By David Eppstein & Vijay Vazirani
No, not to make theoreticians rich! Besides, who will buy your papers anyway? (Quite the opposite, you will be lucky if you can convince someone to take them for free, just for sake of publicity!) What we are proposing is a market in which no money changes hands – a matching market – for matching papers to conferences.
First, a short preamble on how the idea emerged.
Preamble (by Vijay): Soon after my recent Simons talk on Matching Markets, I sent its url to Al Roth. Obviously, I wasn’t expecting a return email. However, the perfect gentleman and ultimate scholar that Al is, he did reply, and mentioned that he did not like my “definition” of matching markets and said, “I guess I would say matching markets are markets because they aggregate information that is held by the participants, which is what markets do (even if they don’t use prices to do it..).” This hit me like lightening from the sky – suddenly it crystallized the innate intuition about markets which I had formed through work on algorithmic aspects of markets! I thanked Al profusely and added, “This definitely helps in me get the right perspective on the notion!”
About a week ago, while updating my talk for a seminar at Columbia University, I included this beautiful insight in it and then a thought occurred: Each PC meeting involves aggregation of information from a large number of agents: PC members as well as external experts. Hence, isn’t a conference a matching market? Excitedly, I sent this question to Al. He replied, “… the conference process, matching papers to conferences, is a market and a particular conference might be a marketplace … ”
When I returned home, my esteemed colleague, David Eppstein, stunned me by declaring that he had thought of a market relevant to our field in which no money changes hands. I immediately knew he was thinking of the conference process. But he got to it out of the blue … and not the long process it took me!
Back to the idea: In the past, matching markets have brought immense efficiency and order in allocation problems in which use of money is considered repugnant, the prime examples being matching medical residents to hospitals, kidney exchange, and assignment of students of a large city to its schools.
At present we are faced with massive inefficiencies in the conference process – numerous researchers are trapped in unending cycles of submit … get reject … incorporate comments … resubmit — often to the next deadline which has been conveniently arranged a couple of days down the road so the unwitting participants are conditioned into mindlessly keep coming back for more, much like Pavlov’s dog.
We are proposing a matching market approach to finally obliterate this madness. We believe such a market is feasible using the following ideas. No doubt our scheme will have some drawbacks; however, as should be obvious, the advantages far outweigh them.
First, for co-located symposia within a larger umbrella conference, such as the
conferences within ALGO or FCRC, the following process should be a no-brainer:
1). Ensure a common deadline for all symposia; denote the latter by S.
2). Let R denote the set of researchers who wish to submit one paper to a symposium in this umbrella conference – assume that researchers submitting more than one paper will have multiple names, one for each submission. Each researcher will provide a strict preference order over the subset of symposia to which they wish to submit their paper. Let G denote the bipartite graph with vertex sets (R, S) and an edge (r, s) only if researcher r chose symposium s.
3). The umbrella conference will have a large common PC with experts representing all of its symposia. The process of assigning papers to PC members will of course use G in a critical way.
Once papers are reviewed by PC members and external reviewers, each symposium will rank its submissions using its own criteria of acceptance. We believe the overhead of ranking each paper multiple times is minimal since that is just an issue of deciding how “on-topic” a paper is – an easy task once the reviews of the paper are available.
4). Finally, using all these preference lists, a researcher-proposing stable matching is computed using the Gale-Shapley algorithm. As is well-known, this mechanism will be dominant strategy incentive compatible for researchers.
With a little extra effort, a similar scheme can also be used for a group of conferences at diverse locations but similar times, such as some of the annual summer theory conferences, STOC, ICALP, ESA, STAC, WADS/SWAT, etc.
Like this:
Like Loading...
Read Full Post »
Turing's Invisible Hand 