IEEE asks journal authors to submit the names of their (non-)preferred reviewers. Given it may not be appropriate to list an acquaintance as a reviewer, how does one judge if an anonymous person will be a good reviewer or not?
Is giving "non-preferred" reviewers akin to the author confiding about his academic relationships to the editor? Do people really need to fill this column up at all?
Just to add an editors perspective even though it partially duplicates the other answers:
Having seen the effect of preferred reviewers in the journal where I serve, I am not altogether happy about these options. The non-preferred category is one that I take seriously because conflicts can lead to bad or poor reviews for reasons other than the science. My problem is instead with the preferred reviewers.
When I assign reviewers I basically steer clear of the preferred reviewers as a rule. In the past when I have selected from this list, I have been extremely disappointed. Almost exclusively I have received "accept" reviews (accompanied by "rejects" from other reviewers) which I have been forced to deem practically useless and look for additional reviewers. I do not mean to say that all persons put on these lists provide poor reviews but what I perceive as the original purpose namely to provide examples of names of persons who's objective opinion the author highly values is rapidly being drowned by names of friends. This has led me to stay clear of this list unless I myself recognize a name as somebody with integrity. I will nevertheless scrutinize that review with a hint of suspicion based on the experience i have. One way to avoid this could be to add a line or two stating why you have selected these names as preferred. This would help evaluate the usefulness of such reviewers. Sadly many submission systems do not allow for such comments except in the submission letter.
So I agree with the others answers, provide particularly non-preferred names but feel free to add also preferred but do not count on the preferred names to be used.