Three essays about healthcare education, depression treatment, and appearance enhancement

Abstract

Social media platforms are becoming increasingly prevalent as health information sources, serving individuals seeking advice and healthcare providers aiming to disseminate knowledge. For example, doctors are using short-video platforms (SVP), such as Douyin and Kuaishou, to provide health education and promote behavioral changes related to public health. However, the misuse of social media may harm users’ well-being. For instance, artificial intelligence (AI) beauty filters on social media may negatively affect users’ mental health by triggering eating disorders and body dysmorphic disorder, potentially leading to depression. On the other hand, AI has great potential to treat depression by offering patient communication, diagnosis, decision-making support, and behavior change interventions. This thesis comprises three essays, each examining a distinct context: audience engagement with health education videos provided by doctors on SVPs, the effect of AI beauty filters on users’ well-being and observers’ evaluations of beauty-filtered selfies and users, and the adoption of AI healthcare agents for depression treatment by depressed patients.

The first essay of this thesis focuses on audience engagement with health education videos on SVPs. Although using SVPs to deliver health education has increased, doctors face a dilemma when creating health education videos. SVPs are primarily used for leisure by the public, which contrasts with the serious nature of health education and poses a dilemma for doctors regarding the appropriate language for education to enhance audience engagement. Using professional medical terminologies may help maintain doctors’ medical competence, but it may suggest that doctors put limited cognitive effort into making the content understandable for laypersons. Conversely, layperson language may enhance audiences’ perceptions of doctors’ cognitive effort, but it risks compromising doctors’ medical competence. Drawing upon the communication accommodation theory (CAT), we develop a research model to understand how, why, and under what conditions medical terminology use affects audience engagement. Results from two experiments and an archival analysis show that medical terminology use negatively affects audience engagement by reducing audiences’ perceived doctors’ cognitive effort. Contrary to our hypothesis, medical terminology use can decrease perceived doctors’ medical competence, which could be explained by the mediating effects of perceived doctors’ low language competence. Doctors’ happy facial emotions exacerbate the negative effect of medical terminology use on audience engagement, whereas serious facial emotions mitigate this negative effect. This study contributes to research on health information systems and CAT research, offering practical implications for doctors and SVP managers.

The second essay of this thesis examines the effect of AI beauty filter usage on the well-being of focal users and how AI beauty filter usage influences observers’ perceptions of users’ morality and authenticity. Previous research on AI beauty filter usage has focused on users’ motivations for editing selfies, considering demographic factors like gender and age and addressing self-presentation and impression management needs. However, there has been limited investigation into the social norms influencing beauty filter usage, and the effects of such usage on focal users remain unclear, particularly regarding who may be more susceptible to positive or negative outcomes. This study develops a research model grounded in self-discrepancy theory to fill these gaps. Survey findings indicate that social norms positively affect AI beauty filter usage frequency. This subsequently leads to more negative moods through perceived self-discrepancy and heightened self-confidence via appearance self-esteem. The positive relationship between usage frequency and perceived self-discrepancy intensifies as the depth of AI beauty filter usage grows. On the other hand, users may achieve impression management purposes by using AI beauty filters. However, it remains unclear how social media observers assess beauty-filtered selfies and how those assessments influence their evaluations of user traits. This study addresses these gaps by drawing on the halo effect and perspective-taking literature. Experimental results show that perceived AI beauty filter usage decreases observers’ perceived selfie authenticity. However, observers who are reminded of their own AI beauty filter usage tend to perceive the selfie as authentic. Perceived selfie authenticity positively affects perceived user authenticity and morality. This essay contributes theoretical insights to self-discrepancy theory and the literature on beauty filters, perspective-taking, and the halo effect. This study offers practical implications for beauty filter users, social media platforms, and health policymakers.

The third essay of this thesis explores how to leverage AI to treat depression by reducing depressed patients’ perceived judgment and simultaneously increasing their perceived accountability. A primary reason for the treatment disparity of depression is the stigma associated with depression, which encompasses negative beliefs, attitudes, and conceptions about depression. Depressed patients may avoid or discontinue treatments due to fear of discrimination from human physicians. AI helps treat depression by minimizing interpersonal interactions, thereby reducing patients’ perceived judgment. However, standalone AI may lack accountability. We propose AI-human hybrids, which integrate AI and human physicians to treat depression. Patients interact directly with AI in AI-human hybrid scenarios, potentially reducing perceived judgment. The AI gathers personal data, diagnoses diseases, and generates treatment plans for human physicians to review and confirm, potentially increasing perceived accountability. Using a trust theory framework, we develop a research model explaining how healthcare agents (i.e., human physicians, standalone AI, and AI-human hybrids) differently affect perceived judgment and accountability, subsequently influencing patient trust. Experimental results show that patients perceive AI-human hybrids as having less judgment than human physicians but more accountability than standalone AI and human physicians. The difference in perceived judgment is more significant among highly depressed patients. Surprisingly, the difference in perceived accountability is less pronounced for them. Perceived judgment and accountability affect patient trust in healthcare agents, with perceived judgment being a more potent predictor. This study contributes to health information systems and trust literature, offering practical implications for depressed patients and healthcare providers.

Date of Award	13 Aug 2025
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Kai H. LIM (Supervisor) & Jingjun David XU (Supervisor)