SFB 673 - "Alignment in Communication"

B6 - Understanding Alignment from Misalignment

Project B6 investigates the cognitive components and processes underlying alignment in sequential verbal behaviour. The project aims to address methodological challenges in previous studies by focusing on break-downs of alignment (misalignment) due to impaired cognitive functioning, closely interlinked with an approach of model generation and validation in robotics. We will specify relevant cognitive components and their contribution to alignment phenomena at different linguistic levels (syntactic and lexico-semantic).

Even though it is widely accepted that communicative competences and conversation skills are based on an interplay of cognitive processes, a full understanding of the cognitive basis is still lacking. Empirical work has identified cognitive functions, which on the one hand contribute to successful communicative behaviour and on the other hand appear to be causally related to break-downs of alignment (misalignment), as has been observed in certain brain-injured populations. Here, a special thematic and methodological focus will be on sequential verbal behavior, either in unconstrained dialogues or in structured tasks performed by two participants. Analyses of the temporal characteristics of sequential verbal behavior will reveal direct effects of preceding moves and anticipated moves of the dialogue partners.

The investigations will be complemented by experiments using artificial intelligent systems (robots and AI systems in general), which exhibit behaviours that can be considered "impaired". By establishing commonalities between impaired humans and "impaired" artificial systems, we intend to contribute to the understanding of (mis-)alignment processes. These experiments will be informed by clinical studies and potential (clinical) causes for misalignment. We will, in particular, consider language games as an instrument to investigate sequential verbal communication in restricted settings with the goal of enhancing our understanding of misalignment at the lexico-semantic level.