- Research Article
- 10.15187/adr.2025.11.38.4.143
An Exploratory Automated Analysis of UX Practitioners’ Interviews: Structuring and Visualizing Cognitive Flows
- Nov 30, 2025
- Archives of Design Research
- Eunsol Lee + 2 more +2
Background : In-depth interviews have long been recognized as a central method in user experience (UX) and design research, providing access to users’ perceptions, reasoning, and contextualized experiences. Unlike surveys or behavioral data, interviews reveal how users construct meaning, shift perspectives, and integrate emotions into their experiences. However, the analysis of interview data remains labor-intensive and highly dependent on manual coding, often resulting in inconsistencies and limited scalability. Conventional methods such as grounded theory and thematic analysis capture themes and concepts but struggle to structurally represent cognitive transitions over time. Recent advances in natural language processing (NLP), embeddings, and clustering offer new opportunities, yet these methods are mostly optimized for short or semi-structured text. Their application to lengthy, layered interview narratives remains limited. To address these challenges, this study introduces an automated framework that segments interview data into cognitive units and visualizes reasoning flows to support more consistent and scalable qualitative analysis.<br/>Methods : We developed a Python-based analytical procedure consisting of four steps: semantic unit segmentation, clustering, cognitive tagging, and visualization. Six cognitive tags were defined: Situational Awareness, Conceptual Clarification, Strategic Application, Reflection & Insight, Perspective Shift, and Constraints & Limitations. A supervised classifier trained in manually labeled utterances automated the tagging process. Visualization tools such as Sankey diagrams and heatmaps were used to illustrate cognitive transitions and centralities. Data collection comprised seven in-depth interviews with UX designers, product managers, and data scientists, resulting in 9.5 hours of recorded material and approximately 110,000 characters of transcribed content.<br/>Results : Applying the automated procedure to a total of 1,041 utterances demonstrated that interview data can be quantitatively structured and analyzed along temporal flows and transition patterns. Sankey diagrams and centrality analyses revealed recurring cognitive transitions and key nodes, allowing the structural characteristics of reasoning flows—often difficult to capture through manual coding—to be objectively identified. These findings highlight the potential of analyzing long-form qualitative interviews in a consistent and reproducible manner.<br/>Conclusions : This study proposes an automated framework for structurally quantifying and visualizing interview data. Rather than merely classifying types of reasoning, the approach expresses utterance-level transitions and flows through quantitative metrics and visual maps, thereby complementing the subjectivity and limitations of traditional qualitative analysis. Implemented entirely with open-source Python tools, the method can be adopted without advanced technical resources and offers a pathway toward the standardization and scalability of qualitative analysis in UX and design research.
Read more