An Embodied Conversational Agent in an eHealth Self-management Intervention for Chronic Obstructive Pulmonary Disease and Chronic Heart Failure: Exploratory Study in a Real-life Setting

Abstract

Background Embodied conversational agents (ECAs) have the potential to stimulate actual use of eHealth apps. An ECA’s design influences the user’s perception during short interactions, but daily life evaluations of ECAs in health care are scarce. Objective This is an exploratory, long-term study on the design of ECAs for eHealth. The study investigates how patients perceive the design of the ECA over time with regard to the ECA’s characteristics (friendliness, trustworthiness, involvement, expertise, and authority), small talk interaction, and likeliness of following the agent’s advice. Methods We developed an ECA within an eHealth self-management intervention for patients with both chronic obstructive pulmonary disease (COPD) and chronic heart failure (CHF), which we offered for 4 months. Patients rated 5 agent characteristics and likeliness of following the agent’s advice before use and after 3 and 9 weeks of use. The amount of patients’ small talk interaction was assessed by log data. Lastly, individual semistructured interviews were used to triangulate results. Results Eleven patients (7 male and 4 female) with COPD and CHF participated (median age 70 years). Patients’ perceptions of the agent characteristics did not change over time (P>.05 for all characteristics) and only 1 participant finished all small talk dialogues. After 3 weeks of use, the patients were less likely to follow the agent’s advice (P=.01). The agent’s messages were perceived as nonpersonalized and the feedback as inappropriate, affecting the agent’s perceived reliability. Conclusions This exploratory study provides first insights into ECA design for eHealth. The first impression of an ECA’s design seems to remain during long-term use. To investigate future added value of ECAs in eHealth, perceived reliability should be improved by managing users’ expectations of the ECA’s capabilities and creating ECA designs fitting individual needs. Trial Registration Netherlands Trial Register NL6480; https://www.trialregister.nl/trial/6480

Attrition in Conversational Agent–Delivered Mental Health Interventions: Systematic Review and Meta-Analysis

“…Our study also found that including any form of visual representation of a CA may be associated with lower attrition rates compared with no visualization at all. This is aligned with many studies on the design of CAs that stressed the importance of design to create positive perceptions of the CA [ 94 ]. However, a recent scoping review reported that visual representation of the CA showed mixed and no association with subjective user experience [ 95 ].…”

Embodied Conversational Agents for Chronic Diseases: Scoping Review

“…Among the studies that reported a study design, mixed method studies (8/36, 22%) [ 10 , 28 , 39 - 44 ] and pilot studies (8/36, 22%) [ 31 , 34 , 38 , 45 - 48 ] were the most commonly used study designs, followed by qualitative studies (6/36, 17%) [ 11 , 26 , 35 , 49 - 51 ], randomized controlled trials (5/36, 14%) [ 32 , 37 , 52 - 54 ], ongoing trials (4/36, 11%) [ 33 , 55 - 57 ], quasiexperimental designs (2/36, 6%) [ 30 ], longitudinal study (1/36, 3%) [ 29 ], single-group nonrandomized trial (1/36, 3%) [ 36 ], and pre- and posttest design (1/36, 3%) [ 27 ]. Of the 36 studies, 30 (83%) were conducted on patients (15/36, 42%) [ 10 , 26 , 32 , 34 - 37 , 40 , 42 , 47 , 52 , 54 - 57 ] or healthy individuals (15/36, 42%) [ 11 , 27 , 29 - 31 , 44 - 46 , 48 - 51 , 53 ] and 6 (17%) [ 28 , 33 , 38 , 39 , 41 , 43 ] on stakeholders such as homecare providers or experts. The mean age of the target population across studies was >40 years, except for 19% (7/36) of the studies [ 30 , 33 , 35 , 47 , 53 , 57 ].…”

Experiences and perceptions of final‐year nursing students of using a chatbot in a simulated emergency situation: A qualitative study

“…A number of studies are currently looking for new approaches to integrate chatbots and AI‐based conversational agents to support health‐related activities, albeit the quality of content still needs to be improved (Park et al, 2019 ; To et al, 2021 ). While the information was found to be adequate, accessible and useful, nearly all participants felt that the content output could be more concise, accurate and employ appropriate length responses and also include other options to motivate participants to explore other related content at their own pace (Stal et al, 2021 ). One possible explanation for this could be the need of different levels of chatbot personalization (intrinsic, extrinsic or a mix of both) in order to create user profiles or user models and support personalized and adaptative features (Fang et al, 2018 ; Kocaballi et al, 2019 ).…”