Verbal conversations are central to sharing information and building shared understanding across diverse perspectives. Yet, they often lack structure and require effort to establish common ground. My research aims to make conversations, especially among creative professionals, fruitful and actionable by leveraging Gen AI and insights from facilitation practices. My prior work focused on identifying key dimensions of AI support for verbal conversations using facilitation cues, including timing, attention, and modality. I conducted empirical studies on how AI-generated cues and conversation modality impact creative production in group conversations, informing design considerations for future facilitation systems. My final dissertation project explores how extending facilitation beyond the conversation can support how insights are extracted, coordinated, and advanced by multiple collaborators.

What is the potential value and role for AI to facilitate real-time creative discussions? The paper explores principles for Generative-AI based conversational support by investigating how humans – playing the role of an AI agent – generate contextual conversational cues to guide an ideation session. We studied n=42 people (14 triads) brainstorming through a remote meeting design probe that allows a wizard facilitator to oversee the ideation and send text-based cues that appear real-time in the ideator interface. Thematic analysis of conversations, cues and post-hoc reflections by facilitators uncovered focal points, strategies and challenges. Notably, 44% of the cues sent out by the facilitators were either dismissed or ignored because they did not notice the cue update. When ideators did notice cues, certain facilitator strategies impacted the conversation more than others. Based on our analysis, we present design opportunities to improve generative AI-based systems to better support real-time creative collaborations.

Large language models (LLMs) have recently soared in popularity due to their ease of access and the unprecedented ability to synthesize text responses to diverse user questions. However, LLMs like ChatGPT present significant limitations in supporting complex information tasks due to the insufficient affordances of the text-based medium and linear conversational structure. Through a formative study with ten participants, we found that LLM interfaces often present long-winded responses, making it difficult for people to quickly comprehend and interact flexibly with various pieces of information, particularly during more complex tasks. We present Graphologue, an interactive system that converts text-based responses from LLMs into graphical diagrams to facilitate information-seeking and question-answering tasks. Graphologue employs novel prompting strategies and interface designs to extract entities and relationships from LLM responses and constructs node-link diagrams in real-time. Further, users can interact with the diagrams to flexibly adjust the graphical presentation and to submit context-specific prompts to obtain more information. Utilizing diagrams, Graphologue enables graphical, non-linear dialogues between humans and LLMs, facilitating information exploration, organization, and comprehension.

This article highlights several game design choices made during the creation of a browser-based game on mitigation strategies for Covid-19. Additionally, it presents a within group comparison of learning gains and self-reported behavioral changes after playing the game. Results show that the short COVIDCampus game has the potential to change college-age players' Covid-19 related mitigation behaviors and it significantly increased players' confidence in asking important health-related questions (Cohen's d=.27). Some implications are discussed.