IJCAI2019 PC 回应审稿质疑：几十年都是这样，而且我们需要创新

雷锋网 AI 科技评论按：人工智能领域顶级学术会议之一的 IJCAI 2019 近日发布了论文录用名单并通知了作者们。IJCAI 2019 今年一共收到 4752 篇有效提交论文，最终的论文收录数量为 850 篇，接收率为 17.9%。

今年 IJCAI 的投稿总数比去年增加了约 37%，所以录用率随着降低到了不足 20% 也算情有可原。不过意外的是，收到录用或者拒稿通知后，许多作者都通过种种途径表达了自己的不满，包括对于打分、审稿人意见、Rebuttal 环节等等，似乎都有大量问题。一时间对于论文审稿过程的批评和讨论沸沸扬扬，其它媒体也进行了报道。

对于这样的舆情，IJCAI 2019 的 PC （program chair，程序主席）Sarit Kraus 给作者们发送了一封邮件，做了一些解释，也给出了一些他的个人观点，希望可以平息作者们的愤懑。雷锋网 AI 科技评论摘录并翻译如下。

When I sent the notification messages on Friday, I was happy to end more than overwhelming six months of working on the reviewing process of IJCAI 2019. On the other hand I was very sad since while I made the authors of 850 papers very happy, I made the authors of 3902 papers very upset and annoyed.  Even though I have many papers accepted to highly refereed conferences, I believe I received many more rejection messages than most of you. As a result, any complaint you have about the reviewing process, its unfairness, its randomness, I am sure I thought the same over the years. I am grateful to all the authors that haven’t sent me a message to complain, and I will try answering the messages of the few that did. From first glance, it will probably not change the decisions.

当我周五把所有的审稿结果都发送出去以后，IJCAI 2019 的长达六个月的繁忙的审稿过程终于开心地结束了。另一方面，虽然我让 850 篇论文的作者们很开心，但另外 3902 篇论文的作者们就一点也不开心，甚至有点烦躁；这也让我不怎么好受。虽然我自己也曾经有多篇论文被声名显赫的会议录用，但我相信我曾经收到过的拒稿通知也要比你们中大多数人更多。所以，大家对于审稿过程、其中的不公平、随机性等等的各种报怨，我相信我在过去的这么多年中也有和你们一样的感受。我很感谢所有没发出报怨之声的作者们，我也会尽量回复那些提出了质疑的声音。不过目前看来，这可能并不会改变审稿结果。

I tried hard to reduce the randomness of the process by decreasing the number of papers each PC and SPC received, by increasing the reviewing time period, and by trying to recruit good ACs/SPCs/PCs and reviewers. Almost all of them did a great job and I am very grateful. However, with 4752 submitted papers, it is extremely difficult. I believe we need to think and being innovative about the concept of highly refereed conferences.

为了减少审稿过程的随机性我做了很多努力，减少了每位 PC 和 SPC（Senior PC）审的论文数量，增加了审稿时间，尽量找了好的 AC（Area Chair，领域主席）/SPC/PC 以及审稿人。几乎所有人都做得很棒，我也非常欣慰。不过，我们毕竟收到了 4752 篇论文投稿，不管怎么样这都是很难的一件事。我觉得我们需要重新思考“高被引顶会”这件事并有所创新。

I strongly believe that conferences are events for hearing new ideas and meeting other researchers and discussing research. My best projects started at conferences. I don’t think they are currently good for evaluations for getting promotion and funding.  We can think about making our journals better. Currently we have a journal track—each paper accepted to AIJ or JAIR with no prior publication in a conference is invited to be presented in IJCAI. We can further limit the number of papers each person can submit to IJCAI or any other innovative ways you all can come up with. We all need to think about it.

我坚定认为会议是聆听新的想法、会见其它的研究者、讨论学术科研的场所，我自己的最好的项目就是从一次会议开始的。但是顶会目前对于获得升职、获得经费并不能起到什么帮助。我们可以考虑把期刊做得更好。目前我们有一个期刊 track，每篇 AIJ 或者 JAIR 期刊接收的论文，如果此前没有在会议中出现的，都会受邀在 IJCAI 展示。我们甚至可以考虑限制每位作者投稿到 IJCAI 的论文的数量，大家也可以一起想想其他的更创新的方式。我们需要共同思考。

In summary, I hope that all of you will come to IJCAI regardless of whether you have papers accepted or not. You can still submit papers to the IJCAI workshops or just come to meet friends and hear interesting and exciting talks.

总的来说，不管你的论文有没有被接收，我都希望大家全都可以来到 IJCAI 会议现场。你仍然可以把论文投稿到 IJCAI 的各个 workshop，或者就是来见一见朋友、听一听有意思的演讲。

Best regards,

Sarit Kraus

IJCAI 2019 Program Chair

对于这封邮件，有学者认为观点比较中肯，可以接受。雷锋网 AI 科技评论也认为，虽然会议在计算机科学领域是非常重要的学术成果评议平台，但是大批量往顶会投论文“灌水”的做法也已经带来了硕士生博士生申请门槛越来越高、参会门票难以购买、审稿人年轻化等问题，这都是需要领域内的学者们反思并设计改进方案的。

另外，AI 研习社近期有「顶会赞助计划」。参与 IJCAI 2019 顶会讨论小组活动，积累研值排行TOP 3，将获得 AI 研习社赞助的「直达顶会」的机票+酒店等参会费用，让你亲临大会现场，和大咖面对面~ 更多详情请查看小组页面 。