Analyzing Actor Behavior in Process Executions

Process executions are driven by actors and machines jointly performing work on incoming cases. Actors typically divide and structure their work into tasks - ,multiple consecutive actions performed together - before handing a case to the next actor. Routines research investigates how and why actors divide and structure work in a process, how it changes over time, and its impact on process executions. So far, process event log data has not been used to investigate these phenomena. We conducted an exploratory case study to identify process mining concepts and data structures that help answering the above questions. We found that modeling case and actor traces together in an event knowledge graph allows to identify instances of task executions; clustering task instances reveals tasks. Extending the event knowledge graph by aggregation wrt. tasks reveals, both, local process models of intra-task behavior, and global process models of inter-task behavior in a case and between actors. We show on the BPIC'17 dataset that querying the extended graph reveals new insights into (changes in) actor behavior, work division, and significant impacts on performance and outcomes.