Ah, so the social science reproducibility problems are hypothesized to be caused by bias (or, less provocatively, filtering), moreso than e.g. bad experimental design. Yes, that would nicely reconcile both perspectives.
There's plenty of room for both. If you have lots of bad experiments and filter the ones you agree with, the combination is even worse because there's an apparent consensus.
