Folks,

For a bunch of decision theorists, we sure do fall easily into 
alternative-focused thinking!!!

Let's try some value-focused thinking instead.  What are the values 
we are attempting to serve with the structure of our review process? 
Values I have heard people articulate are:

  - Quality. We want the set of accepted papers to be of uniformly high quality.
  - Fairness.  We want authors to have a fair chance of having their 
work accepted on its merits whether or not they are well known to the 
insiders in the UAI community.
  - Openness to new ideas.  We want outsiders with interesting new 
ideas to have a fair hearing and a chance to have their papers 
accepted.
  - Cost in time and effort.  We do not want the process to be overly 
burdensome on authors, reviewers, and the program chairs.
  - Cost in dollars.  Obviously, cost in time and effort to volunteer 
reviewers and chairs can be reduced by paying professionals to manage 
certain aspects of the conference.  This cost would have to be borne 
either by conference registrants, by our corporate sponsors, or by 
charging dues to join AUAI.

Can anyone add to the list?

We might consider creating a matrix with major values as the rows and 
options as the columns.  One option obviously is no change to current 
processes.  Other options would involve different modifications aimed 
at improving our score on some of the rows, but possibly at the 
expense of the score on other rows.  The proponent of a given 
modification might describe his/her modified process in enough detail 
to provide an assessment of how it would score on each of the rows. 
In particular, can proponents of double-blind reviewing describe 
processes that we could implement at relatively low cost and would 
provide significant improvement on some of the key attributes of 
value?

Then we could focus our discussion on what our relative weights are 
as a community on the different attributes of value, and on our 
assessments of how well each of the options under consideration 
scores on the different attributes of value.

Once we have structured the problem, we might want to consider taking 
a survey to assess the community's views on the subject.  Putting a 
survey form up on the web is a relatively easy matter.  Of course we 
wouldn't have a scientific sampling process, but it would still be a 
more reliable indicator of community views than the opinions of those 
of us who tend to pipe up on email listservs.  :-)

Kathy


Reply via email to