Peer Review

Just recently I received an email from the organizers of a new conference concerned with the technique of peer review. This events seems to be largely concerned with the review process applied to scholarly papers but it got me thinking about "review by peers", something that I do a lot of.

As a practitioner of Extreme Programming of course I prefer to write production code as one member of a pair. And pair-programming is explicitly an activity needing an effort towards a peer relationship: the higher-powered programmer in the pair has to throttle back a little, and the lower-powered one stretch. Peer-review of papers usually involves a panel of reviewers for each submission, and if you rotate people through pairs often enough (which might turn out to be very often indeed if this experience report[pdf] is anything to go by) then it won't be long before the code changes you make have been thoroughly reviewed by the time it gets checked in.

Then again, during the planning game I like to use Delphi techniques to explore possibilities, reach consensus, and obtain estimates. Delphi works by having a group of, as they are called in the technique, "experts", who are by definition peers, iteratively review a proposal or (answer to a) question. Interestingly, the anonymous nature of the input to a Delphi review would seem to fix some of the problems that occur when using less formal means to reach a consensus.

In development shops that don't do pairing, but do care about quality all the same, you'll often come across code review techniques of one sort or another. Fagan Inspection seems to be the best way to get this done with small groups, although my experience it is ferociously expensive and although produces excellent results perhaps doesn't offer great value-for-money. YMMV. If you have a (potentially) large group available, then the Open Source route is also a good one. Can be very slow, but also very effective, and (like the National Health Service), free of cost at the point of delivery...

And then there's reviewing conference sessions and papers and journal papers and drafts of books. This can be pretty excruciating work. As a reviewer, I whish more authors would heed this advice when preparing their submissions. As an author, I whish more conferences worked along the lines of the PLoPs, which are all about very intensive, hands-on, in the room, right before your eyes peer review.

Although it is (in)famously possible for a clever and resourceful author to get carefully crafted utter tosh published in a supposedly serious journal, there have also been cases of what seems to be genuine fraud (as it were) getting past the review process in vary hard science fields indeed. One has to wonder if this isn't, as with the remarkable ease with which dubious patents may be obtained these days, mostly due to reviewers being snowed under.

On the other hand, the organisers of that conference on peer review I mentioned up at the top are themselves notorious for accepting nonsense papers without review. In a spirit of enquiry I replied politely to the email suggesting that I probably wouldn't be able to attend the conference, but that I would be interested in helping out with the proposed book on peer review--as a reviewer. We'll see what comes of that.

1 comment:

Code review techniques said...

I am always in search of new code review techniques. In different organization different techniques are used.