Workshops/Tutorials
Workshops and tutorials will be held on Friday, September 11.
Persuasion in Conversational Agents: Research Agendas, Tools, and Frameworks
Contact: Julija Vaitonyte, J.Vaitonyte@tilburguniversity.edu
This workshop will combine an invited keynote with roundtable discussions to identify key challenges, research directions, and methodological building blocks for studying persuasion in conversational agents emerging through social interaction. Prior to the workshop, a call for two-page extended abstracts will invite participants to share their perspectives and contribute to shaping the workshop’s topics and discussion points.
Discussions during the workshop will be organized around key ideas and topics, including those stemming from the submissions. The workshop aims to co-construct a research agenda by integrating asynchronous contributions with in-person discussions.
In particular, the workshop will build upon questions such as how persuasion in conversational agents can be studied beyond short-term opinion shifts, how concepts such as manipulation, deception, and “dark patterns” can be operationalized, how researchers can study long-term interaction effects, and what subjective and objective measures are needed to study persuasion in conversational agents.
The workshop is intended for researchers at all career stages, from early-career to established researchers, working across fields including socially interactive agent design and evaluation, machine learning, AI safety, cognitive psychology, neuroscience, sociology, and business/marketing. The workshop will also foster discussions that may lead to research collaborations, including co-authoring a post-workshop journal paper.
Running Virtual Human Experiments in the LLM Era with the IAGO Research Platform
Contact: Johnathan Mell, johnathan.mell@ucf.edu
This half-day tutorial introduces IAGO as a modern research platform for authoring and running controlled virtual human experiments in the LLM era. The central motivation is that current researchers face an uncomfortable tradeoff: traditional scripted systems often provide strong experimental control but limited flexibility, while fully generative LLM-based agents offer greater openness at the cost of reproducibility, controllability, and hallucination risk.
IAGO is designed to occupy a productive middle ground. This focus is closely aligned with IVA 2026’s emphasis on social cognition for human-AI cooperation, since the tutorial centers on how virtual humans can be designed and studied as socially interactive partners in controlled yet realistic settings. It is also well matched to IVA’s broader interest in dialogue, social behavior, behavioral experimentation, and working virtual-agent applications.
The objectives of the tutorial are to present the methodological challenge of building interactive virtual humans that are flexible enough to support realistic interaction but controlled enough for rigorous experimentation; introduce IAGO as a browser-based, increasingly no-code platform for authoring and deploying virtual human experiments; demonstrate how experimenters can configure agents along strategic, emotional, and dialogue dimensions; show how experimenter-authored dialogue pools, LLM-based selection, and selectively enabled generation can support controlled openness in social interaction; discuss how AI-assisted tooling can lower barriers to creating animation pipelines and embodied virtual human experiences; and help attendees identify how such a platform can support IVA-relevant work.
Overall, the tutorial is intended to show how virtual human experimentation can remain scientifically useful and operationally practical in a period increasingly dominated by large generative models.