Adapting Dialog and Pragmatic Traits in Context ADAPTIC

About the Workshop

With the development of LLMs, the possibilities in designing human-like CUIs has been greatly improved. Consequently, CUIs are increasingly adopting human-like pragmatic cues — such as tone of voice, backchanneling, social or emotional expressions, and conversational repair. However, these cues are often blindly mimicked by LLMs, rather than being strategically designed to serve a true communicative function. As LLMs are increasingly being utilized in CUIs across domains requiring divergent user expectations, visual and linguistic interface features provide an opportunity to strategically adapt the interaction for domain specific needs. How can we move beyond the mere illusion of human-like pragmatic cues, and embed these as interface features to design CUIs that actively support mutual understanding and serve user needs in their specific domain context?

This half-day, in-person workshop will kick off this interdisciplinary discourse by inviting researchers, designers, and practitioners to rethink pragmatic cues in CUI design across domains. We aim to bridge the gap between HCI, psycholinguistics, and the social sciences to explore how users appropriate conversational AI across various contexts, transitioning from open-domain models to specialized, domain-specific implementations.

To learn more about participation in the workshop, please refer to the Call for Participation.