One of the major propositions that the critics of IRB review of social research put forth is that minimal risk social research with competent adults should be completely exempt from IRB review. On the surface, this makes some sense. We're talking about research that is unlikely to harm anyone and where adults can decide for themselves whether to participate. Why do we need to review such research? Based on my experience personally reviewing thousands of research protocols in the social sciences, there is one basic problem with this - researchers are human beings. Human beings are not perfect - they overlook things, make mistakes and and can't be totally objective about their own work. If researchers were perfect, if they always took all of the ethical issues into account when planning and conducting their research, then we wouldn't need IRB review. But they are not perfect - none of us are perfect. So, every research activity needs an independent, objective review.
Characteristically, Cohen offers not a single example of a social research project whose ethical content was improved by his or any IRB review, much less one that could only be improved thanks to the broad definitions and coercive rules now used by IRBs. (And no, forcing interviewers to carry lists of mental-health centers doesn't count.) If he has such examples, he should offer them. If not, a vague reference to "thousands of research protocols" is unlikely to persuade a community of scholars trained to think critically and to weigh evidence.