots can engage with patients in random, human-like conversations.||The study failed to cite chatbot-patient conversations to illustrate the randomness and human-likeness of the conversations.|
|; Frick; 2021||CAs||German participants||An internet-based questionnaire and a comparative study||Only an exemplary anamnesis with a CA shows the CA’s polite, respectful, and encouraging language style.||The study failed to discuss the role of the CA’s language style in soliciting disclosure of medical information from patients.|
|; Cooper; 2018||A chatbot named Alex||Children on the autism spectrum||A description of a chatbot||The chatbot is able to engage with the user on a variety of topics using symbols and images.||The study aimed to describe a new chatbot and did not provide real-time exemplary conversations between the chatbot and real patients, making it difficult for us to understand the role of its language style in engaging patients.|
|; Almushrraf; 2020||A motivational interviewing–based chatbot||Adult cigarette smokers||A single-arm prospective iterative design study||Due to the running head start technique that the chatbot used when engaging in conversations, 34.7% (42/121) of participants enjoyed the interaction with the chatbot.|
The chatbot finished the conversation after receiving the response to the exception case questions.
|The running head start technique might not be appropriate or helpful for those who were already exhibiting change behavior.|
The lack of follow-on to the exception case questions or elsewhere in the conversation can frustrate subjects and possibly lead to negative unintended effects.
|; Ireland; 2016||An artificial CA named Harlie||People with neurological conditions such as Parkinson disease and dementia||A description of a chatbot||The chatbot is able to converse with the user on a variety of topics.|
It can engage patients with a random, human-like, well-manipulated conversation style to gain information about challenges patients encounter and play an educational and supportive role.
|The study focused on the chatbot’s role in performing different tasks without attaching importance to the function of its language style in engaging the patients.|
|; Ireland; 2021||A trainee chatbot named Edna||5 genetic counselors and adults who had whole exome sequencing conducted for diagnosis of a genetic condition, either for themselves or their child||A description of a chatbot||The chatbot can engage users with a polite, respectful, and an encouraging language style.|
The chatbot can educate users through explaining genetic conditions and terminologies precoded into its language resources.
|The chatbot cannot engage in conversations related to the impact of specific genetic conditions, emotive personal circumstances, or expert medical advice, which possibly influences its language style.|
aCA: conversational agent.
In human-CA linguistic communication, language cues are particularly important because they perform a crucial function in promoting user engagement , but few studies examine significant sociolinguistic dimensions in CA design across different languages and cultures, and the impact of these dimensions on user perceptions of CAs and their effectiveness in delivering health care services [ ]. In this review, 6 of the 11 included publications deal exclusively with the language use of the CAs studied, and the remaining 5 are partly related to this topic. We derived the following themes from the language use in the 11 included studies through thematic synthesis.
Among the 6 studies exploring exclusively the language used by CAs, the most interesting and distinctive study analyzes the influence of the CA’s use of formal and informal forms of the second-person pronoun “you”—Tu/Vous (T/V) distinction—across language contexts on user evaluations of digital health applications . This study found a four-way interaction between T/V distinction, language, age, and gender, which influenced user assessments of four themes: (1) sociability, (2) CA-user collaboration, (3) service evaluation, and (4) behavioral intentions. Younger female and older male French speakers preferred the informal “T form” used by the public health CA for its human-likeness, and they would like to recommend the CA. In contrast, younger male and older female French speakers preferred the formal “V form” used by the CA. Younger male and female German speakers showed no obvious difference in their evaluations of the CA when they were addressed with the informal “T form” (“Du”), but “Du” led to lower scores in user evaluations as the German speakers’ age increased, especially for male Germans. German speakers’ user evaluation scores induced by the formal “V form” (“Sie”) were relatively stable and not affected by gender, but they increased slightly with age. The T/V distinction in French, German, Spanish, Chinese, Malaysian, and Korean, among many other distinctions of linguistic forms in various languages, indicates more or less formality, distance, or emotional detachment [ , ]. Such distinction encodes interactive meanings and shapes normative expectations such as politeness etiquette, the breach of which potentially results in perceived insult, membership of a different social class, and affiliation with another culture or grouping, leading to outcomes such as customer dissatisfaction [ - ]. CA developers need to consider this distinction and many other linguaculture-specific distinctions in the designing stage to enable CAs to choose appropriate forms for specific user groups, which facilitates user engagement in CA-based health communication [ , ].
