Remember me
Health-related behaviors refer to the actions and choices that individuals make that impact their physical, mental, and emotional well-being. These behaviors can either promote or compromise an individuals’ health []. Health risk behavior refers to an action performed by individuals that, because of its frequency or intensity, increases the likelihood of developing a disease or injury. It may occur whether or not the person is aware of the connection between the behavior and the associated health risks []. Conversely, positive health behaviors encompass actions that contribute to disease prevention, early detection of disease and disability, promotion and enhancement of overall health, and safeguarding against injury risk. These activities aim to maintain and improve one’s well-being [].
Positive health behaviors are crucial for disease prevention, physical well-being, mental and emotional health, longevity and quality of life, productivity and performance, social interactions and relationships, and financial savings []. By proactively making positive behavioral changes, individuals can take control of their health, strive for long-term well-being, and prevent and reduce noncommunicable diseases (NCDs), widely acknowledged as chronic conditions, encompassing afflictions, such as cardiovascular disorders, neoplastic growths, persistent respiratory ailments, and diabetes []. This phenomenon arises because of the shared presence of behavioral risk factors among numerous NCDs, which comprise habits such as tobacco use, inadequate physical activity, unwholesome dietary patterns, and deleterious alcohol intake. NCDs account for 7 of the top 10 global causes of death [,] and cause 41 million deaths annually, equivalent to 74% of global deaths [].
Promoting health prevention through improved health-related behaviors is a cost-effective and low-risk alternative to medication []. However, behavior changes can be challenging. Public health experts have developed a wide range of interventions to facilitate behavior change, including motivational interviewing (MI), which is recognized as highly effective []. MI combines brief interventions and motivational enhancement therapies [] to encourage individuals to adopt healthier lifestyles [,]. MI aims to increase motivation and self-efficacy in adopting health-promoting behaviors using directive communication techniques and a person-centered approach. In MI, the counselors establish a collaborative, empathic, and nonjudgmental relationship with patients, using strategies such as reflective listening, strategic questioning, affirmations, and emphasizing patient autonomy to elicit change talk []. Change talk includes statements expressing an individual’s desires, abilities, reasons, needs, commitments, activations, and steps to modify behavior [].
MI is guided by 4 principles that provide a framework for effective communication and behavior change in the counseling process [-]:
Express empathy: show genuine understanding and empathy toward the individual’s feelings and experiences, fostering a supportive and nonjudgmental environment.Highlight discrepancies: help clients recognize discrepancies between their current behaviors and their desired goals, motivating them to consider change.Roll with resistance: avoid confrontations and work with resistance by understanding its roots and navigating through it collaboratively.Support self-efficacy: encourage belief in the client’s ability to make positive changes, empowering them to act toward their goals.In addition to these principles, MI uses 4 interconnected processes that support individuals in exploring motivations, values, and abilities for positive behavior change [-]:
Engaging: develop a collaborative, trusting relationship with active listening and empathy to create a safe space for discussing changes.Focusing: clarify behavior change goals by exploring motivations, values, and aspirations for intervention directions.Evoking: elicit intrinsic motivation through reflective listening, affirmations, and open-ended questions to empower self-exploration.Planning: collaboratively develop an achievable action plan with specific goals, strategies to overcome barriers, and tailored to individual strengths and preferences.Finally, MI uses 4 essential techniques to facilitate effective communication and engagement [-]:
Open questions: encourage clients to express themselves freely and explore their thoughts and feelings in depth.Affirmations: provide positive and supportive statements that acknowledge client strengths, efforts, and achievements, fostering trust and motivation.Reflections: offer empathetic restatements of clients’ expressions, demonstrating active listening and understanding their perspectives.Summaries: provide condensed recaps of the key points discussed, helping clients organize their thoughts and reinforcing important messages.Accessibility of MI is limited by factors such as cost, logistics, social stigma, convenience, and counselor availability. Consequently, digital technology, including computers, smartphones, tablets, and the internet, is increasingly being used to deliver MI interventions and promote behavior change [-]. These technological approaches use various techniques such as chat rooms, automated responses, emoticons, decision balances, readiness rulers, and open-ended questions, which have demonstrated efficacy in modifying target behaviors []. Technology-based MI interventions have proven to be effective in promoting positive behavioral changes in chronic disease prevention and management. They offer advantages, such as reducing therapist burden and minimizing the need for extensive clinician training. Moreover, they improve access to care for underserved populations, address the stigma associated with disclosing risky behaviors [], and provide cost-effective and easily accessible solutions for patients []. However, it is important to acknowledge that these interventions may have limitations in fulfilling critical components of MI [], such as establishing rapport and demonstrating empathy [,].
