Bio

Erzhen Hu is a fifth-year PhD candidate in the Department of Computer Science at the University of Virginia. Erzhen’s research lies at the intersection of Human-Computer Interaction (HCI), and Computer-Supported Cooperative Work (CSCW), focusing on designing and evaluating Human x AI real-time communication to improve collaborative work. Her work has been published in top HCI venues including CHI, CSCW, and UIST. Her research is supported by a Google PhD Fellowship, and she has interned at Microsoft Research, Google, and Autodesk Research.

Advancing Conversational Dynamics for Real-Time Human-Human and Human-AI Communication in Virtual Space

We spend much of our time in virtual spaces, meeting co-workers, attending conferences, and socializing with friends. As digital and blended environments become central to daily life, we need more flexible, expressive, and low-friction ways to connect across distance. Real-world communication is messy: we interrupt, overlap, whisper, form subgroups, and drift in and out of talk, continually renegotiating who is included or heard. Yet most virtual tools enforce rigid turn-taking, discrete rooms, and constrained awareness cues.

My research explores interactive systems designed to enable flexible group dynamics within real-time human–human and human–AI communication. This proposal introduces a framework for advancing multi-party communication along three dynamics: spatial (group formations), temporal (turn-taking), and authoring (human-AI dialogue design). We show a series of proposed systems that create effective tools to support this vision, including FluidMeet, which supports spatial dynamics with semi-permeable conversation boundaries to mirror how people drift among small groups; OpenMic, which explores temporal dynamics with a malleable “Virtual Floor” for reshaping video feeds and signaling turn intentions in real time; and DialogLab, which enables authoring dynamics by configuring and testing human–AI conversations, bridging live communication with generative scripting and simulation. Together, these contributions advance both the design of interactive systems and the empirical understanding of how humans experience multiparty communication with humans and AI, laying groundwork for more flexible conversational technologies.