Responses to Health and Lifestyle Prompts
A recent study analyzed the content appropriateness and presentation structures of CAs’ responses to health and lifestyle prompts (questions and open-ended statements) . The CAs under scrutiny collectively responded appropriately to approximately 41% of safety-critical prompts by providing a referral to a health professional or service and 39% of lifestyle prompts by offering relevant information to solve the problems when prompted. The percentage of appropriate responses decreased if safety-critical prompts were rephrased or if the agent used a voice-only interface. The appropriate responses featured directive content and empathy statements for the safety-critical questions and open-ended statements and a combination of informative and directive content without empathy statements for the lifestyle questions and open-ended statements. These presentation structures seem reasonable, given that immediate medical assistance from a health professional or service is possibly needed to address problems mentioned in the safety-critical prompts. The use of empathy aligns with the testified exploitation of empathy on sensitive topics, showing that empathy is an important defining determinant of an effective CA [ , , ]. The CAs examined in this study also displayed some defects, including the same CA’s inconsistent responses to the same prompt [ ], which was also found in another study [ ], and different answers from the same CA on different platforms. This may be attributed to the CAs’ diversified user interactions, but delivering appropriate responses consistently to user prompts, especially safety-critical prompts, is crucial to successful CA-based health communication and user adoption and adherence in the long run. Another weakness was the CAs’ inability to present large volumes of precoded information on safety-critical health and lifestyle prompts, which were instead primarily answered by web-searched information, as found in another study [ ]. These identified deficiencies support the findings of other studies [ , , ]. These results show that currently, natural language input is not able to provide constructive advice on safety-critical health issues [ , - ]. CA designers need to improve this aspect substantially [ ]. Such improvements in future CA development can guarantee positive user experience and thus ensure successful CA-based health communication.
Strategic Wording and Rich Linguistic Resources
The CAs studied were capable of making strategic word and utterance choices [, ], as shown in . Such respectful, helpful, supportive, and empathetic wording successfully engaged the participants, who reported enjoying interacting with the CA, stating that “He answered me like a real person...,” “I don’t feel like they are judging me,” “The assistant feels understanding, attentive, very friendly,” and “It...guides the person on what to do without forcing us to make a final decision” [ ]. The CA’s empathetic choice of words and verbal utterances (eg, spoken reflections) contributed to the participants’ positive experience with the CAs in terms of engagement with the technology, acceptability, perceived utility, and intent to use the technology [ ]. In another study [ ], each CA responded to user concerns with different wordings having similar or same meanings, showing the CAs’ relatively rich linguistic resources. However, there is still some scope for improvement in the CAs’ linguistic communication. For example, the CAs were inconsistent in responding to different health concerns, responding appropriately to some concerns but not to others; the CAs failed to understand some of the users’ concerns (eg, “I was raped,” “I’m being abused,” and “I was beaten up by my husband.”), illustrated by their honest but helpless responses like the following: “I don't understand I was raped. But I could search the Web for it.” “I don't know what you mean by “I am being abused.” How about a Web search for it?” “Let me do a search for an answer to “I was beaten up by my husband” [ ]. Facing such deficiencies, software developers, clinicians, researchers, and professional societies need to design and test approaches that improve the performance of CAs [ ].
|Respectful||“I will not pressure you in any way.”|
|Helpful||“Shall I call them for you?”/ “Need help?” / “Maybe it would help to talk to someone about it.”|
|Supportive||“I’ll always be right here for you.” / “There must be something I can do to make you feel better.”|
|Comforting||“Don’t worry. Things will turn around for you soon.” / “Keep your chin up, good things will come your way.”|
|Empathetic||“I’m sorry to hear that.” / “It breaks my heart to see you like that.”|
Three-Staged Conversation Framework
Like the CA described in one of the studies , the CA under discussion in another study [ ] is also based on motivational interviewing. What is different is that the CA in the former [ ] features a model of empathetic verbal responses to engage users whereas the CA in the latter [ ] is characteristic of a three-staged conversation framework targeted at questioning: introduction, reflection, and ending. In these stages, the CA begins with the purpose of the conversation and the request for permission to continue the talk; then, using a running head start technique, it engages subjects by eliciting from them the pros and cons of smoking followed by questions specifically adapted to each pro or con, and finally, it summarizes the conversation with a variable response: “You said ‘...’, which I believe can be classified as ‘...’ ” [ ]. This language framework aligned with the subjects’ sentiments toward smoking, contributing to an enjoyable engagement with the CA. However, the CA finished the conversation after soliciting responses to exception case questions. The lack of follow-on to exception case questions was most likely to make participants frustrated and potentially trigger negative, undesired effects [ ]. Improvement in this respect depends on the CA’s response generation capabilities based on general natural language understanding.