To overcome these limitations, MI interventions delivered through embodied conversational agents (ECAs) have emerged as a promising and innovative approach. ECAs are advanced user interfaces that resemble humans and enable face-to-face conversations with users, incorporating verbal and nonverbal behaviors, such as body movements and facial expressions []. Interacting with ECAs can enhance user engagement and motivation, facilitating long-term use and maximizing their benefits []. ECAs can be developed for various platforms, including PCs and mobile devices, such as smartphones and tablets, enabling continuous use anytime, anywhere. The use of ECAs has been a subject of investigation in the field of clinical psychology, specifically in the context of various conditions, including autism spectrum disorders, major depressive disorder, anxiety disorders, posttraumatic stress disorder, psychotic disorders, and substance use disorders []. They have also been studied to support various health issues, such as overweight, obesity, diabetes, hypertension, and atrial fibrillation [], and for coaching people in a healthy lifestyle, such as physical activity, nutrition, mindfulness, preconception care, stress, blood glucose monitoring, and sun protection []. The use and development of ECAs for health support have significantly increased because of advancements in personal devices (laptops, smartphones, or tablets) and computer techniques (3D game development, speech-to-text, text-to-speech, machine learning, and artificial intelligence) [,].
Using ECAs to deliver MI interventions in the context of improving health behaviors offers several advantages. ECAs can provide consistent, personalized, and nonjudgmental interventions that tailor support to individuals’ specific needs. They can adapt their communication style, tone, and visual cues to enhance engagement and motivation for behavior change []. Compared with text-based interfaces such as Technology-Delivered Adaptations of Motivational Interviewing (TAMI), ECAs have more personalized interactions and better relational skills, potentially providing stronger empathy and social connections []. Although TAMIs may have limited empathy conveyed through textual wording, ECAs can express empathy through verbal and nonverbal behaviors, resembling the approach of a human therapist []. Consequently, incorporating ECAs into MI interventions has the potential to yield improved outcomes [-].
Despite the potential of ECAs delivering MI interventions in improving health-related behaviors, there is limited understanding of their design, development, implementation, evaluation, and effectiveness. Previous reviews have explored ECAs in clinical psychology [] and for coaching people in a healthy lifestyle []; however, they lack a specific analysis of ECA-delivered MI interventions. Similarly, although some reviews discuss TAMIs [,], they do not comprehensively examine ECA-delivered MI interventions. Therefore, a comprehensive review that specifically analyzes ECA-delivered MI interventions is needed.
ObjectivesThe objective of this study was to conduct a scoping review with a focus on analyzing the use of ECAs to deliver MI. This study aimed to address the following questions:
Which health problems are addressed through ECAs providing MI?What are the main characteristics of the ECAs used to provide MI interventions? (eg, device implementation, type of implementation, appearance, dialogue mechanism, and emotional model).How are MI’s principles, processes, and techniques implemented in ECAs for MI interventions?What evaluation protocols and measures are used and what are the primary reported results of interventions through ECAs providing MI?By addressing these questions, this review enhances our understanding of the design, use, evaluation, and impact of ECAs delivering MI interventions, thereby contributing to the knowledge of digital interventions for improving health-related behaviors.
For our review, we adopted the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) methodology [], a recognized framework for conducting systematic and unbiased reviews.
Search StrategySearch SourcesThis review encompassed research papers published in the English language between January 2008 and December 2022, covering the last 15 years. The search encompassed prominent databases, including PubMed, Scopus (Elsevier), IEEE Xplore, ACM Digital, and PsycINFO. The selected databases collectively span the domains of medical, psychological, and computer science literature. An iterative approach was adopted to ensure a thorough search, whereby pertinent papers referenced within the retrieved articles were manually examined, further enhancing the comprehensiveness of the search process.
