ORAHS2025
Abstract Submission

22. Explainable Optimisation in Healthcare

Invited abstract in session HC-4: Innovation 1, stream Sessions.

Thursday, 11:00-12:30
Room: St Olavs, Kunnskapssenteret KA12

Authors (first author is the speaker)

1. Felix Engelhardt
Research and Teaching Area Combinatorial Optimization, RWTH Aachen University
2. Christina Büsing
Lehr- und Forschungsgebiet Kombinatorische Optimierung, RWTH Aachen University
3. Catherine Cleophas
Service Analytics, CAU Kiel University

Abstract

Explainability is an important topic in AI research, since many common techniques are “Black Boxes” for users. In comparison, in optimisation and operations research, we like to think that our models are explainable by nature. However, is that really the case? To actual users such as industry partners, doctors or nurses, a mixed-integer programming solver is as arcane as a deep neural network. The same applies to many combinatorial algorithms and heuristics.

In this session, we raise several examples of explainability in optimisation in the specific context of healthcare optimisation. These include modelling to generate alternatives for rostering problems, counterfactual explanations for (integer) linear programming, and simulation – showcasing the broad range of notions of explainability that might be relevant in practice.

Based on this we would like to discuss the following questions:
* What types of explainability do practitioners in healthcare desire (e.g. counterfactuals vs transparent vs rule-based)?
* To what extent can explainability help with successfully implementing real-world problems?
* What techniques are being used to do so?
* Where are potential deficits in terms of the current explainability of healthcare OR, and where could research be strengthened by including explainability considerations?

Keywords

Status: accepted

Back to the list of papers