I'd need to see a double-blind randomized controlled experiment with tens or hundreds of Earth-like worlds, some with human activity and some without. No n=1 observational study with potentially infinite known and unknown confounding variables can give credible scientific evidence.