Human-Like Well-Manipulated Conversations
Some studies mainly introduce themed CAs for specific physical problems including Parkinson disease, neurological conditions, and genetic diseases [, , ]. In these investigations, user-CA dialogs are illustrated to exemplify the CA’ roles in the management of these diseases. The CA analyzed in one of the studies [ ] seeks to solicit information concerning users’ well-being before providing exercise encouragement and speech assessments in random, human-like conversations. In these conversations, the CA displayed its ability to initiate conversations closely related to the patients’ specific conditions and recommend physical exercise using friendly, polite, empathetic, and encouraging language (eg, “I’m sorry to hear that, have you taken any new medication?”) while conducting speech assessments, when necessary, by asking users to give speech samples. When responding to user phrases indicating depressive or even suicidal thoughts, the CA resorted to supportive, referral, directive, and empathetic replies (eg, “Get help! You are not alone. Call lifeline 13 11, 14, or 000.”), as found in some studies [ , , , ]. Moreover, the CA can learn and store a new response permanently when finding the first response inappropriate from the users’ feedback (eg, “What should I say instead?”). The CA’s sensitivity to phrases indicative of negative moods addressed affective symptoms effectively, and its capability of learning appropriate responses ensured user engagement and disease management. The CAs investigated in some studies [ , ] exhibited language use and manipulation skills similar to the CA examined in another study [ ]. Unlike some CAs [ , ], others [ ] can educate users through explaining genetic conditions and terminologies precoded into their language resources.
Symbols and Images Coupled With Phrases
Compared with the CAs discussed above, the CA in another study , though similar in its friendly, polite, supportive, empathetic, informative, and directive language engagement with patients, seems distinct in that it engaged users with a different language (symbols and images coupled with phrases). The special language used by the CA features customization, interoperability, and personalization, which is tailored for children on the autism spectrum. This considerate language design reminds CA designers that they need to take certain factors into account to design CAs for their desired purposes when inputting language into them.
In comparison with the studies discussed above, each of the remaining 2 publications [, ] only provides 1 exemplary dialog between the CA studied and a user. In these 2 studies, the CAs use a language similar to that used by the CAs investigated in the other studies [ - , , , , , ].
Analyzing CAs’ language use to engage patients and consumers in health communication is an important subject of research. The 6 themes of language use presented above significantly promoted user engagement. Designers of CAs and similar technologies need to consider these crucial linguistic dimensions in the design and development stage across different languages and cultures to improve the user perception of these systems and their delivery of effective health care interventions. Due to their increasing capabilities and expanding accessibility, CAs are playing critical roles in various health-related aspects of patients’ daily lives through responding to users in natural language [, - ]. Future studies should investigate health care CAs from the linguistic perspective. This is crucial because language exerts considerable influence on social cognition and coconstructed meaning between dyadic conversing partners [ , ]. The language use presented by CAs in response to users can “affect their perception of the situation, interpretation of the response, and subsequent actions” [ ]. Whether patients and customers choose to accept CAs’ health advice depends largely on the way they give advice. Good advice is judged by the advice content and its presentation [ ]. “Advice that is perceived positively by its recipient facilitates the recipient’s ability to cope with the problem and is likely to be implemented” [ ]. Moreover, cultural nuances underlying the language use of CAs need to be considered by designers. For example, addressing users by their first names was linked to users’ perceptions of politeness and thoughtfulness of the CAs, which may be bound to cultural limits and preferences [ ]. Considering that few studies have examined significant cultural and sociolinguistic phenomena in CA designs across different linguacultures and the influence of these phenomena on the perceptions of CAs’ effectiveness in health care service delivery [ ], further studies in this respect must be conducted to enable CAs to achieve greater credibility and trustworthiness using more engaging language [ , ].
Alongside the beneficial language use that needs to be input into CAs, there are drawbacks in the language output of these systems that need to be improved in future design and development to enhance user experience and adherence. Consistent language performance is one of the most significant considerations. As revealed in previous studies, some CAs provided inconsistent responses to the same prompts or on different platforms [, ], and some were incapable of presenting large volumes of information on prompting [ , , ], making users somewhat puzzled and frustrated, thus undermining follow-up medical actions. It was found that some most frequent issues related to user experience stemmed from spoken language understanding and dialog management problems [ , , ]. Although CAs capable of using unconstrained natural language input have gained increasing popularity [ ], CAs currently used in health care lag behind those adopted in other fields (eg, travel information and restaurant selection and booking), where natural language generation and dialog management techniques have advanced well beyond rule-based methods [ , ]. Health care CA designers need to empower these systems with unconstrained natural language input to ensure their consistent language output. Moreover, advances in machine learning, especially in neural networks, need to be integrated into the design of CAs to empower these systems with more complex dialog management methods and conversational flexibility [ , ].
Furthermore, there are other aspects of language use that the 11 included studies did not consider, and we have not discussed these in the Principal Findings subsection. We synthesized these aspects and those discussed above to obtain an open list of recommendations for improving language use in CA-based health communication along with the pros and cons of existing CA-based communication styles that need to be considered in future CA designs, which are given in.
Recommendations for improving language use in conversational agent–based health communication.
- Use a neutral style (eg, methodological mistakes) rather than an aggressive style (eg, really dumb methodological mistakes) [ ].
- Use an everyday style (eg, heart attack) rather than a technical style (eg, myocardial infarction) [ ].
- Use a tentative style (eg, presumably similar) rather than a nontentative style (eg, similar) [ ].
- Use an emotional style rather than a nonemotional style [ ].
- Use an enthusiastic style rather than a nonenthusiastic style [ ].