Search TermsTo develop the search query, the authors checked previous relevant reviews and made a list of words related to “embodied conversional agents” and “motivational interviewing.” We used a specific syntax for the query in each database, ensuring a consistent representation of the constructs: (1) ECAs (ie, virtual agent, digital health agent, and virtual assistant) and (2) MI (ie, brief MI, motivational interview, brief motivational intervention, and motivational intervention). The details of the exact terms used to search each database are presented in .
Study Eligibility CriteriaThe inclusion criteria included papers describing ECAs developed for delivering MI interventions targeting health-related behaviors along with an evaluation procedure. In contrast, the exclusion criteria included papers that did not describe ECAs with a human appearance, did not use ECAs for improving health-related behaviors through MI, lacked evaluation, or lacked an explanation of MI implementation.
ECAs with a human appearance were given priority based on acceptability studies that indicate a preference for human agents over abstract, animal-like, and stylized (cartoon-like) agents [-]. In addition, it has been observed that the appearance of an ECA can influence the way users establish trust, communication, and engagement [,,].
Furthermore, studies focusing solely on training health professionals in MI provision were excluded because the primary goal was to assess ECAs delivering MI for specific health-related problems. Papers without evaluations were also excluded, as they often represented ongoing work in the initial stages of ECA development. Similarly, papers lacking an explanation of MI implementation were not included, as accessing this information was necessary to address the implementation of MI in ECAs (question 3). We applied no restrictions on the study setting, study design, study outcome, month, and country of publication.
Study SelectionWe used a multistage screening process as follows. Initially, 3 reviewers (JM, IEE-C, and JM-M) conducted title and abstract screening. Articles that were included by all 3 reviewers proceeded to the second phase, in which the full text was thoroughly reviewed by the same 3 reviewers (JM, IEE-C, and JM-M) for final inclusion decisions. In instances in which multiple papers were associated with the same study or ECA, a hierarchical approach was implemented, favoring the most recent publication. This preference was maintained unless substantial disparities were evident in the assessment methodology, encompassing divergent metrics or distinct target demographic groups.
Data ExtractionAll studies that complied with the stipulated inclusion criteria were included in the review. The data extraction procedure was conducted by 3 independent reviewers (JM, IEE-C, and JM-M), and any discordance that arose was deliberated on through discourse to attain a unanimous consensus. Data extraction was centered on addressing the research objectives delineated in the Objectives subsection. A table template was developed to systematically extract and summarize relevant data such as ECA characteristics, MI implementation, evaluation, and principal outcomes. provides a description of the data extraction process. It is important to note that the data presented in this review were based solely on the information provided in each paper. If a specific piece of information was not mentioned or included in a paper, it was marked as “not mentioned.”
Data SynthesisFollowing data extraction from the included studies, a narrative approach was used to synthesize the data. The data synthesis was carried out by 3 independent reviewers (JM, IEE-C, and JM-M), and any discordances that arose were deliberated upon through discourse to attain a unanimous consensus. The synthesis of the data focused on summarizing and describing the health problems addressed, ECA characteristics (including appearance, dialogue mechanism, emotional model, deployment device, and implementation level), MI implementation (covering intervention type, principles, processes, and techniques), evaluation approach (including the protocol and measures used), and reported results. Microsoft Excel was used for the management of the synthesized data, whereas Mendeley was the tool of choice for reference management. The subsequent section provides a detailed description of the main findings of the included studies.
The search generated 404 entries, of which 97.3% (n=393) were retrieved from digital repositories and 2.7% (n=11) were manually acquired. The distribution across the individual libraries was as follows: 2.7% (11/404) from PubMed, 77.7% (314/404) from Scopus (Elsevier), 1% (4/404) from IEEE Xplore, 15.9% (64/404) from ACM Digital, and 0% (0/404) from PsycINFO. In the primary stage, 354 studies were assessed after eliminating duplicates. After scrutinizing the titles and abstracts, 87.3% (309/354) were excluded based on exclusion criteria. Subsequently, 45 articles underwent a full-text review, of which 31 (68.9%) were excluded. Thus, 14 studies were retained for further analysis. provides an overview of the review phase.
