[UAI] Reviewing protocols (whether or not UAI)

From: Paulusnix@cs.com
Date: Thu Nov 15 2001 - 09:41:37 PST

  • Next message: Ruhul Sarker: "[UAI] 2002 IEEE SMC - Special Session on Evolutionary Optimization"

    Greetings :-

        Professor Langseth argues that the values of "professionalism" and
    "review quality" are better served by a double-blind review procedure
    (neither reviewers nor authors are told one another's identities, reviewers
    are not told each other's identities) than by a single-blind procedure
    (reviewers are told authors' identities, but not vice versa, and reviewers
    are not told one another's identities).

        A third alternative, open review (common knowledge of all identities
    among all immediate stakeholders, all of whom read all of the reviews) is
    practiced by _Behavioral and Brain Sciences_, unless an individual reviewer
    requests anonymity (in which case that reviewer's identity is not told to the
    other participants).

        Of course, the fruit of the BBS process is _public_ peer review (with
    all names and comments disclosed, and with author rebuttal). But any
    scientific publication is fair game for public scrutiny once it is in print.
    BBS' editor (about to step down after long service), is the distinguished
    Stevan Harnad, a productive scientist and innovator in the effective
    dissemination of scientific information.

        Obviously, plausible arguments can be mustered in favor of and in
    opposition to all three of these methods. Open review is interesting, I
    think, since it proceeds from the premise that the nub of the problem is
    accountability. That scientific debate is otherwise characterized by its
    "openness" offers another thread of argument in favor of open review.

        With respect to double-blind review, I think it is reasonable to ask
    whether anonymity can in fact be achieved. Typically, the paper is available
    on the Internet at the author's website and in many cases, reviewers have
    been selected according to criteria which tend to ensure that they would know
    who is working on what problem by what methods. Conversely, under all three
    protocols, referees write a prose narrative to explain their verdict, and
    many people have "signature" writing styles. (Fortunately, my own writing
    style is hard to spot, much as Philippe Smets' is.)

        I think it is clear that "anonymity failure" would be especially likely
    in a small community. If that community's members are also adepts at
    inference under uncertainty, then unless they are deluded, "failure"
    predictably collapses into "sham." Even in a large community where median
    inferential prowess is doubly uncertain, like the AAAI/IJCAI general
    population which lately practices double-blind reviewing, the collapse into
    "sham" can be only too evident.

        Best regards to all.

                                                            Paul Snow



    This archive was generated by hypermail 2b29 : Thu Nov 15 2001 - 09:49:37 PST