New Collaborative Project: Artificial Intelligence That Recognizes Causal Relationships
The German Federal Ministry of Education and Research (BMBF) is funding the new collaborative project CausalNet with nearly two million euros. The goal: to develop a new generation of machine learning over the next three years that can understand cause-and-effect relationships. This ability will make AI applications more flexible, efficient, and reliable.
Challenge: From Correlation to Causation
Current machine learning models are primarily based on correlations rather than causal relationships. In other words, they make predictions based on probabilities without identifying true causes. This leads to limitations in certain areas and can affect the models’ performance.
"The current capabilities of trained models are already incredibly impressive. However, in certain applications, we are particularly interested in causal questions or causal abstractions – in understanding how these models work and how we can manipulate or control them," explains Prof. Stefan Bauer, AI scientist at Helmholtz Munich and one of the CausalNet project partners.
The potential of artificial intelligence in medical applications is also limited by the fact that current programs cannot establish causal connections. A model that links cause and effect could enable more targeted therapy decisions. The same applies to applications in science, business, and the public sector.
Collaborative Research for a New AI Paradigm
"We aim to develop novel methods for integrating causality into machine learning models," says Professor Stefan Feuerriegel. He is the head of the Institute of Artificial Intelligence in Management at Ludwig Maximilian University (LMU) and spokesperson for CausalNet. To incorporate the principle of cause and effect into future AI models, experts from LMU, Helmholtz AI, the Technical University of Munich (TUM), the Karlsruhe Institute of Technology, and Economic AI GmbH are collaborating.
"The ability to integrate causal relationships into AI models is key to enabling trustworthy, fair, and ethically sound decisions in complex systems in the long term," says Prof. Niki Kilbertus, who leads a research group for Helmholtz AI at Helmholtz Munich. "True causal explanations are crucial for avoiding biases and unfair patterns that often occur when models only recognize correlations. By considering causality, we can improve not only the accuracy but also the transparency of AI decisions – an essential step for the responsible use of AI in society."
The team plans to tackle the unique challenges of causal machine learning in high-dimensional environments using tools from representation learning, the theory of statistical efficiency, and specific machine learning paradigms. "Moreover, we will derive the effectiveness and robustness of our methods through theoretical results," says AI expert Feuerriegel. This is important to ensure the reliability of the proposed methods. "Then, we will bring causal machine learning into real-world applications and demonstrate its tangible benefits for the economy, public sector, and scientific discoveries."
Open Source for Broad Application
CausalNet aims to promote the practical use and further development of its findings by making the developed software, tools, and results publicly available under the open-source principle. "Over the next three years, we will elevate machine learning to a new level, making AI applications more flexible, efficient, and robust," says Feuerriegel.
