Scientific Explanation

Scientific explanation, a core goal of science, is undergoing renewed scrutiny, particularly concerning its application in artificial intelligence (AI). Current research focuses on bridging the gap between philosophical understandings of explanation (e.g., causal-mechanistic models) and the practical challenges of creating interpretable AI models, often employing techniques like disentangled representation learning and symbolic regression to achieve more transparent and understandable results from complex data like images. This work aims to develop robust benchmarks for evaluating scientific understanding in both humans and machines, ultimately improving the reliability and trustworthiness of AI-driven scientific discovery and knowledge dissemination.