Regarding geographic distribution, most of the 14 studies originated from the United States (n=10, 71%), followed by the Netherlands (n=3, 21%) and Australia (n=1, 7%). The first study emerged in 2012, and the most recent in 2022. Notably, the number of studies has nearly doubled over the last 5 years. provides a summarized overview of the general results, and provides details of the included studies [-,-]. The following sections detail the key findings that address the defined review objectives.
Figure 1. PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews) flowchart of the study selection. ECA: embodied conversational agent; MI: motivational interviewing. Table 1. Overview of the general results (n=14).CharacteristicPublications, n (%)ReferencesPublication typeaECA: embodied conversational agent.
bMI: motivational interviewing.
Health Problems AddressedIn terms of the health problems addressed by ECAs, of the 14 studies, 5 (36%) focused on reducing unhealthy alcohol use [,,,,]. In addition, 3 (21%) studies aimed to promote physical activity and healthy eating [,,]. Moreover, 2 (14%) studies focused on promoting women’s preconception health [,], whereas another 2 (14%) studies focused on supporting cognitive learning [,]. Furthermore, 1 (7%) study centered on counseling patients in medication-assisted treatment for opioid use disorder [], and 1 (7%) study aimed to support brain injury rehabilitation decision-making processes []. Among those focused on promoting physical activity and healthy eating, 2 of 3 specifically targeted promoting both physical activity and fruit and vegetable consumption [,], whereas the remaining 1 out of 3 solely focused on promoting physical activity [].
The fact that the primary use of ECAs providing MI to support the reduction of alcohol use is not surprising, considering the strong support for MI’s efficacy (without ECAs) in addressing substance use, the very issue for which MI was originally designed []. Among the various drugs addressed, there is substantial evidence that traditional MI has a positive effect on reducing alcohol use []. This finding is consistent with previous reviews that reported TAMIs [,], wherein substance use behaviors, including alcohol use, were the most frequently addressed unhealthy change behaviors.
The next set of health-related behaviors most addressed by ECAs offering MI were those promoting physical activity, healthy eating, and women’s preconception health (5/14, 36%). Similar to the support for alcohol use reduction, this finding aligns with the MI literature, as the promotion of these healthy behaviors represents the second most addressed category of traditional MI []. In addition, previous systematic reviews of TAMIs [,] identified the promotion of these healthy behaviors as part of the second most commonly addressed health-related problems through these technological solutions.
The remaining reviewed studies addressed various other health-related problems, such as supporting cognitive learning, counseling patients in medication-assisted treatment for opioid use disorders, and providing support for brain injury rehabilitation for decision-making processes. The limited number of studies addressing these health problems could be attributed to their focus on more specific populations, necessitating further research to gather better evidence on the positive results of ECAs offering MI for such health-related issues. Notably, none of the reviewed studies explored the use of ECAs that provide MI to support the treatment or recovery of mental health problems. This represents a significant research opportunity, considering that MI is now increasingly used to address mental health disorders, such as anxiety, depression, and others []. Conversely, ECAs, when not providing MI, are also more commonly used for mental health treatments, including depression, anxiety, psychotic disorders, and other conditions []. Consequently, the development and evaluation of ECAs offering MI as a complement to psychotherapeutic interventions would be a relevant research endeavor to assess whether the outcomes for mental health problems can be improved compared with the current approach of using MI and ECAs separately.
Characteristics of the ECAsOverviewThe main aim of ECAs is to implement natural and intuitive interactions with the users. These advanced interfaces are then equipped with a body that interacts multimodally by using verbal, para-verbal, and nonverbal behaviors. A socially intelligent manner of interaction with users should consider an adequate emotional behavior that helps to produce ECA’s reactions externalized by natural, expressive speech, and nonverbal behaviors []. On the basis of these ECA characteristics, we are interested in how the ECA’s appearance, the underlying dialogue mechanism, and the internal emotional model are designed and implemented in the included studies. Moreover, the degree of implementation (fully or not) and the type of devices in which the ECAs can be deployed were also analyzed. presents the characteristics of ECAs.