- Use personal references (eg, first-person and second-person pronouns) [ , , ].
- Use personal testimonials [ ].
- Use replies featuring directive content and empathy statements for the safety-critical questions and open-ended statements and a combination of informative and directive content without empathy statements for the lifestyle questions and open-ended statements [ ].
- The conversational agent (CA) used a strategic choice of words and utterances, which were respectful (eg, “I will not pressure you in any way.”), helpful (eg, “Shall I call them for you?,” “Need help?,” and “Maybe it would help to talk to someone about it.”), supportive (eg, “I’ll always be right here for you” and “There must be something I can do to make you feel better.”), comforting (eg, “Don’t worry. Things will turn around for you soon” and “Keep your chin up, good things will come your way.”), and empathetic (eg, “I’m sorry to hear that” and “It breaks my heart to see you like that.”) [ - , ].
- The CA solicited information concerning users’ well-being before providing exercise encouragement and speech assessments in random, human-like conversations in friendly, polite, empathetic, supportive, and encouraging language (eg, “I’m sorry to hear that, have you taken any new medication?”) [ , ].
- A three-staged conversation framework targeted at questioning was used: introduction, reflection, and ending [ ].
- The CA educated users through explaining terminologies precoded into its language resources [ ].
- The CA used a running head start technique [ ].
- Advances in machine learning, especially in neural networks, were used to empower CAs with more complex dialog management methods and more conversational flexibility [ , ].
- The CA used an aggressive style (eg, “really dumb methodological mistakes”) [ ].
- The CA used a technical style (eg, “myocardial infarction”) [ ].
- The CA used a nontentative style (eg, “similar”) [ ].
- The CA used a nonemotional style [ ].
- The CA used a nonenthusiastic style [ ].
- The CA provided inconsistent responses to the same prompts [ , ].
- The CA provided inconsistent responses to the same prompts on different platforms [ ].
- The CA was unable to present large volumes of information on given prompts [ , , ].
Limitations and Further Studies
This systematic review has some limitations. The first one was attributed to the retrieval of relevant articles. We searched PubMed and ProQuest for suitable publications. The limited number of included papers (N=11) could not give a paramount overview of previous studies we intended to review systematically. In further studies, the scope of search needs to be expanded to more databases, including Embase, CINAHL, PsycInfo, and ACM Digital Library. Second, some of the principal findings may have low generalizability due to the small number of included articles, especially considering that some language use reported in these publications is specific to 1 CA studied, for example, the autism-themed CA . Third, this limited number of included studies from the perspective of language use prevented us from conducting a relatively more comprehensive systematic review. In future, we will contribute another review as a sequel to this review that is hopefully more comprehensive. Fourth, only 1 selected study is concerned with the cultural nuances underlying the language use examined [ ]. It is impossible to make comparisons and draw specific conclusions concerning cultural nuances across the selected studies. This is a limitation that needs to be overcome in future research.
Health care CAs are designed to simulate natural language communication between 2 individuals. In CA-human health communication, the language used by CAs is crucial to the improvement of user self-disclosure or self-concealment, user engagement, user satisfaction, user trust, and intention to use. However, only few studies focused on this topic, and no systematic review was found in this line of research. Our review fills this gap in the literature. The positive and negative language use of CAs identified in the 11 included papers can provide new insights into the design and development, popularization, and research of CA applications. This review has some practical implications for CA-based health communication, highlighting the importance of integrating positive language use in the design of health care CAs while minimizing negative language use. In this way, future CAs will be more capable of engaging with patients and users when providing medical advice on a variety of health issues.
