Rebooting AI Research Publications Strategy: Continuing the Vardi-Halpern discussion, and how to address reviewer burden

The societal and technological impact of AI is likely to be profound, garnering unparalleled attention on social media and news outlets. There is an urgent need to understand precisely how the recent advancements in AI will impact various fields, ranging from education to the biological sciences. Unfortunately, the dissemination of AI research is currently facing massive bottlenecks.

AI papers are getting submitted to conferences in large numbers, and program committees are struggling. Conference venues in computer science are traditionally treated as archival publications. But as argued by Moshe Vardi, although conferences promise predictability, in terms of submission and decision timing, this is an illusion as papers get bounced from one venue to another until they arbitrarily find a reviewer pool – aligned in spirit and mind perhaps – which accept the paper. Joseph Halpern has argued for some changes within ACM, based on a two-step process. However, both viewpoints ignore the challenge of finding appropriate reviewers for journal submissions, and get timely reports. Besides, conference venues provide an excellent model for dissemination of quickly developing areas.

We advocate a comprehensive strategy consisting of three parts, all supported by a two-step process: (a) conferences for urgent reports, (b) public-facing reviews (or at least accessible to handling editors) for rejected papers, and (c) a timely review process for journals, inspired by but hopefully not as formidable as, the tenure approval process, where authors seek out confirmed reviewers.