We organized, together with Hendrik Buschmeier and Sina Zarrieß, the 29th Workshop on the Semantics and Pragmatics of Dialogue (SemDial), this year called “Bialogue”. More than 50 researchers from Computer Science, Linguistics, Psychology and other disciplines met at CITEC to discuss latest findings and advancements in conversational interaction, dialogue and language technology.
Amelie and Dominik gave oral presentations at the German Cognitive Science Conference (KogWis 2025), held in Bochum. Presenting work carried out in TRR 318 (in subproject A01 and C05, respectively), Amelie talked about using cognitive partner models for adaptive explanation generation, while Dominik described how the, possibly biased, diagnostic reasoning of medical experts can be modeled computationally.
Stefan gave a keynote at the Workshop on Interactive Task Learning in Human-Robot Co-Construction (ITL4HRI) at IEEE RO-MAN in Eindhoven (NL), speaking about the cognitive and communicative mechanisms of co-constructive task learning in human-robot interaction.
Our comprehensive paper on how to make human-agent collaboration in dynamically changing environments sufficiently flexible and adaptive (i.e., fluid) just got published in Frontiers in Robotics & AI. We describe dynamic collaboration patterns observed in humans, discuss how those can be transferred to human-agent teams, and describe what models of theory of mind reasoning must and can be implemented in AI agents to that end.
Can artificial intelligence work side by side with doctors to help them make better diagnoses? An article features our research dedicated to answering this question. It explains how we are working on an interactive AI system that accompanies doctors as they make a medical diagnosis by reviewing and evaluating assumptions in dialog with them.
Florian went to the Int. Conference on Hybrid Human-AI Interaction (HHAI 2025), held in Pisa (Italy) and presented work on collaborative AI agents that combine task intelligence (perception, planning, execution) with socio-cooperative intelligence (theory of mind, multi-agent collaboration, language-based communication) to enable fluid human-AI teams.
Contributing to the workshop “Social Robots in the Public Sector”, organized by the RuhrBots project at HSPV NRW in Gelsenkirchen, with a talk on how social robots can act as socially intelligent partners and peers in early language learning by pre-school children.
Fabian has contributed to the AAAI 2025 Spring Symposium on “Current and Future Varieties of Human-AI Collaboration” in San Francisco by presenting our work on studying dynamic, adaptive human-AI collaboration and teamwork in the game environment “Cooperative Cuisine” (extended version of “Overcooked!”).
We opened our lab and set up research demos and showcases — our Family & Friends Day took place on Jan 24th, and a lot of family member, friends, colleagues, and alumni showed up! The day was full of fun conversations, interesting interactions between humans and machines, and hearty getting-together moments. Visitors could interactively collaborate with AI agents in the Cooperative Cuisine environment, talk to the embodied conversational agent MAVERICK, learn about a board game through adaptive explanations generated by the SNAPE system, or feel like a physician working together with the AI model ASCODI to solve medical diagnosis problems. On top of it there was so much more, especially food, fun and socializing 🙂
Can AI help improving the diagnostic reasoning of physicians to make them arrive at diagnoses that are better justifiable? We present “ASCODI”, an XAI-based interactive reasoning support system for this, in a paper at the MAI-XAI Workshop at the ECAI 2024 conference. This is work done in Project C05 of the TRR 318 Constructing Explainability.