14 Nov 2025
15:00 Doctoral defense Via video conference (link meet.google.com/fwq-xcep-wxr)
Topic on
Conceptual Modeling and Automated Planning for Knowledge-Intensive Processes Under Uncertainty
Student
Sheila Katherine Venero Ferro
Advisor / Teacher
Cecília Mary Fischer Rubira - Co-advisors: Leonardo Montecchi and Julio Cesar dos Reis
Brief summary
Knowledge-Intensive Processes (KIPs) are unpredictable, context-sensitive, and heavily dependent on knowledge, information, and data. A fundamental challenge in modeling KIPs is managing their inherent uncertainty, which arises from complex, often unknown factors that can change during execution or vary from case to case. As a result, it is difficult to predetermine the process structure, specifically which activities should be performed and in what sequence. Therefore, the structure must emerge progressively through collaborative decision-making by knowledge workers, who rely on their expertise and situational awareness as they adapt to evolving objectives and new information. Given these complexities, traditional process modeling approaches struggle to represent KIPs effectively. In response, data-centric approaches, such as artifact-centric, object-aware, and case management, have been proposed, offering greater flexibility by focusing on user decisions and data. However, most of these approaches still fail to address the multifaceted nature of KIPs, particularly their inherent uncertainty. To address this gap, this thesis proposes an integrated approach that combines data-centric modeling and automated planning techniques to support PICs under uncertainty. It defers structural decisions to runtime, enabling dynamic adaptation to evolving circumstances. The approach combines knowledge workers and computational agents in a continuous cycle of planning, execution, supervision, and replanning. At its core, the approach models PICs as planning problems using Markovian Decision Processes (MDPs), which explicitly capture uncertainty through probabilistic state transitions and support optimized decision-making. The proposed solution is supported by METAKIP, a metamodel developed in this thesis that allows conceptual specification of both the planning domain and the planning problem. From this model, a MDP is automatically generated and solved, producing plans in the form of process model fragments. The resulting plans are suggested to knowledge workers to support decision-making and guide the construction of the process structure at runtime. The solution is based on the MAPE-K (Monitor, Analyze, Plan, Execute, and Knowledge) autonomic cycle, providing a continuous cycle of adaptation. To validate the feasibility and effectiveness of the approach, we developed MAPKIP (Modeling, Analysis, and Planning of Knowledge-Intensive Processes), an online prototype tool that integrates the PRISM model checker. A healthcare case study demonstrates the potential of the approach to support informed and adaptive decision-making in uncertain environments.
Examination Board
Headlines:
| Cecília Mary Fischer Rubira | IC / UNICAMP |
| Marcelo Fantinato | EACH / USP |
| Elisa Yumi Nakagawa | ICMC / USP |
| Eliane Martins | IC / UNICAMP |
| Bruno Barbieri de Pontes Cafeo | IC / UNICAMP |
Substitutes:
| Breno Bernard Nicolau de França | IC / UNICAMP |
| Paulo Henrique Monteiro Borba | CIn / UFPE |
| Claudia Maria Lima Werner | COPPE/UFRJ |