Impact of AI Chatbot modality and belief congruence on health information : In the context of influenza vaccination

다국어 초록 (Multilingual Abstract)

This study, drawing on media synchronicity theory (MST), investigates whether user evaluations of identical pro-vaccination messages vary when delivered by AI chatbots in different modalities (text vs. voice). Furthermore, it examines the extent to which these evaluative outcomes are contingent upon belief congruence—specifically, whether the message aligns with or contradicts the user’s pre-existing belief on influenza vaccination. A 2 × 2 between-subjects online experiment was conducted with 400 adults in South Korea. Participants engaged in a three-minute interaction with a Dialogflow-based AI chatbot that consistently delivered pro-vaccination messages about effectiveness, safety, and necessity. After the interaction, participants reported their perceptions of chatbot credibility, message credibility, and attitude toward influenza vaccination. Three separate 2 × 2 ANOVAs were conducted to test the effects of chatbot modality, belief congruence, and their interaction on each outcome. Results showed significant main effects of chatbot modality and belief congruence on chatbot credibility, message credibility, and attitude toward vaccination. The chatbot modality × belief congruence interaction emerged only for attitude toward vaccination. Specifically, among belief-incongruent participants, attitudes were more favorable in the voice condition than in the text condition, whereas no chatbot modality difference emerged among belief-congruent participants. From an MST perspective, the attitude interaction suggests that chatbot modality-related differences in key media capabilities can matter more when belief incongruence makes coordination of meaning during interaction more salient. Overall, the study highlights chatbot modality as an important design consideration in vaccine-related AI communication, particularly when message acceptance is less straightforward due to belief conflict.

국문 초록 (Abstract)

본 연구는 AI 챗봇의 모달리티(텍스트 vs. 보이스)와 챗봇이 전달하는 메시지의 입장과 사용자의 신념 일치 여부(일치 vs. 불일치)에 따라 인식 및 태도가 어떻게 달라지는지를 Media Synchronicity Theory (MST) 관점에서 살펴보았다. 대한민국 성인 400명을 대상으로 2 × 2 피험자 간 온라인 실험을 진행하였다. 참여자들은 독감 예방접종의 효과성, 안전성, 필요성에 관한 메시지를 제공하는 Dialogflow 기반 AI 챗봇과 3분간 대화하고, 실험 후 챗봇 신뢰성, 메시지 신뢰성, 독감 예방접종 태도를 측정하였다. 챗봇 신뢰성, 메시지 신뢰성, 독감 예방접종 태도 변수별로 2 × 2 분산분석(ANOVA)을 수행한 결과, 세 변수 모두 AI 챗봇 모달리티와 신념 일치 여부에 따른 차이가 확인되었다. 그러나 AI 챗봇 모달리티와 신념 일치 여부의 상호작용 효과는 독감 예방접종 태도에서만 확인되었다. 챗봇이 제공하는 메시지의 입장과 자신의 기존 신념이 일치하지 않는 경우 보이스 챗봇은 텍스트 챗봇보다 더 긍정적인 태도를 유발하였다. 반면, 신념이 일치하는 경우는 두 모달리티 간 차이가 뚜렷하지 않았다. 이러한 결과는 자신의 신념과 충돌하는 정보를 처리하는 과정에서, 보이스 챗봇이 메시지 수용을 더욱 효과적으로 유도할 수 있음을 보여준다. 나아가 독감 예방접종처럼 의견 차이가 발생할 수 있는 주제에서 건강 정보를 전달할 때, 정보의 내용뿐만 아니라 챗봇 모달리티의 선택이 커뮤니케이션 성과에 영향을 미칠 수 있음을 시사한다.

목차 (Table of Contents)

• Chapter 1. Introduction 1
• Chapter 2. Literature Review 5
• 1. AI Chatbots as Interactive Health Communication Agents 5
• 2. Chatbot Modality in Health Communication 8
• 3. Belief Congruence and Communication Processes in Chatbot Interactions 14
• 4. Interaction Effects of Chatbot Modality and Belief Congruence 17
• 5. Credibility Evaluations and Vaccination Attitudes in Chatbot Interactions 21
• 6. Research Model 23
• Chapter 3. Methodology 26
• 1. Research Design 26
• 2. Participants 28
• 3. Procedure 29
• 4. Stimuli 31
• 1) Platform 31
• 2) Content 32
• 3) Interface 33
• 4) Exposure 34
• 5) Logging 34
• 5. Measurement 35
• 1) Pre-Existing Beliefs Toward Vaccination 35
• 2) Chatbot Credibility 35
• 3) Message Credibility 36
• 4) Attitude Toward Vaccination 37
• 5) Attention and Manipulation Checks 37
• Chapter 4. Results 39
• 1. Effects of Chatbot Modality and Belief Congruence on Attitude Toward Vaccination 40
• 2. Effects of Chatbot Modality and Belief Congruence on Chatbot Credibility 41
• 3. Effects of Chatbot Modality and Belief Congruence on Message Credibility 41
• Chapter 5. Discussion 45
• Chapter 6. Implications 50
• 1. Theoretical Implications 50
• 2. Practical Implications 51
• Chapter 7. Limitations and Future Research 52
• References 54
• Appendix 63
• <Appendix 1> Stimulus Materials 63
• <Appendix 2> Chatbot Intents 65
• <Appendix 3> Survey Materials 67
• Korean Abstract 81