WOCHAT: Workshop on Chatbots and Agentic Technologies

Motivation

Conversational AI is undergoing a structural transformation. Dialogue systems are evolving from reactive turn-level responders into agentic systems capable of planning, reasoning, collaborating, and adapting over extended interactions. This transition raises foundational research questions that remain underexplored within dialogue research.

WOCHAT 2026 is motivated by three critical gaps:

• Agentic Dialogue Modeling: While large language models enable fluent responses, goal-driven planning, long-term coherence, and task-level autonomy remain open research challenges. Dialogue systems must integrate decision-making, memory, and structured planning while maintaining conversational naturalness.

• Multi-Agent and Multimodal Interaction: Human communication is multimodal and collaborative. Dialogue research must address how agents integrate linguistic, visual, and contextual signals, and how multiple agents coordinate, negotiate, and share information within conversational settings. The intersection of multi-agent systems and dialogue modeling represents a fertile but fragmented research space.

• Evaluation Beyond Surface Fluency: Traditional automatic metrics inadequately capture higher-order dialogue qualities such as reasoning, commonsense grounding, emotional nuance, and theory-of-mind. The community lacks robust, reproducible evaluation methodologies for agentic and socially aware conversational systems. Addressing this evaluation gap is essential for scientific progress and responsible deployment. 

Additionally, real-world applications in finance, legal reasoning, cultural heritage, and cybersecurity expose the limitations of current systems in terms of robustness, explainability, and trustworthiness. These domains provide stress tests that surface core methodological challenges in dialogue research.

By focusing on these issues, WOCHAT 2026 directly aligns with SIGDIAL’s mission to advance the theory and practice of discourse and dialogue systems. The workshop emphasizes methodological rigor, reproducibility, and cross-disciplinary integration between dialogue modeling, multi-agent coordination, affective computing, and evaluation science.