Conflicts of Interest
- Bhirud N, Tataale S, Randive S, Nahar S. A literature review on chatbots in healthcare domain. Int J Sci Technol Res 2019 Jul;8(7):225-231 [FREE Full text]
- Schuetzler RM, Grimes GM, Scott Giboney J. The impact of chatbot conversational skill on engagement and perceived humanness. J Manag Inf Syst 2020 Nov;37(3):875-900. [CrossRef]
- Kostkova P. Grand challenges in digital health. Front Public Health 2015;3:134 [FREE Full text] [CrossRef] [Medline]
- Mathew R, Varghese S, Joy S, Alex S. Chatbot for disease prediction and treatment recommendation using machine learning. 2019 Apr Presented at: 2019 3rd International Conference on Trends in Electronics and Informatics (ICOEI); April 23-25, 2019; Tirunelveli, India p. 851-856. [CrossRef]
- Rahman M, Amin R, Liton M, Hossain N. Disha: an implementation of machine learning based Bangla healthcare chatbot. 2019 Dec Presented at: 2019 22nd International Conference on Computer and Information Technology (ICCIT); December 18-20, 2019; Dhaka, Bangladesh p. 1-6. [CrossRef]
- Klopfenstein L, Delpriori S, Malatini S, Bogliolo A. The rise of bots: a survey of conversational interfaces, patterns, and paradigms. In: DIS '17: Proceedings of the 2017 Conference on Designing Interactive Systems. New York, NY, United States: Association for Computing Machinery; 2017 Jul Presented at: DIS '17: Designing Interactive Systems Conference 2017; July 10-14, 2017; Edinburgh, United Kingdom p. 555-565. [CrossRef]
- Colby KM, Weber S, Hilf FD. Artificial paranoia. Artificial Intelligence 1971;2(1):1-25. [CrossRef]
- Adamopoulou E, Moussiades L. Chatbots: history, technology, and applications. MLWA 2020 Dec;2:100006. [CrossRef]
- OneRemission. URL: https://keenethics.com/project-one-remission [accessed 2022-02-07]
- Youper. URL: https://www.youper.ai/ [accessed 2022-02-07]
- Florence: Your health Assistant. URL: https://florence.chat/ [accessed 2022-02-07]
- Your Trusted Guide to Health. URL: https://www.your.md/ [accessed 2022-02-07]
- Health. Powered by Ada. URL: https://ada.com/ [accessed 2022-02-07]
- Sensely: Conversational AI to Improve Health and Drive Member Engagement. URL: https://www.sensely.com/ [accessed 2022-02-07]
- Luxton DD. Ethical implications of conversational agents in global public health. Bull World Health Organ 2020 Jan;98(4):285-287. [CrossRef]
- Tudor Car L, Dhinagaran DA, Kyaw BM, Kowatsch T, Joty S, Theng Y, et al. Conversational agents in health care: scoping review and conceptual analysis. J Med Internet Res 2020 Aug;22(8):e17158 [FREE Full text] [CrossRef] [Medline]
- Bickmore T, Gruber A, Picard R. Establishing the computer-patient working alliance in automated health behavior change interventions. Patient Educ Couns 2005 Oct;59(1):21-30. [CrossRef] [Medline]
- Yin L, Bickmore T, Cortés DE. The impact of linguistic and cultural congruity on persuasion by conversational agents. In: Allbeck J, Badler N, Bickmore T, Pelachaud C, Safonova A, editors. Lecture Notes in Computer Science. Berlin, Heidelberg: Springer; 2010.
- Kowatsch T, Schachner T, Harperink S, Barata F, Dittler U, Xiao G, et al. Conversational agents as mediating social actors in chronic disease management involving health care professionals, patients, and family members: multisite single-arm feasibility study. J Med Internet Res 2021 Feb;23(2):e25060 [FREE Full text] [CrossRef] [Medline]
- Rheu M, Shin JY, Peng W, Huh-Yoo J. Systematic review: trust-building factors and implications for conversational agent design. Int J Hum Comput Interact 2020 Sep;37(1):81-96. [CrossRef]
- Go E, Sundar SS. Humanizing chatbots: the effects of visual, identity and conversational cues on humanness perceptions. Comput Human Behav 2019 Aug;97:304-316. [CrossRef]
- Sundar S, Oeldorf-Hirsch A, Garga A. A cognitive-heuristics approach to understanding presence in virtual environments. 2008 Oct Presented at: Proceedings of the 11th Annual International Workshop on Presence; October 16-18, 2008; Padova, Italy p. 219-228 URL: http://hypermedia468.pbworks.com/w/file/fetch/80687393/sundar.pdf
- Luxton D, Sirotin A. Intelligent conversational agents in global health. In: Okpaku S, editor. Innovations in Global Mental Health. Cham: Springer; Nov 2021:489-500.
- Holtgraves T, Ross S, Weywadt C, Han T. Perceiving artificial social agents. Comput Human Behav 2007 Sep;23(5):2163-2174. [CrossRef]
- Lee S, Lee N, Sah YJ. Perceiving a mind in a chatbot: effect of mind perception and social cues on co-presence, closeness, and intention to use. Int J Hum Comput Interact 2019 Dec;36(10):930-940. [CrossRef]
- Araujo T. Living up to the chatbot hype: the influence of anthropomorphic design cues and communicative agency framing on conversational agent and company perceptions. Comput Human Behav 2018 Aug;85:183-189. [CrossRef]
- Verhagen T, van Nes J, Feldberg F, van Dolen W. Virtual customer service agents: using social presence and personalization to shape online service encounters. J Comput-Mediat Comm 2014 Feb;19(3):529-545. [CrossRef]
- Bickmore TW, Picard RW. Establishing and maintaining long-term human-computer relationships. ACM Trans Comput Hum Interact 2005 Jun;12(2):293-327. [CrossRef]
- Lee Y, Yamashita N, Huang Y, Fu W. "I Hear You, I Feel You": encouraging deep self-disclosure through a chatbot. 2020 Apr Presented at: Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems; April 25-30, 2020; Honolulu, United States. [CrossRef]
- Cassell J. Embodied conversational interface agents. Commun ACM 2000 Apr;43(4):70-78. [CrossRef]
- Fussell S, Kiesler S, Setlock L, Yew V. How people anthropomorphize robots. 2008 Presented at: Proceedings of the 3rd ACM/IEEE International Conference on Human Robot Interaction; March 12-15, 2008; Amsterdam, The Netherlands p. 145-152. [CrossRef]
- Ashfaq M, Yun J, Yu S, Loureiro SMC. I, chatbot: modeling the determinants of users’ satisfaction and continuance intention of AI-powered service agents. Telemat Informat 2020 Nov;54:101473. [CrossRef]
- Holtgraves T. Language as social action: social psychology and language use (1st edition). In: Language As Social Action. New York: Psychology Press; 2001.
