Best Practices for the Care and Feeding of a Conference Organization
Download
Report
Transcript Best Practices for the Care and Feeding of a Conference Organization
IBM Research
Best Practices for the Care and Feeding of a
Program Committee, and Other Thoughts on
Conference Organization
Fred Douglis
IBM T.J. Watson Research Center
Best Practices
7/1/2016
© 2008 IBM Corporation
IBM Research
WOW … It’s WOWCS
First reaction: what a great idea!
Second reaction: will anyone actually submit?
– Immediate fear: spend time writing something up, only to find
out it’s been canceled
Final reaction: duty
– I’ve chaired USENIX’98, USITS’99, and a few other
conferences
– Most went well but there were some issues
End goal: guidebook (wiki) for program/general chairs
– https://wiki.usenix.org/bin/view/Main/Conference/CollectedWisdom
2
Best Practices
7/1/2016
© 2008 IBM Corporation
IBM Research
Wiki
Goals
– Living document
– Allow discussion
Two types of info eventually
– Guide for new program chairs or conference organizers
• Review management software
• Best practices
– Ideas for new ways to do things
• How well do rebuttals work? Ratings? Bidding?
3
Best Practices
7/1/2016
© 2008 IBM Corporation
IBM Research
Problem 1: Bad Reviewers
Not everyone is a good PC member
– Some never do anything
– Some do really crummy reviews
– Some are too argumentative, dominate the rest of the PC, or
have other personality issues
How to find who is a known bad reviewer?
– Past experience
– Word of mouth… but people tend not to know to ask
How to expand word of mouth?
– Reviewer database?
4
Best Practices
7/1/2016
© 2008 IBM Corporation
IBM Research
Problem 2: Inexperienced Reviewers
Always want to bring some “new blood” into a PC
– Search for repeat authors who haven’t served (need a script to
scour Google Scholar for this)
– Caveat: not everyone knows what to expect and may not be
cut out for it. Don’t take too many people you don’t know if
you can help it
Tips for ensuring the best PC
– Set expectations early (# reviews, timing, … no surprises!)
– Have multiple deadlines: force people to miss early rather
than all at once just before the decisions
– Help with calibration (average scores of reviewers compared
to the scores of their peers on same papers)
5
Best Practices
7/1/2016
© 2008 IBM Corporation
IBM Research
Problem 3: PC Composition
Avoid inbreeding, or the appearance of it
– Overlap from year to year
– Institutional overlap
There are enough people who will have published at an
established conference that a chair should draw on them
rather than outside the conference community
– Best fit
– Reward participation
– Avoid problems with calibration
6
Best Practices
7/1/2016
© 2008 IBM Corporation
IBM Research
PC Bells and Whistles
Options for running a PC
– Rebuttals (Seem to be a nice idea, but add lead time)
– Reviewer ratings (hard to calibrate; don’t want to insult)
7
Best Practices
7/1/2016
© 2008 IBM Corporation
IBM Research
Running a Conference
Sponsorship
– Some institutions make it a pleasure to run a conference, and
some make it much harder
– Doing it alone has risks (e.g., liability): who “owns” the
conference risks?
Setting expectations
– It’s hard to know just how many submissions a new conference
will get: better to be swamped (and add PC members) than miss
all your targets
– Publicity is amazingly critical (who knew?)
Scheduling
– Watch for conflicts and hope the others watch too
– Consider the impact of rejected papers from one conference
being routed to another. (Feed reviews?)
8
Best Practices
7/1/2016
© 2008 IBM Corporation
IBM Research
Backup
9
Best Practices
7/1/2016
© 2008 IBM Corporation
IBM Research
Problem 4: (Self)plagiarism
Simultaneous submission seems to be a big
problem
– Least publishable units
– True plagiarism
We need a better mechanism for detecting
– Serendipity doesn’t cut it
10
Best Practices
7/1/2016
© 2008 IBM Corporation
IBM Research
Concrete Proposals/Challenges
Plagiarism
– Neutral agent to collect papers and analyze for overlap
– Get multiple organizations to buy in
Reviewer database
– Neutral agent to collect feedback, and if someone gets
multiple negative reports, somehow blacklist
See Internet Computing columns Sept-Oct/NovDec 2007
11
Best Practices
7/1/2016
© 2008 IBM Corporation