From Black-Box to Collaborative: Position Paper on Human-Guided Trace Clustering

MCML Authors

Udo Schlegel

→ Group Thomas Seidl
 Database Systems, Data Mining and AI

Gabriel Marques Tavares

Dr.

→ Group Thomas Seidl
 Database Systems, Data Mining and AI

Thomas Seidl

Prof. Dr.

Director

Database Systems, Data Mining and AI

Abstract

Trace clustering techniques partition event logs into coherent groups of process variants but fundamentally lack objective ground-truth labels against which cluster quality can be meaningfully assessed. Current evaluation practices rely on proxy metrics, such as model fitness and silhouette scores, that fail to address the inherently task-dependent and user-specific nature of meaningful process variant groupings. This position paper argues that the field urgently needs human-guided frameworks combining visual analytics and active learning to systematically incorporate domain expertise into trace clustering. We propose that coordinated visual interfaces with strategic query mechanisms, soliciting constraints, prototype labels, and goal-directed feedback, represent the missing systematic paradigm for embedding subjective expertise into algorithmic clustering. Widespread adoption of such frameworks would transform trace clustering from black-box automation into collaborative intelligence, enabling scalable, interpretable trace clustering aligned with real organizational analysis needs.

inproceedings STS+26

---

Vipra @EuroVis 2026

Visual Process Analytics Workshop at the Eurographics Conference on Visualization. Nottingham, UK, Jun 08-12, 2026.

---

Authors

U. Schlegel • G. M. Tavares • D. Schuster • T. Seidl

---

Links

DOI

---

Research Area

 A3 | Computational Models