Interfaces: Implementing user-centered ComAI

• Applicant: Professor Dr. Rainer Malaka
• Subject Area: Image and Language Processing, Computer Graphics and Visualisation, Human Computer Interaction, Ubiquitous and Wearable Computing ; Communication Sciences
• Term: since 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 516511468

With the rise of Large Language Models (LLMs) and their adaptation to human-computer communication based on human feedback, users are increasingly expecting human-like interaction from ComAIs. However, these models designed specifically for human-computer communication face mainly two problems: First, they rely only on the data they are trained with, which is often biased and insufficient. Second, they are only created to produce text and responses, but there is no validation of the truthfulness of the generated output during the training. Therefore, when designing and building ComAIs, it is important that users know that such problems are a feature of LLMs. Interfaces are crucial in helping users to identify problematic information and evaluate the data sources’ quality and reliability. Against this background, the project uses the example of conversational bots to examine the design and implementation of ComAI interfaces as a dimension of their socio-material constitution. Investigating both the conversational and paralinguistic features of interfaces, we research which implementation features of the interface design influence user-alignment and how.

• DFG Programme: Research Units
• Subproject of: FOR 5656: Communicative AI: The Automation of Societal Communication