[IJCAI-PRICAI 2020] Discussions about summary reject The summary-reject decisions have provoked intense reactions from some authors. I understand the disappointment that can come from such a notification, especially for students submitting their first paper. The whole process was probably not clear enough for authors, despite the CFP, the FAQ and the video posted on social media. I will try to explain a bit more. The tremendous increase in number of submissions to generalist AI conferences has led to a significant decrease of the reviewing quality because the reviewer resource is not infinite. During last IJCAI, a panel has been organized, during which we agreed that we cannot continue with the same process anymore. Summary reject is the most common idea proposed fo reducing the reviewing load. Summary reject can be implemented in several ways. A possible scenario is to ask one senior PC to spend about one hour on each paper assigned to him/her and to decide whether the paper deserves to enter the full review process or not. The advantage is that the senior PC writes a short review explaining the decision. The drawback is that the future of the submission exclusively depends on whether a single individual likes the paper or not. Another scenario is to send the paper to a large enough number of senior PCs, in our case between 7 and 10, and each of them spends 1/10 to 1/7 of one hour on the paper and votes to say whether the paper has chances of being accepted to IJCAI. The advantage of this scenario is the much higher robustness to individual opinions. The drawback is that we cannot ask a senior PC to write a thorough review for each of them. I chose the second scenario. Now that I have clarified the why and how of summary reject, let me answer a few of the comments I have heard or received. 1. "Desk rejecting 2000 papers is ridiculous". --> Summary reject is *not* desk reject. Senior PCs were explicitly asked to assess the value of the contribution, not the format or the fact that the paper fits the scope of IJCAI (which already led to 346 desk-rejected papers). After the votes, each paper was processed by an area chair, who analyzed the votes and the brief comments from the senior PCs, and in many cases the area chair had a look to the paper before proposing a decision. 2. "You killed IJCAI, whose main purpose is to give authors a feedback on their current research, especially to students". --> Unfortunately, given the large scope and the number of submissions, IJCAI can no longer afford being a forum where to seek feedback about our work. For such feedback, conferences with more focused scopes are probably more appropriate. 3. "You could at least send us the comment of the area chair". --> This was my original plan. But after having checked many of the papers, I understood it was not a satisfactory solution. Papers with many negative votes usually do not have any comment from the area chair. Borderline papers have two-three lines comments. I did the test on Saturday: I sent the comment from the area chair to the first authors who sent me an email to ask for it. This didn't close the debate because two-three lines are not enough to convince an author about the weaknesses of their submission. 4. "The result of this selection is random". --> As it is the first time we implemented this, I cannot claim the validity of the experience. A careful statistical analysis will be conducted at the end of the review process. The pros and cons of the approach will be drawn for future IJCAIs. (Just to break a bit the myth of the 'three reviews' accuracy: https://cacm.acm.org/blogs/blog-cacm/181996-the-nips-experiment/fulltext). 5. "It is so rude and unfair to receive a notification without any feedback for a hard work of several months. If I knew I would have never submitted to IJCAI". --> Though I understand the feelings it can raise, let me quote the CFP: "By submitting a paper, authors acknowledge that they are aware of the possibility of receiving a summary rejection notification within one month after submission without any written review. Authors also accept that all decisions of the program committee (i.e., summary reject, standard reject, or accept) will be final and cannot be appealed." I sincerely hope that these clarifications will help some authors to better understand how and why things have gone like that for their paper. Best wishes, Christian Bessiere Program chair of IJCAI-PRICAI 2020