> Why do we need to bother with repeatable experiments for social sciences?

So we aren't lead into error by well-known cognitive biases and/or outright fraud, both of which are known to exist as they can be demonstrated through repeatable experiments.