Saturday, December 10, 2016

Less Exhausting, More Effective: Flipped Conferences*



If you want to get to business immediately, go to the 4th paragraph please.

--

I wrote the first draft of this like mid-August and wanted to post it shortly after. However, I kept delaying. It has been a tough year for the world in general. I, being a Turkish citizen, am like a non-life-threatening train-wreck since the coup attempt in Turkey (which happened midsummer). I still go to work regularly, act normal as much as possible, and do my best. But some days sleeping or getting out of bed is an extreme challenge. Terrible things happen in this world all the time, but some have the power to impact you more personally than others. If I consider my hopes and dreams as a small garden, some people burnt that garden completely midsummer. Now whenever I try to grow some new flowers there, some people kill them with a hammer.  It will take some time to restore that garden, but we are working on it. Naturally, being in this state turns something like posting about how to improve our conferences into a trivial or unimportant errand. Hence, my lame excuse to you for waiting too long to post this.

--

On the way back home after SIGMOD this year, I met my Partner in Crime, Helena Kotthaus, for dinner in Palo Alto. She was in Bay Area that week for a conference as well and was going back to Dortmund in a couple of days. We were both extremely exhausted because of the conferences we attended. The main conclusion of that night was that we were officially not-young anymore. We intended to go to a bar after dinner, but instead, first stopped at an independent bookstore (Bell's Books) and hung out there till we were told that it was closing time, then went to have coffee at Coupa Café and discussed how we could change our conferences to be less exhausting and more effective.

--

There has been various discussions about how we should change the way we run conferences in the recent years in our community. The premier conferences in our field have grown too big. It is very exhausting to keep track of all the papers and demos you are interested in as many sessions with many talks run in parallel, make sure everything goes well with your own presentation(s) or demo(s), and also meet/catch-up with people in your community.

While it might seem desirable to depart from few larger conferences and have many smaller workshops instead, the two serve different purposes. Workshops are extremely effective to gather people from a narrow research area. Since there is a single track in most workshops and the duration is typically shorter than the duration of a conference, the gurus of that research area stay in the workshop venue to follow majority of the talks. Therefore, workshops can actually provide higher recognition and networking opportunities, especially, for students. On the other hand, workshops tend to be organized by a smaller group of people, which may or may not change from one year to another. Therefore, they run into the danger of targeting the same group of people year after year sometimes excluding people from other fields or the newer generation.

Conferences bring together people from wider-variety of fields and prevent us from thinking that our own field is the center of the universe. They open up more opportunities for cross-field collaborations, and hence, more interesting research topics. In addition, they can teach us about fields that we do not have a good background on.

What can we do to make our conferences less exhausting and more effective then? We thought about adopting a structure similar to flipped classrooms, flipped conference, where the talks for accepted papers are given before the conference and the conference becomes a venue mainly for panels, discussions, demos, tutorials, and poster sessions.

The majority of the time in a conference goes to talks. In last SIGMOD, each session had seven papers to be presented. Even though the talk times were short, it was extremely hard to follow most of the talks in a session due to their density. At the end of the conference, one of the questions in the conference survey was whether to have talks for all papers or some specific subset of papers. We find it unacceptable to just pick a subset of papers for talks. Each accepted paper should have as equal opportunity as possible to be represented in a conference. Otherwise, they shouldn't be accepted. Therefore, we should focus on eliminating the burden of having too many talks even if many papers get accepted in a premier conference.

In a flipped conference, the authors of the accepted papers would record their talks before the conference and upload them to conference website (maybe as early as or shortly after the camera-ready deadline). The conference attendees can then watch these talks beforehand picking time-slots where they can better focus on the talks they are interested in or skim them during their flight to the conference location.** Then, they can mark the specific papers they would like to further dive in, read the paper itself in detail or target the authors of that paper for a fruitful discussion during the actual conference. One might say that lightning talks (as we do in some workshops) or short videos (as they do in SIGGRAPH) at the beginning of the conference would be sufficient for attendees to better navigate themselves in a conference. However, these don't always give a good idea about how interesting a paper is for you.

Giving the talks prior to the conference would be helpful for paper authors as well. The talks that are scheduled for the last day of the conference or in parallel with a talk from a prominent member of the community tend to get less attention. This way, we give each author, especially students, higher chances of being noticed.

The conference would be mainly composed of several poster and demo sessions that are less dense than the ones that we typically have, so that one can pay attention to all the posters and demos in a room. This would allow more interactive discussions among the conference attendees. In the meantime, the panels and tutorials would continue to take place as usual. One can also add discussion sessions on sub-fields in the community, where people discuss the accepted papers in that field. This would allow including more papers in a session, different sessions can have different number of papers, and some papers can be included in multiple sessions if they span multiple sub-fields.

Overall, flipped conferences can lead to a more interactive, less exhausting, and more beneficial conference experience for all attendees.


* I know that we are not really inventing this term. The idea of a flipped conference have been suggested by some folks already (e.g., here, here, here, ...), even though the implementation details are different from suggestion to suggestion.

** Initially, I suggested uploading these videos just a week before the conference would be enough. However, My Twin Sister, Duygu Ceylan, thinks that people might not allocate time to watch videos before the conference. Therefore, I thought uploading them earlier (like closer to camera ready deadline) might be better. Though, she confirmed that the premier conferences in the field of computer vision and graphics (her fields) are also quite exhausting and sometimes papers can get stale by the time they are presented at the conference. 

No comments:

Post a Comment