
Multi-Level Abstractions and
Causal Modeling for Enhanced
Reinforcement Learning
Integrating expert and domain knowledge into deep reinforcement learning agents through multi-level abstractions and causal modeling. We want to foster interpretable, more robust DRL agents that can generalize across environments.
About
MAC-MERLin aims to integrate established expert and domain knowledge into artificial neural networks by leveraging multi-level abstractions in combination with causal modeling. Building on the established concepts of causality and abstraction, the project improves the interpretability of Deep Reinforcement Learning (DRL) agents and aligns them more closely with human understanding and expert insights, while at the same time making them more robust, better able to generalize across environments, and better tuned to real-world complexity.
The project addresses four fundamental challenges that currently make the practical deployment of neural-network decision making difficult: modeling, robustness of DRL agents, their ability to generalize to new environments, and their interpretability. By closing the gap between the theoretical and practical aspects of DRL, MAC-MERLin contributes to applications across smart cities and power grids, intelligent supply chains and production systems, autonomous driving, drug discovery, medical treatment and bioengineering.
Publications
2026
Chaotic and complex dynamics expose the limits of counterfactual reasoning
2026 — Scientific Reports (Sci Rep)
Per-Domain Generalizing Policies: On Learning Efficient and Robust Q-Value Functions
2026 — Proceedings of the Thirty-Sixth International Conference on Automated Planning and Scheduling. International Conference on Automated Planning and Scheduling (ICAPS-2026), June 27 - July 2, Dublin, Ireland
What are AI systems? Rethinking the core definition in the EU AI Act
2026 — Computer Law & Security Review
PGTG: Procedurally Generated Grid-Based Traffic Gym
2026 — Reinforcement Learning Journal
2026 — Workshop on Planning in the Era of LLMs at ICAPS-26
2025
Learning Per-Domain Generalizing Policies Using Offline Reinforcement Learning
2025 — Proceedings of the Ninth Workshop on Bridging the Gap Between AI Planning and (Reinforcement) Learning (PRL @ ICAPS 2025). Workshop on Bridging the Gap Between AI Planning and Reinforcement Learning (PRL-2025), located at ICAPS'25, November 10, Melbourne, Australia
PyDSMC: Statistical Model Checking for Neural Agents Using the Gymnasium Interface
2025 — Quantitative Evaluation of Systems and Formal Modeling and Analysis of Timed Systems - Second International Joint Conference, Proceedings. International Joint Conference on Quantitative Evaluation of Systems and Formal Modeling and Analysis of Timed Systems (QEST+FORMATS-2025), August 26-28, Aarhus, Denmark
Visual Analysis of Action Policy Behavior: A Case Study in Grid-World Driving
2025 — World Conference on Explainable Artificial Intelligence. xAI: World Conference on Explainable Artificial Intelligence (xAI-2025), July 9-11, Istanbul, Turkey
Using Action-Policy Testing in RL to Reduce the Number of Bugs
2025 — Proceedings of the Eighteenth International Symposium on Combinatorial Search. International Symposium on Combinatorial Search (SoCS-2025), August 12-15, Scotland, United Kingdom
Structural Causal Bottleneck Models
2025 — UAI 2025 Workshop on Causal Abstractions and Representations. Causal Abstractions and Representations (CAR), located at UAI 2025, July 25, Rio de Janeiro, Brazil
A Roadmap for Responsible Robotics: Promoting Human Agency and Collaborative Efforts
2025 — IEEE Robotics & Automation Magazine
Disentangling AI Alignment: A Structured Taxonomy Beyond Safety and Ethics
2025 — AISoLA 2024. International Conference on Bridging the Gap between AI and Reality (AiSoLA-2024)
Justifications for Democratizing AI Alignment and Their Prospects
2025 — AISoLA 2025. International Conference on Bridging the Gap between AI and Reality (AiSoLA-2025)
On the Complexities of Testing for Compliance with Human Oversight Requirements in AI Regulation
2025 — International Conference on Bridging the Gap between AI and Reality. International Conference on Bridging the Gap between AI and Reality (AiSoLA-2025)
How do we assess the trustworthiness of AI? Introducing the trustworthiness assessment model (TrAM)
2025 — Computers in Human Behavior
The Landscape of Causal Discovery Data: Grounding Causal Discovery in Real-World Applications
2025 — Proceedings of the Fourth Conference on Causal Learning and Reasoning. Conference on Causal Learning and Reasoning (CLeaR-2025)
2025 — Proceedings of the Fourth Conference on Causal Learning and Reasoning. Conference on Causal Learning and Reasoning (CLeaR-2025)
Separation-Based Distance Measures for Causal Graphs
2025 — Proceedings of The 28th International Conference on Artificial Intelligence and Statistics. International Conference on Artificial Intelligence and Statistics (AISTATS-2025), May 3-5, Mai Khao, Thailand
Bridging the Gap Between AI Planning and Reinforcement Learning
2025 — AAAI 2025 Workshop Proceedings (PRL Workshop). Workshop on Bridging the Gap Between AI Planning and Reinforcement Learning (PRL-2025), February 25
Per-Domain Generalizing Policies: On Validation Instances and Scaling Behavior
2025 — The 35th International Conference on Automated Planning and Scheduling. International Conference on Automated Planning and Scheduling (ICAPS-2025), November 9-14, Melbourne, Australia
2024
Motion Primitives as the Action Space of Deep Q-Learning for Planning in Autonomous Driving
2024 — IEEE Transactions on Intelligent Transportation Systems