Human-AI Interaction

 Conversational user interfaces (CUIs) have existed since the 1960s, but recent AI advancements have brought them back into everyday life. Conversation is the most natural way humans interact, making CUIs a powerful tool for emotional and informational support. However, current research often lacks a holistic approach, focusing only on isolated aspects like technical development or ethics. We take a human-centered perspective, considering interface design, system development, and ethical implications together to create CUIs that align with human needs and values.

We aim to design conversational agents that are both harmless and helpful, supporting users emotionally and informationally. With expertise in AI, human-AI interaction, and HCI, we explore key aspects of CUI design. Our research includes analyzing AI-induced harms, such as sexual harassment complaints in Replika, where user reviews revealed serious ethical concerns. In another study, we worked with teens to understand how they would design a chatbot for online safety. These insights provide a foundation for creating safer, more supportive conversational agents while guiding future research on ethical AI development.

AI ethics faces a significant gap between the rapid deployment of AI tools and the ethical concerns they raise, particularly in generative AI. While the field is evolving, interdisciplinary collaboration is needed to address the challenge of incentivizing organizations to prioritize ethics before deploying large-scale AI systems. Users currently lack control and transparency regarding how their data is used, which is a key issue requiring urgent attention.

Our research, grounded in computer science, mathematics, and ethical philosophy, focuses on making AI ethics more relevant to computer and data scientists rather than just ethicists. We have found that many prioritize computing power and speed over ethical considerations. By identifying ways to shift this preference, we aim to drive structural changes that embed ethical awareness into AI development and deployment. 

Role of AI in Youth Online Social Support: In our studies we wanted to understand young adults’ perceptions support provided by various common sources (AI, therapist, adult mentor, and peer) to four help-seeking messages on topics that youth frequently seek support about (relationships, suicidal thoughts, self-expression, and physical health). We provided insights into how young people perceive AI-generated responses in peer support situations. We found that young people preferred AI-generated responses about less sensitive topics (relationships, self-expression, and physical health). For more sensitive discussions (suicidal thoughts), participants preferred responses from the adult mentor.