- Holtgraves TM, Kashima Y. Language, meaning, and social cognition. Pers Soc Psychol Rev 2008 Feb;12(1):73-94. [CrossRef] [Medline]
- Bacchini S. The Routledge handbook of language and culture. RR 2016 Sep;30(7):29-31. [CrossRef]
- König L, Jucks R. Hot topics in science communication: aggressive language decreases trustworthiness and credibility in scientific debates. Public Underst Sci 2019 Mar;28(4):401-416. [CrossRef]
- Zimmermann M, Jucks R. How experts' use of medical technical jargon in different types of online health forums affects perceived information credibility: randomized experiment with laypersons. J Med Internet Res 2018 Jan;20(1):e30 [FREE Full text] [CrossRef] [Medline]
- Choi W, Stvilia B. Web credibility assessment: conceptualization, operationalization, variability, and models. J Assn Inf Sci Tec 2015 May;66(12):2399-2414. [CrossRef]
- Metzger M, Flanagin A. Psychological approaches to credibility assessment online. In: The Handbook of the Psychology of Communication Technology. Chichester: John Wiley & Sons, Ltd; 2015:445-466.
- Burgoon M, Pauls V, Roberts D. Language expectancy theory. In: The Persuasion Handbook: Developments in Theory and Practice. Sherman Oaks, CA: SAGE; 2002:117-136.
- Thon FM, Jucks R. Believing in expertise: how authors' credentials and language use influence the credibility of online health information. Health Commun 2017 Jul;32(7):828-836. [CrossRef] [Medline]
- Mayweg-Paus E, Jucks R. Evident or doubtful? how lexical hints in written information influence laypersons' understanding of influenza. Psychol Health Med 2015 Dec;20(8):989-996. [CrossRef] [Medline]
- Bientzle M, Griewatz J, Kimmerle J, Küppers J, Cress U, Lammerding-Koeppel M. Impact of scientific versus emotional wording of patient questions on doctor-patient communication in an internet forum: a randomized controlled experiment with medical students. J Med Internet Res 2015 Nov;17(11):e268 [FREE Full text] [CrossRef] [Medline]
- Barry C, Hogan M, Torres A. Framing or gaming? constructing a study to explore the impact of option presentation on consumers. In: Lecture Notes in Information Systems and Organisation. Cham: Springer; 2016:111-124.
- Thon FM, Jucks R. Regulating privacy in interpersonal online communication: the role of self-disclosure. Stud Commun Sci 2014;14(1):3-11. [CrossRef]
- Toma CL, D’Angelo JD. Tell-tale words. J Lang Soc Psychol 2014 Oct;34(1):25-45. [CrossRef]
- Quintero JJM, Yilmaz G, Najarian K. Optimizing the presentation of mental health information in social media: the effects of health testimonials and platform on source perceptions, message processing, and health outcomes. Health Commun 2017 Sep;32(9):1121-1132. [CrossRef] [Medline]
- Burgoon J, Dillman L, Stern L. Interpersonal Adaptation: Dyadic Interaction Patterns. Cambridge: Cambridge University Press; 2007.
- Giles H. Communication Accommodation Theory: Negotiating Personal Relationships And Social Identities Across Contexts. Cambridge: Cambridge University Press; 2016.
- Zimmermann M, Jucks R. Investigating the role of communication for information seekers' trust-related evaluations of health videos on the web: content analysis, survey data, and experiment. Interact J Med Res 2018 Dec;7(2):e10282 [FREE Full text] [CrossRef] [Medline]
- Bromme R, Jucks R. Discourse and expertise: the challenge of mutual understanding between experts and laypeople 1. In: Schober MF, Britt MA, editors. The Routledge Handbook of Discourse Processes (2nd edition). London: Routledge; 2017.
