News

Abstract:

Collecting and analyzing qualitative data is an integral part of human-computer interaction (HCI) research. Qualitative data can provide deeper insight into research participants’ impressions of systems as well as the motivations behind their behaviors with and reactions to those systems. The insights gained through this information can be invaluable for forming hypotheses that lead to system improvements and increase understanding of how people interact with systems.  The last decade has seen increased interest for involving natural language processing (NLP) in the qualitative analysis (QA) process across multiple scientific fields. While NLP approaches for QA are certainly not perfect, they have been shown to enhance QA results. Despite the advantages offered by these techniques, papers within the HCI community rarely mention use of these approaches.  Additionally, involvement of NLP approaches in QA has typically been limited to organizing, visualizing, or annotating data.  This work endeavors to explore the extent to which large language models (LLMs) can be leveraged to generate themes or higher-level codes and relevant exposition for QA.