Explanations help computers learn how causes really work
xWhyL: Causal Interactive Learning
Machine LearningArtificial Intelligence
Summary
Understanding why things happen helps people learn about cause and effect, but computers usually ignore this. The authors developed a way for computers to learn from explanations given by experts, improving their ability to find cause-effect relationships beyond just looking at data. Their system can tell when explanations are wrong and does not just accept them blindly. This method shows how human explanations can make computer models of cause and effect more accurate and reliable.
What this means in practice
- •For data science teams: Use expert explanations to improve algorithms that find cause-effect relationships in complex datasets.
- •For machine learning developers: Integrate human feedback through explanations to make causal models that better distinguish correct from incorrect assumptions.
Authors
Nicholas Tagliapietra, Florian Peter Busch, Moritz Willig, Matej Zečević, Lavdim Halilaj, Juergen Luettin, Kristian Kersting
Abstract
Explanations are central to causal reasoning, and cognitive science has long established that the human drive to explain is itself a mechanism for learning about causality. Despite this, learning from those abductive signals is largely ignored in artificial intelligence. While explainable AI (XAI) increasingly draws on causal models to generate explanations, the converse direction about what explanations can do for causality remains largely unexplored. To fill this gap, we propose xWhyL, a formal framework connecting causality and XAI by learning causal models from explanations. We develop a mathematical theory that translates explanations into a learning signal complementary to observational data, and demonstrate how it enables overcoming the limits of observational causal discovery. As explanations can be derived from incorrect beliefs and clash with data, a tension we call the Causal Tug-of-War, we prove conditions under which our framework rejects misspecified explanations rather than absorbing them. Our practical instantiation, Causal Interactive Learning (CIL), shows how expert explanations can efficiently support causal discovery and distinguish correct from incorrect explanations.