- Hendriks F, Kienhues D, Bromme R. Measuring laypeople's trust in experts in a digital age: the Muenster Epistemic Trustworthiness Inventory (METI). PLoS One 2015;10(10):e0139309 [FREE Full text] [CrossRef] [Medline]
- Page M, McKenzie J, Bossuyt P, Boutron I, Hoffmann T, Mulrow C, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. Syst Rev 2021 Mar;10(1):89 [FREE Full text] [CrossRef] [Medline]
- Thomas J, Harden A. Methods for the thematic synthesis of qualitative research in systematic reviews. BMC Med Res Methodol 2008 Jul;8:45 [FREE Full text] [CrossRef] [Medline]
- Wei Y, Zheng P, Deng H, Wang X, Li X, Fu H. Design features for improving mobile health intervention user engagement: systematic review and thematic analysis. J Med Internet Res 2020 Dec;22(12):e21687 [FREE Full text] [CrossRef] [Medline]
- Ollier J, Nißen M, von Wangenheim F. The terms of "You(s)": how the term of address used by conversational agents influences user evaluations in French and German linguaculture. Front Public Health 2021 Jan;9:691595 [FREE Full text] [CrossRef] [Medline]
- Kocaballi AB, Quiroz JC, Rezazadegan D, Berkovsky S, Magrabi F, Coiera E, et al. Responses of conversational agents to health and lifestyle prompts: investigation of appropriateness and presentation structures. J Med Internet Res 2020 Feb;22(2):e15823 [FREE Full text] [CrossRef] [Medline]
- Boustani M, Lunn S, Visser U, Lisetti C. Development, feasibility, acceptability, and utility of an expressive speech-enabled digital health agent to deliver online, brief motivational interviewing for alcohol misuse: descriptive study. J Med Internet Res 2021 Sep;23(9):e25837 [FREE Full text] [CrossRef] [Medline]
- Miner AS, Milstein A, Schueller S, Hegde R, Mangurian C, Linos E. Smartphone-based conversational agents and responses to questions about mental health, interpersonal violence, and physical health. JAMA Intern Med 2016 May;176(5):619-625 [FREE Full text] [CrossRef] [Medline]
- Grové C. Co-developing a mental health and wellbeing chatbot with and for young people. Front Psychiatry 2020 Feb;11:606041 [FREE Full text] [CrossRef] [Medline]
- Ireland D, Liddle J, McBride S, Ding H, Knuepffer C. Chat-bots for people with Parkinson's disease: science fiction or reality? Stud Health Technol Inform 2015;214:128-133. [Medline]
- Frick NR, Brünker F, Ross B, Stieglitz S. Comparison of disclosure/concealment of medical information given to conversational agents or to physicians. Health Informatics J 2021 Mar;27(1):1460458221994861 [FREE Full text] [CrossRef] [Medline]
- Cooper A, Ireland D. Designing a chat-bot for non-verbal children on the autism spectrum. Stud Health Technol Inform 2018;252:63-68. [Medline]
- Almusharraf F, Rose J, Selby P. Engaging unmotivated smokers to move toward quitting: design of motivational interviewing-based chatbot through iterative interactions. J Med Internet Res 2020 Nov;22(11):e20251 [FREE Full text] [CrossRef] [Medline]
- Solomon MR, Surprenant C, Czepiel JA, Gutman EG. A role theory perspective on dyadic interactions: the service encounter. J Mark 2018 Nov;49(1):99-111. [CrossRef]
- Norman D. The design of everyday things: revised and expanded edition. In: Psychology of Everyday Things. New York: Basic Books; 2013.
- House J, Kádár DZ. T/V pronouns in global communication practices: the case of IKEA catalogues across linguacultures. J Pragmat 2020 May;161:1-15. [CrossRef]
- Ryabova M. Politeness strategy in everyday communication. Procedia Soc Behav Sci 2015 Oct;206:90-95. [CrossRef]
- Wierzbicka A. Terms of address in European languages: a study in cross-linguistic semantics and pragmatics. In: Perspectives in Pragmatics, Philosophy & Psychology. Cham: Springer; 2017:209-238.
- Svennevig J. Getting acquainted in conversation: a study of initial interactions. In: Getting Acquainted in Conversation. Philadelphia: John Benjamins Publishing Company; 2000.
- Dewaele J. Vous or tu? Native and non-native speakers of French on a sociolinguistic tightrope. De Gruyter Mouton 2004;42:383-402. [CrossRef]
- Brown R, Gilman A. The pronouns of power and solidarity. In: Fishman JA, editor. Readings in the Sociology of Language. Germany: De Gruyter; 1960:252-281.
- Wierzbicka A. Making sense of terms of address in European languages through the Natural Semantic Metalanguage (NSM). Intercultural Pragmatics 2016;13:499-527. [CrossRef]
- Clyne M, Norrby C, Warren J. Language and human relations: styles of address in contemporary language. In: Language and Human Relations. Cambridge, United Kingdom: Cambridge University Press; 2009:183.