Table 2. Characteristics of the embodied conversational agents.StudyAppearance (gender)Appearance (ethnicity)Dialogue mechanismEmotional modelDevice implementationImplementation levelLisetti et al []Both (female and male)Matched the ethnicity of the usersRules drivenRecognize user’s emotion and show emotionsPCPrototypeLisetti et al []Both (female and male)Matched the ethnicity of the usersRules drivenRecognize user’s emotion and show emotionsPCFull systemFriederichs et al []Both (female and male)Did not match the ethnicity of the usersPredefined or sequentialOnly shows emotionsPCFull systemJack et al []FemaleMatched the ethnicity of the usersRules drivenNot mentionedPCFull systemSchouten et al []FemaleDid not match the ethnicity of the usersRules drivenMentioned but not implemented (a Wizard-of-Oz study)PCWizard-of-OzOlafsson et al []FemaleDid not match the ethnicity of the usersRules drivenOnly shows emotionsTablet or smartphoneFull systemTielman et al []FemaleDid not match the ethnicity of the usersRules drivenOnly shows emotionsTablet or smartphoneFull systemJack et al 2020 []FemaleMatched the ethnicity of the usersRules drivenOnly shows emotionsPCFull systemOlafsson et al []FemaleDid not match the ethnicity of the usersRules drivenOnly shows emotionsTablet or smartphoneFull systemOlafsson, et al []FemaleDid not match the ethnicity of the usersRules drivenOnly shows emotionsPCPrototypeBoustani et al []Both (female and male)Matched the ethnicity of the usersRules drivenOnly shows emotionsPCFull systemHocking and Maeder []FemaleDid not match the ethnicity of the usersRules drivenNot mentionedPCPrototypeRubin et al []FemaleDid not match the ethnicity of the usersRules drivenOnly shows emotionsTablet or smartphoneFull systemSchouten et al []Both (female and male)Did not match the ethnicity of the usersRules drivenMentioned but not implemented (a Wizard-of-Oz study)PCWizard-of-OzECA AppearanceThe visual appearance of ECAs is a crucial aspect to be considered when they are used as the primary means of interaction. According to Cassell [], the way an ECA looks captures a user’s attention and affects their cognitive and interactive abilities. ECAs have an advantage over chatbots because they can communicate nonverbally through facial expressions and body movements. Therefore, it is important to appropriately represent the embodiment to convey this type of communication. Moreover, simply having an embodied form in interactions with an agent enhances social outcomes, including motivation [], the cornerstone of MI.
Of the 9 reviewed studies, only 4 (44%; 29% of the 14) mentioned a previous study as the basis for selecting the appearance of the ECA [,,,]. Conversely, none of the other studies (10/14, 71%) mentioned any previous study for selecting the ECA’s appearance [-,,,-]. In terms of the ECA’s gender, most ECAs use a female character (9/14, 64%) [,,,,,-], whereas the rest of them present characters of both gender (5/14, 36%) [,,-], and no one provides only a male representation. Among the works that use a female representation, only 2 [,] reported a study about users’ preferences for ECA’s gender, where female ECAs were rated more positively than male ECAs. The greater use of female ECAs may be based on previous studies where young, female ECAs are preferred over male ECAs in health applications [], on stereotypes where women are seen as more suitable for healing activities [], or on previous studies using ECAs in psychotherapy where most users selected a female ECA over a male ECA when both options were offered []. Nevertheless, other studies point out the lack of consensus among users regarding the preferences of ECA’s gender in eHealth []. Thus, more studies are necessary to clearly identify the effect and influence of ECA’s gender on users, particularly when ECA is used to support behavior change.
Ethnicity is another demographic characteristic that influences the visual appearance of ECAs. Among the 14 reviewed studies, only 2 (14%) [,] developed the ECA’s appearance to match the ethnicity characteristics of the target users; 3 (21%) of them offer the option to select the preferred ECA’s ethnic background [,,], whereas most of them did not consider the ethnicity of the target users (9/14, 64%) [,,,,,-]. Although some previous studies highlighted the importance of ethnicity concordance between the ECA and the users in terms of preferences [] and ECA’s task efficacy [], others did not find evidence supporting ethnicity concordance as a predictor of perceived similarity (which, in turn, is associated with higher satisfaction, trust, and liking toward the ECA) [] or an impact on higher social presence ratings []. Because of the lack of consensus regarding user-ECA ethnicity concordance, an effective strategy is to offer users a set of ECAs with different ethnic backgrounds, as provided in the studies by Lisetti et al [], Boustani et al [], and Lisetti []. This approach allows users the opportunity to choose the appearance with which they feel most comfortable, thereby increasing the adoption of this type of interface and, consequently, improving its efficacy toward behavior change.
