This study aims to understand how students co-construct understanding of artificial intelligence (AI) through collaborative argumentation. A dual-dimensional Epistemic Network Analysis was conducted with data from seven university students, to capture the interconnected relationship between structure and conceptual understanding in an evolving argumentative discourse. It was revealed that participants tended to describe AI performance with reference to human performance, and their actions to seek supporting evidence for their ideas played an important role in connecting phenomenal understanding with evaluative understanding. It was also found that participants conducted more self-challenge as a community rather than inter-personal challenge to advance knowledge construction.