Toward a Complementary Use of Artificial Intelligence for Causal Analysis and Mapping: Experimental Approaches of Qualitative Causal Coding and Data Visualization

This novel study proposes an experimental prototype of using artificial intelligence (AI) for qualitative causal analysis and mapping, aimed at lowering person-hours, financial resources and language barriers, as well as improving generalizability and scalability in causal coding and data visualization. Utilizing three AI Large language models (LLM), this work attempted descriptive, causation, and pattern coding on Korean and English narratives and produced relevant directed acyclic graph visualizations. Benchmarking findings highlight the feasibility of complementary use of AI for scalable, resource-efficient, and language-inclusive causal inference in time, labor, and cost-constrained research environments within Korean social sciences. Results also show that while each model has distinct strengths, cross-model triangulation enhances methodological rigour, and data visualization can be achieved using AI tools; a comparative analysis with a validated human-coded case confirmed that LLM can replicate much of the causal structure, though complex loops require human-AI iteration. This study proposes LLM-assisted qualitative causal workflow model via humane-AI collaboration, ranging from traditional elemental coding, multi-level causal inference and non-linear causal mapping, while acknowledging limitations in prompt sensitivity, over-generalization, procedure standardization, and adherence to ethics guidelines.