ECA Dialogue MechanismVerbal communication is of the utmost importance in the successful implementation of MI techniques, encompassing activities such as posing open-ended questions, making affirmations, and summarizing key points. Therefore, it is crucial to evaluate how ECAs offering MI incorporate these techniques when interacting with users. Most ECAs in the reviewed studies (13/14, 93%) used a rules-driven dialogue mechanism and determined the next dialogue based on the progress of the session [,,,-]. Only 1 of 14 (7%) studies used a predefined and sequential dialogue mechanism []. Furthermore, in 1 study that used a rule-driven dialogue mechanism, it also presented the first steps toward a machine learning–based dialogue engine and presented results on training a machine learning model to classify utterances [].
Evidently, rule-driven approaches are widely used. In such approaches, the ECA’s responses to the user are determined based on predefined rules that consider information about the session’s progression. Alternatively, some studies opt for a more simplistic mechanism that relies on a prescripted dialogue for both ECA and user inputs. The prevalence of rule-driven approaches is not surprising, as natural language understanding, turn-taking management, and natural language generation present challenging obstacles. These challenges are prone to errors, which can have significant consequences in health applications, potentially leading to harmful outcomes [].
Remarkably, only one of the reviewed studies introduced a dialogue mechanism that harnesses machine learning algorithms trained with annotated counseling sessions. This sophisticated mechanism automatically generates subsequent counseling actions at each turn of the dialogue, allowing unconstrained user speech []. The emergence of large language models, such as GPT-3/4, holds great promise in addressing many of the challenges in natural language processing and can significantly enhance MI sessions delivered by an ECA. Preliminary efforts have already been made to leverage GPT to automatically generate reflections on patients’ statements regarding their behavior []. Nevertheless, the use of large language models in psychotherapy necessitates careful consideration of certain factors, including risk detection, transparency, and bias, to ensure safe and effective therapeutic sessions [].
ECA Emotional ModelIn terms of ECAs’ behavior, the conveying of coherent emotional reactions toward user inputs is one of the characteristics of ECAs that facilitates the building of trust and a social bond with the user. There is evidence that artificial agents that express emotions and mood as part of their behavior are perceived to be more empathic, resulting in higher trust []. In addition, the emotions expressed by ECAs influence how users respond to them []. The expression of emotions has a communicative function: an emotion is used to communicate social feedback and empathy []. The communication of empathy in ECAs involves 2 processes: the recognition of the user’s emotion or emotions or experiences and the simulation of emotional cues in the ECA through, for example, facial expressions and body movements. It is widely accepted that empathy involves both cognitive and affective attributes []. Cognitive attributes of empathy involve understanding another individual’s experience and to communicating that understanding, whereas affective attributes involve emotional expressive responses to someone else’s display of emotions []. Thus, as empathy is one of the key principles of MI, it is important to know whether (and how) ECAs providing MI sessions implement emotional or empathic communication with the user.
Of the reviewed studies, 8 of 14 (57%) mentioned that the ECA included some mechanism to display empathy without explicitly recognizing the user’s emotions [-,,,-]. Another 2 of 14 (14%) participants recognized the user’s emotion and incorporated a mechanism to show empathy [,]. In 2 of 14 (14%) studies, it was not mentioned whether the ECA used any mechanism to display empathy [,]. Finally, 2 of 14 (14%) studies mentioned that the ECA displays empathic reactions but did not implement a mechanism to produce them, as they presented Wizard-of-Oz scenarios [,].
As can be seen, only one ECA demonstrates the capability to assess the user’s emotional state by analyzing real-time facial expressions. This ECA conveys empathy through techniques, such as reflective listening, head nodes, and facial expressions [,]. Other studies have implemented diverse emotional reactions and used empathic communication with the user using verbal utterances, such as simple reflections or emphasizing the positive aspects of the user’s current attitude [,]. In addition, nonverbal behaviors such as smiling, hand flips, and nodding were also used []. To provide effective empathic reactions, both cognitive and emotional, during interactions with the user, ECAs should incorporate a mechanism for automatically detecting and recognizing the user’s emotional state. The latest developments in multimodal interaction, where the automatic recognition of emotions is based on visual, audio, and speech recognition [], and the nonverbal behavior of the ECA is generated through different channels [], would benefit ECAs that provide MI sessions.
