Date: Mon, 1 Feb 93 21:56:39 PST
From: Steve Minton <minton@ptolemy.arc.nasa.gov>
To: Jon Doyle
Subject: Editorial Board Invitation

Dear Jon,

    I am pleased to invite you to join the editorial board of a new
journal, the "Journal of Artificial Intelligence Research". The
primary means of distributing this journal will be over the internet.
The journal is being established by Jaime Carbonell, Dan Weld, Oren
Etzioni, Matt Ginsberg, Rich Korf, Ken Forbus, Bart Selman, Paul
Rosenbloom, and myself. I am a co-director (along with Oren and Dan)
of a non-profit corporation, The AI Access Foundation, that is being
set up to run the journal. First let me outline the basic idea behind
the journal, and then I'll give you the details regarding the
editorial board.

Electronic publication has the potential to radically alter the
way that scientific information in AI is disseminated. We intend
to establish a high-quality, refereed journal of Artificial Intelligence
that will be distributed electronically. In addition, each complete volume
of the journal will be available in a bound volume, marketed by Morgan 
Kaufmann. In establishing this journal, we have the following three goals:

1. Free access to scientific articles for individuals. At least for the 
   first year, the electronic version of the journal will be distributed 
   at no cost over the internet. (In the future, we may have to
   charge a minimal subscription fee, but hopefully not). 
   The noncommercial reproduction of articles for educational and 
   research purposes will be explicitly permitted.

2. Greatly reduced time to publication. (See below)

3. Exploring new capabilities provided by the electronic medium. Electronic
    publication opens up some interesting possibilities, including:
    A) online text searching 
    B) an electronic newsgroup associated with the journal 
       where published articles can be discussed and commented upon.
    C) enabling authors to publish appendices with pointers to new work.
    D) enabling authors to publish appendices containing code/data.
   We have definite plans for the first three. The fourth is
   contingent upon the resolution of certain legal issues, and so we
   will be holding off on this for a while.

The electronic journal will be organized into volumes and issues, as
with a traditional journal. As an issue is ``published'', it will be
advertised online (via newsgroups and email).  Articles will be made
available in postscript, although authors may submit additional
versions in other formats if they wish. (An ASCII text version will
also be available to facilitate text searching.)  Articles will be
distributed by FTP and by an automated email system. CMU has agreed
to serve as a distribution site. (Other universities have also expressed
an interest). The bound volume will be published semi-annually by
Morgan Kaufmann.

The primary responsibility of editorial board members will be to
review papers, thereby providing a consistent level of quality in the
reviewing process.  Ideally, papers will be reviewed by two members of
the editorial board and one outside reviewer. In order to
substantially reduce the time to publication as compared to
traditional journals in AI, we have established the following policies
for reviewing:

1. Reviewers will be responsible for returning their reviews 
   within 21 days after receiving the paper. (This might seem extreme, 
   but keep reading).
2. Before a reviewer is sent an article, he will be asked whether or not
   he has the time to review the article. Reviewers will be encouraged
   to accept the paper only if they can realistically guarantee
   a 3 week response. If an editorial member declines to review an article,
   this will NOT be considered anti-social, even if the reviewer has 
   declined on several previous occasions. (I am assuming that anyone who
   joins the editorial board will do so in good faith.)
3. Editorial members will review at most 3 papers per year. (Actually,
   I think that most members will review less than 3 a year).
4. Editorial board members will serve a two-year term. 

As one of the action editors for the journal "Machine Learning", I
have found that if you ask reviewers if they are interested in a paper
before sending it to them, it is rarely a problem getting it back on
time. Reviewers usually put an article on their "stack" and then get
to it sometime near the end of the reviewing deadline. As we all know,
the actual process of reading and reviewing an article is typically
accomplished over a few days, and thus a 3-week reviewing period
should be workable.  (I note that in fields such as biology, reviewers
are regularly expected to return their reviews in a week or two.) The
3-week review period will be extended if the paper is longer than
usual, or in the paper is submitted in the 6 weeks preceding a major
conference deadline (AAAI, IJCAI, or if the paper is in a subfield
such as ML or KR, the major conference for that subfield).
With these reviewing policies, authors can expect to hear about their
papers in approximately 4 weeks. If an author is willing to revise his
paper promptly, then a 3-4 month submission-to-publication period is
realistic. 

For the first year or so, I will be in charge of selecting reviewers
and handling the review process. (NASA is funding my time as executive
editor, as well as funding administrative help.)  Once the process has
been established and the kinks are worked out, we will transition to a
system where a small set of action editors will be selected to help
handle the review process. (This system has worked quite well for the
ML journal).

As you can see, the responsibilities of editorial board members will
be light. Basically, you will be asked to review up to three articles a
year, but almost certainly fewer. You will have the option of turning
down articles if you are too busy.

I sincerely hope you will accept our invitation to join the editorial
board. Although there are already some good journals in the field,
this journal presents some new opportunities. We have the potential to
significantly alter the way AI results are published.  This may not
only affect our field, but other areas of computer science may be
influenced as well by our success or failure.  It is an experiment
worth participating in. If we are successful, it will mean that
quality articles will be available for free (or at very low cost),
that you don't necessarily need to go to the library to look something
up, that articles can be published much more quickly, that programs
and data will be publishable along with articles, etc.

In addition to our goal of providing free on-line access to the new
journal, we also feel strongly about the necessity of its being very
high quality. A highly-respected editorial board will go a long way
towards establishing the initial credibility of the journal -- which
is particularly crucial in an experimental endeavor such as this -- as
well as helping to assure the long-term quality of the final product.
In choosing an editorial board, we are looking for researchers who are
not just "well-known", but who have made significant research
contributions. As someone whose contributions and technical judgement
are highly regarded, you would be doing us a service by joining the
editorial board.

I am looking forward to hearing from you, with a positive response I hope!
Please feel free to send me any questions, concerns or comments as well.

Regards,
Steve Minton