This is a comment on research in education (from Tom Bennett) but it could apply across the whole of social policy:
I'm sure these people are engaged in the most rigorous of science, but the area that it addresses is devilled with darkest, emptiest aspects of bad educational research: small intervention groups, interested parties, cognitive bias, short term studies, conclusions that don't necessarily follow from the data, an aversion to testing a theory to destruction, etc. This matters, because huge and enormously expensive wheels are turning in education ministries around the world. Children's lives are chained to this wheel. Poor children can't afford to fix the mistakes of state education, as middle-class children can, through tutoring and familial support.
Yet we persist with allowing ideological bias and personal preference to be presented as research by social scientists. As we keep saying, if you really want evidence-based policy you need to start with robust evidence not ideological bias.