I would say objective reality is dubious, and wonder what claims do "survive testing"?
On the one hand, human behaviour, culture and influences is extremely fragmented, to the point where I wonder what is it we're trying to model? The subject we're trying to predict is part of this culture and influences, and probably shifting so fast it's questionable we can, with accuracy, predict much.
On the other hand we have the interpretor, the constructor of the study, who also is colored by culture and influences to the point where we must ask why is a certain model being constructed?What is predicted and why? How are people's actions interpreted to fit that model?