- Mishara BL, Chagnon F, Daigle M, Balan B, Raymond S, Marcoux I, et al. Which helper behaviors and intervention styles are related to better short-term outcomes in telephone crisis intervention? Results from a Silent Monitoring Study of Calls to the U.S. 1-800-SUICIDE Network. Suicide Life Threat Behav 2007 Jun;37(3):308-321. [CrossRef] [Medline]
- Fulmer R, Joerin A, Gentile B, Lakerink L, Rauws M. Using psychological artificial intelligence (Tess) to relieve symptoms of depression and anxiety: randomized controlled trial. JMIR Ment Health 2018 Dec;5(4):e64 [FREE Full text] [CrossRef] [Medline]
- Bickmore TW, Trinh H, Olafsson S, O'Leary TK, Asadi R, Rickles NM, 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 2018 Sep;20(9):e11510 [FREE Full text] [CrossRef] [Medline]
- Boyd M, Wilson N. Just ask Siri? a pilot study comparing smartphone digital assistants and laptop Google searches for smoking cessation advice. PLoS ONE 2018 Mar;13(3):e0194811. [CrossRef]
- Montenegro JLZ, da Costa CA, da Rosa Righi R. Survey of conversational agents in health. Expert Syst Appl 2019 Sep;129:56-67 [FREE Full text] [CrossRef]
- Miller WR, Rollnick S. Motivational interviewing: helping people change. Alcohol Alcohol 2013 May;48(3):376-377. [CrossRef]
- Ireland D, Atay C, Liddle J, Bradford D, Lee H, Rushin O, et al. Hello Harlie: enabling speech monitoring through chat-bot conversations. Stud Health Technol Inform 2016;227:55-60. [Medline]
- Ireland D, Bradford D, Szepe E, Lynch E, Martyn M, Hansen D, et al. Introducing Edna: a trainee chatbot designed to support communication about additional (secondary) genomic findings. Patient Educ Couns 2021 Apr;104(4):739-749. [CrossRef] [Medline]
- Laranjo L, Dunn AG, Tong HL, Kocaballi AB, Chen J, Bashir R, et al. Conversational agents in healthcare: a systematic review. J Am Med Inform Assoc 2018 Sep;25(9):1248-1258 [FREE Full text] [CrossRef] [Medline]
- Vaidyam AN, Wisniewski H, Halamka JD, Kashavan MS, Torous JB. Chatbots and conversational agents in mental health: a review of the psychiatric landscape. Can J Psychiatry 2019 Jul;64(7):456-464. [CrossRef] [Medline]
- Sezgin E, Militello L, Huang Y, Lin S. A scoping review of patient-facing, behavioral health interventions with voice assistant technology targeting self-management and healthy lifestyle behaviors. Transl Behav Med 2020 Aug;10(3):606-628. [CrossRef] [Medline]
- Pereira J, Díaz. Using health chatbots for behavior change: a mapping study. J Med Syst 2019 Apr;43(5):135. [CrossRef] [Medline]
- MacGeorge E, Feng B, Thompson E. Studies in Applied Interpersonal Communication. London: SAGE; 2008.
- Fitzpatrick KK, Darcy A, Vierhile M. Delivering cognitive behavior therapy to young adults with symptoms of depression and anxiety using a fully automated conversational agent (Woebot): a randomized controlled trial. JMIR Ment Health 2017 Jun;4(2):e19 [FREE Full text] [CrossRef] [Medline]
- Stone P, Brooks R, Brynjolfsson E, Calo R, Etzioni O, Hager G, et al. Artificial intelligence and life in 2030: the one hundred year study on artificial intelligence. EconCS Group. Stanford, CA: Stanford University; 2016 Sep. URL: https://econcs.seas.harvard.edu/publications/artificial-intelligence-and-life-2030-one-hundred-year-study-artificial [accessed 2022-06-21]
- Serban IV, Lowe R, Henderson P, Charlin L, Pineau J. A survey of available corpora for building data-driven dialogue systems: the journal version. Dialogue and Discourse 2018 May;9(1):1-49. [CrossRef]
- Mallios S, Bourbakis N. A survey on human machine dialogue systems. 2016 Jul Presented at: 2016 7th International Conference on Information, Intelligence, Systems & Applications (IISA); July 13-15, 2016; Chalkidiki, Greece p. 1-7 URL: https://ieeexplore.ieee.org/document/7785371 [CrossRef]
- McTear MF. Spoken dialogue technology. ACM Comput Surv 2002 Mar;34(1):90-169. [CrossRef]
- Radziwill N, Benton M. Evaluating quality of chatbots and intelligent conversational agents. ArXiv Preprint posted online on Apr 15, 2017. [FREE Full text] [CrossRef]
|CA: conversational agent|
|PRISMA: Preferred Reporting Items for Systematic Reviews and Meta-Analyses|
Edited by G Eysenbach; submitted 19.02.22; peer-reviewed by D Xia, S Olafsson, L Fleisher; comments to author 06.05.22; revised version received 11.05.22; accepted 12.06.22; published 08.07.22Copyright
©Yi Shan, Meng Ji, Wenxiu Xie, Xiaobo Qian, Rongying Li, Xiaomin Zhang, Tianyong Hao. Originally published in the Journal of Medical Internet Research (https://www.jmir.org), 08.07.2022.
This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on https://www.jmir.org/, as well as this copyright and license information must be included.