Howdy,
As far back as I can remember, the SIGGRAPH technical papers community has complained about the technical papers program. As I have gotten older I’ve begun to realize that *every* technical papers community complains about their program, whether it is the review process, acceptance rates, or the additional comparisons required by reviewer #2, etc. It seems just to be human nature. There is almost always a sense that the “grass is greener” and I am often reminded of the Churchill quote: “Democracy is a very bad form of government, but I ask you never to forget: all the others are so much worse.”
But, I do think that something has shifted over the last five years. Page counts and author lists have gotten longer; the number of hours of work to produce a SIGGRAPH Technical Paper is far greater than when I was a grad student. This trend is not, in and of itself, bad and is, in part, due to progress in the field; over the past decades the community has built up an enormous body of prior art for reviewer #2 to point to, and much of that work is available in open-source repositories. A very positive effect is that SIGGRAPH technical papers are highly novel, very rigorous, and contain thorough analysis; I think the quality of our papers rivals the top journals in all of science, as evidenced by the fact that ACM Transactions on Graphics (ToG) has the highest impact factor of any ACM journal that publishes novel research (apparently ACM Computing Surveys scores higher).
But, as with all things, there are downsides. I would argue that the increased amount of work required to go from research results to a paper published at SIGGRAPH has had some significant deleterious consequences.
First, we have shifted somewhat from basic research—gaining insights, running experiments, and solving problems—to extensive evaluation. Extensive evaluation is certainly important. But, I would argue that this is separate from contributions to basic research. In fact, I would argue that authors proposing a new idea are the very worst people to be evaluating that idea; no one is going to spend as much time understanding or tweaking the parameters of others’ methods as they will their own.
Second, the amount of work required to publish means that graphics researchers publish fewer papers, garner fewer citations, and are jumping ship to neighboring conferences that have just as much credibility in a tenure review, but require less work.
Over the years, I have heard many proposals to change the SIGGRAPH process: accept more papers, accept fewer papers, re-impose page limits (granted there was never an official limit, but in 2001 if you went over 8 pages, you must have made a *very* significant contribution; Hertzmann and colleagues “Image Analogies” was a major outlier at the time at 14 pages). But, one idea has recently started gaining traction: introducing conference papers / short papers / technical briefs. There is little consensus on what to call such a program and there are many other details to also figure out.
To be clear, the idea that is building momentum is not to create a “low-bar” SIGGRAPH paper. Whatever form this program would take it would have the same high level of intellectual merit and technical novelty. But, given page limits, papers may, for example, have only a demonstration of an idea, rather than a complete systems with an exhaustive comparison to prior work. As with conferences like CVPR, authors would have to choose which details and evaluations to include; they would be forced to focus on what they consider most important and what will likely benefit future readers. Philosophically, I would argue that in any optimization it is very difficult to find a good solution in the absence of constraints, but that once you start introducing constraints, e.g. page counts, you are forced to make decisions to honor the constraints and quickly converge to a (probably local) minimum.
While the review process would be rigorous and may include a rebuttal, it would not include a second review cycle or mandatory revisions, further reducing the workload for the authors and the committee. The door would remain open for publishing extended versions in a journal, and we may be able to work with the PACM on CGIT to invite papers to submit such versions, perhaps with reviewer continuity.
These are the broad-brush outlines of the sort of program gaining momentum. Let me mention some of the positive and negative potential consequences.
On the positive side, we may stop losing researchers to other communities and bring more researchers into the SIGGRAPH community. Since I joined the EC, I have seen plenty of graphs which show that SIGGRAPH submissions, and in fact all graphics submissions, are basically flat while CHI/CVPR/NeurIPS are skyrocketing. Sure, some of that is simply the fact that Machine Learning is the hottest thing in Computer Science right now, but you can publish Machine Learning papers at SIGGRAPH. More practically, with this new program the community can publish more papers and members of our community will appear stronger on all of the imperfect, but influential, metrics from G-number to publication count; resulting in more funding, more hiring, and more recognition. And, in my role as SIGGRAPH chair, the most important thing is keeping members of the community engaged and sharing their work in our publications.
On the negative side, this may draw some papers away from the regular Technical Papers program. Personally, I think this is okay; the work is still in SIGGRAPH, though in a different program. I think it would be important that authors decide on where to submit; maybe this new program has a deadline two weeks later than Technical Papers and it is disallowed to submit to both.
In my view, the potentially bigger downside is a decrease in submissions to our specialized conferences, which historically have hoped to pick up rejections from SIGGRAPH. One idea I have heard suggested was to allow papers accepted to this new program to give oral presentations at specialized conferences and posters at the SIGGRAPH and SIGGRAPH Asia conferences. Such an approach would presumably provide the specialized conferences with great content. Of course, it is also possible, though maybe counterintuitive, that a such a program would generate *more* submissions to our specialized conferences; while authors may be tempted to send rejected Technical Papers to ToG or hold them for the next upcoming SIGGRAPH or SIGGRAPH Asia, papers rejected from non-journal program may well find their way into the specialized conferences.
Of course, if we implement such a program, there would be all sorts of unintended consequences that we can only guess at now, just as there were when we made the decision in 2002 to include SIGGRAPH and SIGGRAPH Asia papers in special issues of ToG. There are also many questions that need answering before we could move forward, and that is why I am writing this essay, to invite the community to be part of this conversation.If you think this is a great idea, we want to know. If you think this is a terrible idea, we also want to know. If you have ideas on how this broad-brush proposal could be improved or how details could be filled in, we want to hear from you.
In this spirit we will be creating a survey and holding a town hall later this month to get feedback from the community. This page will be updated with more information as it becomes available.
Adam