Brinkman, RubenMannhardt, FelixMennens, Robin J. P.Scheepens, Roeland J.Arleo, Alessiovan den Elzen, Stefvon Landesberger, TatianaRehse, Jana-RebeccaPufahl, LuiseZerbato, Francesca2024-05-212024-05-212024978-3-03868-254-7https://doi.org/10.2312/vipra.20241103https://diglib.eg.org/handle/10.2312/vipra20241103Process mining discovers process models from an organization's event logs. Discovered process models are used by process analysts to understand and improve real-life processes. Interpretability of such discovered process models by actual users is crucial for efficient and effective usage of models in process analysis. A large body of work, including empirical studies, investigates how users interpret process models and their visualization. However, the focus is on manually created process models for documentation or specification. There is little work on the influence of discovered process models visualization on interpretability by users. Often discovered models are augmented with frequencies and deviations from an event log, which leads to even more complex visualizations. We contribute a user study with 12 participants with varying level of process mining expertise and derive a ranking of 15 issues for interpretability in discovered process model visualizations. We derived five requirements for an improved process model visualization that we, subsequently, implemented in a prototype visualization. A preliminary validation of the prototype among a subset of participants showed promising results and, orthogonal, the identified issues may be useful for future research and work on the interpretability of discovered process models.Attribution 4.0 International LicenseCCS Concepts: Applied computing → Business intelligence; Human-centered computing → Empirical studies in visualizationApplied computing → Business intelligenceHuman centered computing → Empirical studies in visualization10.2312/vipra.202411036 pages