Detta forskningsseminarium fokuserar på metodologiska tillvägagångssätt inom samhällsvetenskapen, med särskild tonvikt på metoder som används mindre ofta eller som sträcker sig över ämnesgränserna. Serien är öppen för alla som som är intresserade av att veta mer om detta eller dela sina kunskaper på området.
"When Is Noise a Surprise? Abductive Inference in Computational Analysis"
Dr Elena Bogdanova, Associate Professor in the Department of Sociology and Work Science.
In this seminar, Elena will explore the role of surprise in computational research and its implications for abductive inference and theorising.
Drawing on an ongoing study of how sustainable investment is defined and constructed in the annual and sustainability reports of Swedish occupational pension providers between 2015 and 2025, she will discuss her use of BERTopic to move beyond theoretical categories established in earlier research. The analysis produced an unexpected finding: some of the passages most explicitly concerned with sustainability were classified as noise rather than forming stable topics.
Is this computational ‘noise’ telling us something theoretically interesting about how sustainable investment is constituted, or does it instead reflect the reporting genre, the composition of the corpus, or modelling choices?
Using this empirical puzzle as a starting point, the seminar will consider when unexpected computational results can become theoretically productive surprises, and what this means for abductive inference at a time when machine learning and generative AI are increasingly involved in identifying and summarising patterns in qualitative data.
The seminar should be of interest to anyone working with computational methods, AI and research methods, qualitative or mixed-methods research, or questions of theory development and inference.