Conversational AI (such as chatbots, voice assistants and embodied agents) are becoming increasingly prevalent and capable. For example, you could talk to a chatbot to help you with personal health goals, or talk to an embodied agent in a computer game to provide a more immersive experience. (In addition to producing more accurate and efficient agents), a goal of practitioners has been to create more natural, engaging, and valuable interactions between humans and conversational AI. However, it is not always clear how best to design conversational AI to engage with users given both the myriad of social cues available and potential differences in beliefs and perceptions surrounding interactions with AI compared to interactions with humans.
Motivated by these challenges, I will discuss several of our recent human-computer interaction (HCI) studies that investigate features that can make conversational AI more engaging. Specifically, we conducted empirical studies to investigate the effect of changing social cues in healthcare-related chatbots and its impact on user engagement, and explored user perceptions of interacting with large language model (LLM) driven agents in a commercially available video game. From this we hope to provide insight into design choices for conversational AI as well as spark discussion.
This post was originally published on 3rd party site mentioned in the title of this site