Process mining (PM) and visual analytics (VA) share a common goal: helping people understand complex behavior from data. PM provides techniques to analyze process event data, i.e., records of activities performed over time, to discover, compare, and assess the processes that generated them. VA combines computational analysis with interactive visual interfaces to support the exploration and interpretation of complex data. Its iterative human-in-the-loop approach lets users dynamically steer exploration, formulate and refine hypotheses, and incorporate domain knowledge, supporting sensemaking under uncertainty.

Bringing these two perspectives together is particularly promising for unstructured or loosely structured processes, i.e., processes whose behavior does not follow clearly defined execution paths. Such processes are common in domains such as healthcare, smart environments, software systems, and user-interaction data, where behavior is highly variable, may evolve over time, and exhibits only weak regularities. Traditional PM techniques often summarize behavior in global process models that, for highly unstructured data, become overwhelmingly complex, and provide little support for understanding relevant behavior. VA, in turn, focuses on designing interactive visual methods to address this complexity, but less attention has been paid to methods tailored to process-oriented and constraint-based data.

This Dagstuhl Seminar will investigate how interactive visual process analytics can support the exploration of unstructured processes. Rather than assuming that one global model can adequately describe the data, the seminar will explore approaches that help analysts identify and relate local, context-dependent behavioral patterns, move between levels of detail while preserving traceability, and use interaction and domain knowledge to refine their interpretation. This combination is mutually beneficial: PM can draw on VA for interactive, human-in-the-loop exploration, while VA can leverage PM concepts and representations to explore behavioral, temporal, and constraint-based relationships in event data.

These three directions translate into concrete starting points for the seminar. Declarative process modeling as an exploratory paradigm raises the question of how models based on constraints rather than fixed execution paths can serve as interactive representations of local, context-dependent behavioral patterns, supporting reasoning and exploration under variability and ambiguity. Multi-level abstraction considers how analysts can navigate different levels of granularity when the appropriate level of detail is not known in advance, and calls for interactive and progressive visual methods that make granularity explicit, navigable, and explainable. Interactive visual analytics provides the integrating principle, letting analysts steer exploration, compare behavior, adapt abstractions, relate local patterns, and incorporate domain knowledge into sensemaking. Across these directions, the seminar will address broader questions on human sensemaking under uncertainty, heterogeneous, incomplete and evolving event data, exploratory workflows and iterative hypothesis refinement, scalable and progressive computational and visual methods, user guidance and interpretability, interactive storytelling, and empirical evaluation. Representative datasets from healthcare, personal productivity, and process mining analysis workflows will ground discussions in concrete problems.

The seminar’s ultimate goal is to develop a focused research agenda for the interactive analysis of unstructured processes, identifying key conceptual, methodological, and technical challenges. Interdisciplinary working groups will explore concrete research questions and develop initial concepts and approaches, with the intention of continuing these collaborations through joint publications, funding initiatives, and broader community activities. The seminar continues the exchange begun with the Dagstuhl Seminars "23271 – Human in the (Process) Mines" and "25152 – Multi-Faceted Visual Process Mining and Analytics", which produced lasting interdisciplinary collaborations, a special issue in the Information Systems journal, and the VIPRA (Visual Process Analytics) workshop series, among others.

Creative Commons BY 4.0

Manuel Resinas, Christian Tominski, Katerina Vrotsou, and Francesca Zerbato