Research

The rapid expansion of AI technologies has led to diverse AI agents. However, AI performance has traditionally been evaluated based on systemic criteria, overlooking users' subjective experiences, particularly those shaped by emotional intelligence. Drawing on the Theory of Reasoned Action (TRA), existing research has focused on technological aspects of AI adoption, including models like the Technology Acceptance Model (TAM) and the Unified Theory of Acceptance and Use of Technology (UTAUT). Recognizing the need for a more nuanced approach, I applied semantic differential scales, which consider both cognitive and affective dimensions of human attitudes. In addition to Osgood et al.'s foundational dimensions—evaluation, potency, and activity (EPA)—I incorporated two additional models: Moshkina's and Bartneck et al.'s. Exploratory factor analysis (EFA) was employed to test the validity and reliability of the measurement scales. The result revealed six key factors that shape user attitudes toward AI agents: intelligence, activity, expressiveness, usability, mood, and naturalness. This research offers both theoretical and practical contributions. It presents a comprehensive tool for investigating user attitudes toward AI agents and provides new insights into how AI agents can be evaluated from a human-centered perspective. 

*Published in the International Journal of Social Robotics in April 2023

“Trusting an Intelligent Virtual Assistant”

(Advisor: Prof. Sunkyong Lee)

This study investigated how task types (functional vs. social) and gendered voices (female vs. male) of Siri, an intelligent virtual assistant (IVA), impact users' perceptions of social presence and trust. This research was built upon prior U.S.-based studies that examined gender and task effects on trust in Siri. My role involved translating English survey questionnaires into Korean, ensuring cultural and linguistic relevance. In an online experiment, we randomly assigned individuals to one of four conditions, each interacting with Siri for specific tasks. Multivariate analyses revealed significant differences in trust based on task type, with higher trust levels for functional tasks. While social tasks are designed to enrich user experience through entertainment and casual interaction, they do not necessarily increase trust or social presence compared to functional tasks when interacting with Siri. However, unlike previous research, Siri's gendered voice had no significant effect on trust, which implies that the gender effect overall was much more complex and nuanced. Published in Behaviour & Information Technology in January 2024, the study replicated findings on task effects from prior research. This study emphasizes the need for replication in HMC research to establish external validity and understand cultural differences in user interactions with IVAs.

*Published in Behaviour & Information Technology in January 2024

“Consumer Account Sharing Journey for OTT Services”

(Advisor: Prof. Yoonhyuk Jung)

Smart Media Service Research Center (SSRC) 

This project aimed to deepen our understanding of consumer behavior in Over-The-Top (OTT) services, focusing on account sharing, where users split subscription fees with family or friends. While previous studies explored the attributes and privacy concerns of account sharing, little attention has been given to the consumer decision journey. Using the consumer decision journey model (Court et al., 2009), my colleagues and I conducted interviews and surveys based on social representation theory to analyze decision-making from initial consideration to service adoption. Since the consumer decision journey model highlights how consumers don't follow a linear path but cycle through phases of consideration, evaluation, purchase, and post-purchase, it seemed plausible to explain the complex account sharing processing: how users revisit the decision to share accounts to reduce costs and mitigate privacy concerns, particularly in the post-purchase phase where users continuously evaluate their subscription experience and sharing benefits. Our findings revealed a strong path dependency, with users typically sharing accounts with family members to reduce costs and address privacy concerns. These insights led to the development of an account-sharing journey model, expanding the traditional customer journey framework. This research earned the Excellence Prize in the 2021 Student-Led Creative Research Competition by the Smart Media Service Research Center (SSRC), highlighting evolving consumer practices within OTT services.

*Awarded Excellence Prize in 2021 Student-led Creative Research Presentation

“User Perception on Hate Music”

(Advisor: Prof. Yoonhyuk Jung)

This research was inspired by a news article criticizing a music platform for recommending an artist's album despite their involvement in a hate crime. As digital music grows, "Hate music," which contains discriminatory or hateful content, has become a social issue. Using the Elaboration Likelihood Model (ELM), we categorized hate music into two types: "hate content" (lyrics promoting hate or violence) and "hateful conduct" (artists involved in hate crimes like murder or sexual offenses). Hate content was considered a central route, reflecting the message's substance, while hateful conduct was treated as a peripheral route, related to the artist's actions. We aimed to compare users' perceptions of these two categories. Through surveys and statistical analysis (Paired t-Test, Unpaired t-Test, One-Way ANOVA), we found that users viewed hate content more negatively than hateful conduct, aligning with ELM's framework of message quality versus source reliability. Notably, younger users (teens) were less sensitive to both forms of hate music. Additionally, users with lower self-efficacy perceived hate content as more harmful than hateful conduct. This study offers valuable insights for managing hate music on digital platforms, highlighting its societal impact on promoting prejudice and discrimination. 

*Published in Information Society & Media in November 2021