Research Lab Maria Becker CHAI Lab: Computational Humanities & AI

PI: Dr. Maria Becker

 

Our mission: Exploring the interface between Digital Humanities and Social Sciences, AI, NLP, and Quantitative Linguistics

At the CHAI Lab, we investigate a variety of research questions from the humanities and social sciences using computational approaches within an interdisciplinary project team, with a special focus on large language models and methods from (corpus) linguistics, computational linguistics, and natural language processing (NLP). We are particularly interested in:

• how people interact with AI systems (human–AI interaction) and the associated technical as well as socio-ethical challenges,
• how to develop AI- and NLP-based methods to address research questions in the humanities and social sciences—such as political science, sociology, and economics,
• how to apply AI- and NLP-based techniques to analyze and optimize linguistic data in the life sciences, for example in medical communication, and
• how to combine AI- and NLP-based methods with linguistic approaches for the study of textual data and linguistically motivated research questions — for instance, how quantitative-statistical methods can be used to evaluate fine-grained annotations, how complex speech-act patterns in heterogeneous text types can be detected and analyzed automatically, and what opportunities AI models offer in this context.

In CHAI Lab we work on several projects in the areas of medical communication and science communication, in which Maria Becker serves as lead investigator:

The Thematic Research Network Knowledge in Context: Language and Thought in the Natural and Life Sciences is a research consortium funded under the Excellence Strategy of the German federal and state governments, in collaboration between Mannheim University Medical Center and the Departments of Computer Science, German Studies, and Translation Studies at Heidelberg University. In this project, we investigate the transfer of medical knowledge to lay audiences and the role that varying prior knowledge plays. Our goal is to use computational methods to identify linguistic factors that affect the comprehensibility of medical texts depending on patients’ background knowledge, language proficiency, and socio-cultural contexts.

The BMBF-funded project Moralizations in Science Communication (MoWiKo) is a cooperation between Heidelberg University and the Karlsruhe Institute of Technology. In an interdisciplinary team of (computational) linguists, communication scientists, and ethicists, we analyze the forms and effects of moralising discourse in a variety of science-popularising genres using both automated and qualitative methods. Our focus is on debates around AI, energy security, and food supply.

The project Annotation and Automatic Analysis of Moralizing Practices in Diverse Knowledge Domains (AMora), also funded under the Excellence Strategy, involves the creation of datasets drawn from different text types and knowledge fields in which moralising speech acts and their linguistic features are annotated both manually and automatically. We then use these datasets to develop computational models for the automated detection and analysis of moralising strategies in heterogeneous texts.