Operations Research 2025
Abstract Submission

44. Reassessing Validation & Verification in AI-Driven Modeling and Simulation

Invited abstract in session WE-7: Simulation, Data & Decision Support, stream Simulation and Quantum Computing.

Wednesday, 16:30-18:00
Room: U2-205

Authors (first author is the speaker)

1. Dominic Weller
Bundeswehr Office for Defense Planning, German Armed Forces

Abstract

The German Armed Forces regularly conduct scientific studies in the field of analytical and
constructive modelling and simulation (M&S) to support decision-making and capability
planning. Artificial Intelligence (AI), particularly machine learning (ML), is becoming an
integral part of these projects. However, due to partly insufficient results from previous studies and the rapid development of AI, questions arose if existing verification and validation (V&V) approaches and established M&S processes are still adequate.

Therefore, we conducted a comprehensive internal study to find answers regarding these concerns. This included
1. a meta-analysis of previous studies and thereby identifying recurring challenges and lessons learned,
2. a systematic analysis of the current state of research and an intensive exchange on the topic with industry partners, NATO experts, and academic institutions,
3. the contextualization and definition of the problems related to the use of ML and AI within the M&S context, therefore providing a practitioners perspective and
4. the transfer of the findings to our guidelines, procedures, and processes to identify
areas for action and to suggest best practices.

The results show that most of the identified problems often have a basic character and not necessarily originate from the application of AI itself. We must apply current M&S and V&V methods consistently and a deeper understanding of AI/ML technologies must be established. Only when the basic prerequisites of M&S are met, advanced methods such as Explainable AI (XAI) and corresponding frameworks may contribute to supporting certain steps in the simulation process.

Keywords

Status: accepted

Back to the list of papers