Device Deployment and Level of ImplementationThe most commonly used devices to present ECAs were PCs (10/14, 71%) [-,,-,,]. Of the 10 studies where the ECA was developed for PCs, 3 mentioned that the ECA was specifically developed for web access on PCs [,,]. The remaining ECAs (4/14, 29%) were designed for use on mobile devices, such as tablets or smartphones [,,,]. It is evident that most of the reviewed ECAs were designed for use on PCs. Considering the significant advancements in smartphones and tablets, such as their high computational capabilities for rendering and animation tasks, it is expected that ECAs would be optimized for these devices to facilitate their anytime and anywhere use. However, a plausible explanation for this trend exists. The ECAs discussed in this review have primarily been developed for research purposes, where strict control over study participants is necessary. Thus, researchers could supply participants with the necessary devices and collect data under controlled conditions. This contrasts with commercial applications, in which mobile devices are deployed on a large scale. Nonetheless, future ECA developments should capitalize on smartphone capabilities, including integrated sensors for detecting user activities, as this can enhance and personalize MI-based recommendations. In terms of implementation level, of the 14 reviewed studies, 9 (64%) presented a full-system development [-,,,,,]. In addition, 3 of 14 (21%) studies presented a prototype of the ECA [,,], whereas the remaining 2 of 14 (14%) studies used a Wizard-of-Oz scenario [,]. In a Wizard-of-Oz scenario, a real person controls the movements and dialogues of the ECA. Prototypes and Wizard-of-Oz scenarios are frequently used during the development phase to assess the specific characteristics of the ECA. Given that some of the reviewed works describe preliminary and pilot studies, it is not surprising that the use of these prototypes serves as a foundation upon which more comprehensive versions can be built, informed by the findings derived from these initial evaluations.
MI ImplementationMI Implementation and Theoretical BackgroundAmong the 14 analyzed studies (), 9 (64%) used traditional MI [-,,,-], whereas 5 (36%) used brief MIs [,,,,]. Within the subset of brief MI studies, the Drinker’s Check-Up method was used in 3 of 5 studies [,,], the Screening, Brief Intervention, and Referral to Treatment model in 1 of 5 studies [], and the Feedback, Responsibility, Advice, Menu, Empathy, and Self-Efficacy model in 1 of 5 studies []. MI was implemented as the core of ECAs in 11 of 14 (79%) studies [-,,,-] and as a component in 3 of 14 (21%) studies [,,].
The transtheoretical model was explicitly specified in 8 of 14 (57%) studies to understand the stages of change [,,,,,-]. In addition, other theories were incorporated in 7 of 14 (50%) studies [,-,,,] to support specific intervention activities, including Self-Determination Theory [], Formal Scaffolding Theory [,], Small Talk Theory [,], and Provider-Patient Communication Theory []. One (7%) study used techniques such as emotional recognition and mindfulness to enhance intervention effectiveness [].
Overall, most of the reviewed studies used traditional MI, whereas a smaller proportion used brief MI. The choice between these 2 approaches might be influenced by various factors, such as setting, time constraints, available resources, and individual or population needs. Traditional MI allows for comprehensive exploration, whereas brief MI uses a more condensed and targeted approach. MI was identified as the central component in 11 (79%) studies, often combined with other elements in 3 (21%) studies. Half of the studies (n=7, 50%) incorporated supplementary theories to support their intervention. Further research is required to evaluate the effectiveness of ECAs using traditional MI and brief MI in diverse settings and populations. In addition, exploring the integration of MI with other techniques or theories to enhance its effectiveness and address individual needs requires further research and development.
