AI As Tool/Subject: Methodological Challenges In ML and LLM Social Research

This session examines how Machine Learning (ML), Large Language Models (LLMs), and related AI systems are reshaping social-scientific research. These tools are increasingly used for many analytical tasks like text annotation, qualitative coding, discourse analysis, summarisation, classification and synthetic data generation. At the same time, they raise important methodological, epistemological, and ethical questions about validity, transparency, reproducibility, and the delegation of interpretive authority, ever so relevant in respect of the increasing adoption and usage of AI systems in general.
We invite contributions that explore how ML and LLMs can be integrated into quantitative, qualitative, and mixed-methods research designs to support and innovate social research. Relevant topics include the use of AI to analyse unstructured data, support different aspects of scientific discovery, and assist with interpretive workflows or the modelling of complex phenomena.
The session also welcomes work that treats AI systems themselves as objects of empirical investigation. This includes model inspection methods such as auditing, probing, prompt experiments, scenario-based testing, bias and fairness evaluation, comparisons across models, model versions, and human subjects or annotators, and analyses of how AI systems respond to social identities, political frames and cultural contexts.
By bringing together empirical, methodological, and critical perspectives, the session aims to reflect on the role of ML and LLMs in research and the social world, promote interdisciplinary dialogue, and advance methodological innovation with and through emerging technologies.