More rigor is always good, but there’s also value in studying messy real-world behavior as it happens