Table 3. Motivational interviewing (MI) implementation characteristics.StudyMI typeMI implementation levelMI principlesMI processesMI techniquesLisetti et al []Brief MICompleteEmpathy, develop discrepancy, and roll with resistanceEngaging, focusing, evoking, planning, and tracking of behavior changeAffirmations, reflective listening, and summariesLisetti et al []Brief MICompleteEmpathy, develop discrepancy, and roll with resistanceEngaging, focusing, evoking, planning, and tracking of behavior changeAffirmations, reflective listening, and summariesFriederichs et al []MICompleteDevelop discrepancy and support self-efficacyFocusing, evoking, and planningOpen questions, reflective listening, summaries, and provide information or adviceJack et al []MICompleteExpress empathyEngaging, focusing, evoking, planning, and tracking of behavior changeProvide information or adviceSchouten et al []MIComponentEmpathy, develop discrepancy and self-efficacyEngaging, focusing, and evokingAffirmations, reflective listening, and provide information or adviceOlafssonet al []MICompleteEmpathy, develop discrepancy, roll with resistance, and self-efficacyEngaging, focusing, evoking, and tracking of behavior changeAffirmations, reflective listening, summaries, and provide information or adviceTielmanet al []Brief MIComponentExpress empathy and develop discrepancyFocusing and evokingSummaries and provide information or adviceJack et al []MICompleteExpress empathyEngaging, focusing, evoking, planning, and tracking of behavior changeProvide information or adviceOlafsson et al []MICompleteEmpathy, develop discrepancy, roll with resistance, and self-efficacyEngaging, focusing, evoking, and tracking of behavior changeAffirmations, reflective listening, summaries, and provide information or adviceOlafsson et al []MICompleteNot specifiedPlanning and tracking of behavior changeReflective listeningBoustani et al []Brief MICompleteEmpathy, develop discrepancy, roll with resistance, and self-efficacyFocusing, evoking, and planningAffirmations, reflective listening, and provide information or adviceHocking and Maeder []MICompleteRoll with resistance and self-efficacyFocusing, evoking, planning, and tracking of behavior changeSummaries and provide information or adviceRubin et al []Brief MICompleteDevelop discrepancyFocusing and evokingAffirmations, reflective listening, summaries, and provide information or adviceSchouten et al []MIComponentEmpathy, develop discrepancy, and self-efficacyEngaging, focusing, and evokingAffirmations, reflective listening, and provide information or adviceMI Core Principles, Processes, and Techniques Implemented in the ECAWhen analyzing the core principles of MI (empathy, discrepancy, rolling with resistance, and self-efficacy) [-] in the 14 reviewed studies, the following frequencies were observed (): 3 of 14 (21%) studies included all 4 principles [,,], 4 of 14 (29%) studies included 3 principles [,,,], 3 of 14 (21%) studies included 2 principles [,,], and 4 of 14 (29%) studies included 1 or no principles [,,,]. The specific frequencies for each principle were as follows: empathy was expressed in 10 of 14 (71%) studies [,,-], discrepancy was developed in 10 of 14 (71%) studies [,,-], rolling with resistance was addressed in 6 of 14 (43%) studies [,,-,], and self-efficacy was supported in 7 of 14 (50%) studies [,-,,,]. These results indicate significant variation in the inclusion of core MI principles. Notably, only 21% of the studies encompassed all 4 principles, and 50% included 2 or fewer principles. These findings highlight a deficiency in adhering to MI principles, which are crucial for maintaining fidelity and integrity of the interventions, as they form the foundation for behavior change and motivation. Self-efficacy and the principle of rolling with resistance exhibited the lowest inclusion rate. This suggests a failure to address resistance or ambivalence toward change, emphasizing the need for activities that can enhance beliefs in positive behavior changes.
Regarding the basic processes of MI [-] (engaging, focusing, evoking, and planning), the frequencies observed in the 14 studies were as follows (): 4 of 14 (29%) studies included all 4 processes [,,,], 7 of 14 (50%) studies included 3 processes [,-,,,], 2 of 14 (14%) studies included 2 processes [,], and 1 of 14 (7%) studies included 1 process []. The specific frequencies for each process were as follows: the engaging process was included in 8 of 14 (57%) studies [,,,,-], the focusing process was present in 13 of 14 (93%) studies [-,-,], the evoking process was incorporated in 13 of 14 (93%) studies [-,-,], and the planning process was included in 8 of 14 (57%) studies [-,,,,,]. In addition, the ECAs introduced a supplementary process, namely, tracking of behavior change, which was found in 8 of 14 (57%) studies [,,,-]. Similarly, the implementation of MI processes in ECAs showed a varied inclusion. Only 29% of the studies encompassed all 4 processes, with
Comments (0)