AdONE Seminar - Monday, May 18, 2026: Prof. Giovanni Pantuso (University of Copenhagen)

Solution of arbitrary stochastic programs with decision-dependent uncertainty by discretization

In this talk we discuss stochastic programs with decision-dependent uncertainty. Unlike classical stochastic programs, here the distribution of the uncertain parameters depends on the decision vector through an arbitrary distribution map. Problems of this type are notoriously challenging. We show that solutions to such problems, within arbitrary precision, can be obtained by means of (sequences of) piecewise-constant approximations constructed on partitions of the feasibility set. In particular, optimal objective values and solutions emerge in the limit as the resolution of the partition is refined. The concept is illustrated through numerical examples. 

Date: May 18, 2026

Time: 5 pm, s.t.

Place: online 

 

AdONE Seminar - Monday, June 8, 2026: Speaker tba

Title: tba

Date: June 8, 2026

Time: 5 pm, s.t.

Place: tba

 

AdONE Seminar - Monday, June 15, 2026: Prof. Fabrizio Maria Maggi (Free University of Bozen-Bolzano)

AI-augmented Business Process Management

This talk will discuss the evolution of AI-augmented Business Process Management focusing on how artificial intelligence can enhance process automation, improvement, and adaptability to change. In this context, we investigate the use of automated planning as a core mechanism to support framed autonomy. We define the process environment as a frame of multifaceted constraints, including declarative rules, procedural models, resource availability, costs, and deadlines. Each constraint is associated with a violation cost, capturing the criticality of the constraint in real-world process execution. Given a running process instance (i.e., a prefix) that may violate this frame, automated planning is used to compute an optimal continuation that minimizes the overall violation cost. In this way, the system can recommend or enact process adaptations that are the most feasible ones based on the given costs. Complementary to this perspective, we are also studying Neuro-Symbolic Predictive Process Monitoring (PPM) as another key direction in AI-augmented BPM. In this setting, predictive models based on neural networks are enriched with symbolic knowledge, enabling them to handle exceptional or rare situations that are not sufficiently represented in historical data. While neural components provide strong predictive capabilities, symbolic knowledge supports reasoning beyond the data, improving robustness and adaptability to process change.

Date: June 15, 2026

Time: 5 pm, s.t.

Place: Boltzmannstr. 3, 85748 Garching, room MI 01.10.011

 

AdONE Seminar - Monday, June 22, 2026: Prof. Christina Imdahl (TU Eindhoven)

Title: How to Account for Behavioral Newsvendors: The Robust Buyback Contract to Address Response Uncertainty

Normative (expected-profit-maximizing) theory assumes that decision-makers are fully rational, but in reality, they deviate from the optimal response. In contract negotiations, contract parameters are often optimized based on the assumption of rational behavior. Deviations from rational behavior can lead to significant costs for all parties involved. To address this, we propose robust optimization to obtain contract parameters that account for deviations from rational behavior. We apply this optimization approach to obtain the robustly optimal contract parameters for the buyback contract when there is limited knowledge of the buyers' responses. We develop a parameterization of the robust contract, called the naive robust contract, that is based on the pull-to-center effect.  We compare the naive robust contract to the normative, the naive robust and various behavioral contracts based on established behavioral response models in a lab experiment. The results show that a robust contract, built on simple assumptions, yields higher average supply chain profits than the normative contract. The robust contract also weakly dominates all other behavioral contracts, meaning that it significantly outperforms them in either average supply chain profit or risk. This approach offers a practical way to set contract parameters without relying on assumptions about the distribution of responses, except for the support.

Date: June 22, 2026

Time: 5 pm, s.t.

Place: Arcisstr. 21, room Z534/Z536 (0505.Z1.534Z/0505.Z1.536Z)

 

AdONE Seminar - Monday, July 6, 2026: Speaker tba

Title: tba

Date: July 6, 2026

Time: 5 pm, s.t.

Place: tba