<p>Problem-finding is a foundational process for creativity and inquiry-based learning, yet it remains difficult to observe and evaluate because its epistemic dynamics unfold internally and over time. While learning analytics and artificial intelligence offer new possibilities for capturing learning processes, few frameworks explicitly align instructional activities with systematic, minimally obtrusive data collection that renders problem-finding trajectories analytically visible. This study proposes and empirically examines a process-oriented analytic framework for capturing and analyzing learners’ problem-finding processes through multimodal trace data. The framework integrates a structured five-step problem-finding task with continuous capture of textual, visual, and temporal traces as analyzable process evidence. An AI-mediated inquiry activity is used as an illustrative case to instantiate perspective-extending support while simultaneously generating rich trace data. An exploratory study with 28 participants in higher education demonstrates the feasibility of the framework. Quantitative mixed-effects modeling revealed systematic developmental change from perceptual noticing to more conceptual reasoning across task steps, while qualitative functional coding traced shifts in epistemic framing. In addition, uptake-related variation was associated with subsequent differences in category diversity and abstractness, suggesting that externally introduced perspectives may become visible in learners’ later reasoning processes. Although illustrated through an AI-mediated inquiry activity, the proposed framework is not tied to a specific tool or domain. Instead, it offers a generalizable approach for aligning inquiry design with trace-based process analytics, enabling systematic investigation of problem-finding as it unfolds.</p>