Automated Scientific Discovery with Software Engineering and AI
by Mathieu Acher
09/01/2025
DiverSE Coffee
Rennes, France
Abstract
This talk explores the intersection of artificial intelligence and software engineering in advancing automated scientific discovery. Key themes include the role of variability in experiments, the importance of reproducibility and replicability, and the challenges of applying AI across diverse scientific fields. Case studies, such as numerical program analysis, soccer data interpretation, and prompt variability testing, illustrate the transformative potential of AI in structuring data, analyzing results, and testing hypotheses. The discussion emphasizes the need for a balance between automation and human interpretation to ensure robust and generalizable scientific progress.