LLM-powered conversational agents are increasingly influencing our decision-making, raising concerns about “sycophancy”–the tendency for LLMs to excessively agree with users even at the expense of truthfulness. While prior work has primarily examined LLM sycophancy as a model behavior, our understanding of how users perceive this phenomenon and its impact on user trust remains significantly lacking. In this work, we conceptualize LLM sycophancy along two key constructs: conversational demeanor (complimentary vs. neutral) and stance adaptation (adaptive vs. consistent). A 2 × 2 between-subjects experiment (N = 224) revealed complex dynamics: complimentary LLMs that adapted their stance reduced perceived authenticity and trust, while neutral LLMs that adapted enhanced both, suggesting a pathway for manipulating users into over-trusting LLMs beyond their actual capabilities. Our findings advance user-centric understanding of LLM sycophancy and provide profound implications for developing more ethical and trustworthy LLM systems.

With the development of natural language processing (NLP), open-domain chatbots can operate as companions. The success of open-domain chatbots depends on understanding how positive and negative expectancy violations affect both utilitarian and hedonic gratification, which in turn influences user satisfaction and continued use. Therefore, this study examines how violations of expectations influence the hedonic and utilitarian evaluations of open-domain chatbots using the expectancy-violation theory. Participants (n = 204) interacted with an improvising AI chatbot and reported their satisfaction. Results indicated that both hedonic and utilitarian satisfaction were higher when negative expectations were met compared to when expectations were violated. However, no significant difference was found in hedonic gratification between the chatbot’s expected performance and positively unexpected performance. Conversations meeting expectations elicited more utilitarian satisfaction than positively surprising interactions. These findings highlight the dynamics between value types and expectation violations in AI chatbot evaluations.

In e-commerce, Augmented Reality (AR) employs computer vision and artificial intelligence (AI) techniques to enhance the shopping experience through personalized recommendations based on users’ physical data. However, concerns regarding privacy and perceived intrusiveness can undermine the persuasive appeal of personalized AR for e-commerce experiences. Can overtly communicating information transparency mitigate such concerns? Two between-subjects online experiments were conducted, revealing that consumers were drawn to the AR-powered e-commerce site due to its heightened perceived immersion and usefulness but also considered it more intrusive than the non-AR site. However, when both websites were AR-powered, the one with high information transparency significantly enhanced perceived privacy protection, reducing perceived intrusiveness. Perceived transparency positively mediated the effects, particularly among individuals with lower pre-existing trust in algorithms. These findings have implications for theory pertaining to the use of AR in strategic communications and the design of information transparency for AR-powered e-commerce.

Propelled by their remarkable capabilities to generate novel and engaging content, Generative Artificial Intelligence (GenAI) technologies are disrupting traditional workflows in many industries. While prior research has examined GenAI from a techno-centric perspective, there is still a lack of understanding about how users perceive and utilize GenAI in real-world scenarios. To bridge this gap, we conducted semi-structured interviews with (N = 18) GenAI users in creative industries, investigating the human-GenAI co-creation process within a holistic LUA (Learning, Using and Assessing) framework. Our study uncovered an intriguingly complex landscape: Prospects – GenAI greatly fosters the co-creation between human expertise and GenAI capabilities, profoundly transforming creative workflows; Challenges – Meanwhile, users face substantial uncertainties and complexities arising from resource availability, tool usability, and regulatory compliance; Strategies – In response, users actively devise various strategies to overcome many of such challenges. Our study reveals key implications for the design of future GenAI tools.

Recommender systems (RS) have become increasingly vital for guiding health actions. While traditional systems filter content based on either demographics, personal history of activities, or preferences of other users, newer systems use social media information to personalize recommendations, based either on the users’ own activities and/or those of their friends on social media platforms. However, we do not know if these approaches differ in their persuasiveness. To find out, we conducted a user study of a fitness plan recommender system (N = 341), wherein participants were randomly assigned to one of six personalization approaches, with half of them given a choice to switch to a different approach. Data revealed that social media-based personalization threatens users’ identity and increases privacy concerns. Users prefer personalized health recommendations based on their own preferences. Choice enhances trust by providing users with a greater sense of agency and lowering their privacy concerns. These findings provide design implications for RS, especially in the preventive health domain.

Automated machine learning (AutoML) is envisioned to make ML techniques accessible to ordinary users. Recent work has investigated the role of humans in enhancing AutoML functionality throughout a standard ML workflow. However, it is also critical to understand how users adopt existing AutoML solutions in complex, real-world settings from a holistic perspective. To fill this gap, this study conducted semi-structured interviews of AutoML users (N = 19) focusing on understanding (1) the limitations of AutoML encountered by users in their real-world practices, (2) the strategies users adopt to cope with such limitations, and (3) how the limitations and workarounds impact their use of AutoML. Our findings reveal that users actively exercise user agency to overcome three major challenges arising from customizability, transparency, and privacy. Furthermore, users make cautious decisions about whether and how to apply AutoML on a case-by-case basis. Finally, we derive design implications for developing future AutoML solutions.

The outbreak of COVID-19 forced schools to swiftly transition from in-person classes to online or remote offerings, making educators and learners alike rely on online videoconferencing platforms. Platforms like Zoom offer audio-visual channels of communication and include features that are designed to approximate the classroom experience. However, it is not clear how students' learning experiences are affected by affordances of the videoconferencing platforms or what underlying factors explain the differential effects of these affordances on class experiences of engagement, interaction, and satisfaction. In order to find out, we conducted two online survey studies: Study 1 (N = 176) investigated the effects of three types of videoconferencing affordances (i.e., modality, interactivity, and agency affordances) on class experience during the first two months after the transition to online learning. Results showed that usage of the three kinds of affordances was positively correlated with students' class engagement, interaction, and satisfaction. Perceived anonymity, nonverbal cues, and comfort level were found to be key mediators. In addition, students' usage of video cameras in class was influenced by their classmates. Study 2 (N = 256) tested the proposed relationships at a later stage of the pandemic and found similar results, thus serving as a constructive replication. This paper focuses on reporting the results of Study 1 since it captures the timely reactions from students when they first went online, and the second study plays a supplementary role in verifying Study 1 and thereby extending its external validity. Together, the two studies provide insights for instructors on how to leverage different videoconferencing affordances to enhance the virtual learning experience. Design implications for digital tools in online education are also discussed.

Does assembling one's robot enhance the quality of our interaction with it? And, does it matter whether the robot is a utilitarian tool or a socially interactive entity? We examined these questions with a 2 (Assembler: Self vs. Others) × 2 (Expectation Setting/Framing: Task-oriented robot vs. Interaction-oriented robot) between-subjects experiment (N = 80), in which participants interacted with a humanoid desktop robot (KT-Gladiator 19). Results showed that participants tended to have more positive evaluations of both the robot and the interaction process when they set it up themselves, an effect that is positively mediated by a sense of ownership and a sense of accomplishment, and negatively mediated by perceived process costs of setting up the robot. They also tended to evaluate the robot and the interaction more positively when they expected it to be task-oriented rather than interaction-oriented. Implications for theory and design of robots are discussed.