reviewing

[Thomas G. Dietterich]

Hi Juergen,

It may be that Nobel prize winners don't benefit from peer review, but
I know that many of my papers have been improved as a result of peer
review.  

If you read the acknowledgements sections of the papers published in
Machine Learning, Neural Computation, and JMLR, you will often see the
authors thanking the referees.  I know of several cases where the
referees not only identified bugs in proofs, but helped strengthen
theorems and simplify proofs.

Why is there this difference between machine learning and other
discplines?  Perhaps because in our young discipline research is not
nearly as competitive as it is in mature fields such as physics and
biology.  A difficulty with physics and biology is that you and your
competitors are studying the *same system* and trying to answer
exactly the *same questions*.  It is a race to publish, "priority"
matters, and unscrupulous reviewers can delay a paper unfairly.

But in computer science, perhaps because it is a "Science of the
Artificial", most work involves developing frameworks, perspectives,
analytical techniques, and models, and these rarely compete directly.
For all of these, peer review can play an important role in checking
the proofs, clarifying the ideas, and improving the presentation.  In
this way, I think peer review advances the field rather than holding
it back.

--Tom

-- 
Thomas G. Dietterich              Voice: 541-737-5559
Department of Computer Science    FAX:   541-737-3014
Dearborn Hall 102                 URL:   http://www.cs.orst.edu/~tgd
Oregon State University
Corvallis, OR 97331-3102