AAAI

AI Submissions Surge Past 32,000, Pressing for Review Accountability

The submission count at AAAI has climbed past 32,000, with just a day left before the deadline.

4 min readMachine Learning

A submission number in the low 32,000s at AAAI, with a full day still on the clock, is the kind of number that makes you pause. It is not just a data point; it is a signal about the sheer scale of the research community's appetite and the pressure cooker environment we have built. When a researcher looks at that ticker and wonders where we are heading, they are not complaining about the volume of work. They are pointing at a systemic issue: the machinery of evaluation is groaning under the weight of its own success. We have optimized for the volume of submissions while quietly accepting a review process that often feels like a black box, where the difference between a paper being withdrawn, rejected, or accepted can hinge on a single, often unaccountable, reviewer's mood.

This feeling of powerlessness is not isolated to the AAAI submission counter. It is the same anxiety that surfaces when we see Neurosurgery Match Requirements Highlight Growing Pressure on Medical Students or when we read about the shifting, confusing demands in Navigating AI/ML Job Requirements: A Shift in Expected Skills. In each case, the individual is asked to perform at a high level within a system that offers little transparency in return. The researcher's specific ask, that reviews and names for withdrawn or rejected papers be made public, is not about public shaming. It is about accountability. It is about creating a feedback loop where reviewers know their words carry weight beyond an anonymous decision letter, and where authors can learn from a process that currently feels more like a lottery than an evaluation.

We have to be honest about what is being proposed and why it matters. Making reviews public for rejected or withdrawn papers is not a trivial change; it would alter the incentive structure for everyone involved. Reviewers might be more measured, perhaps more constructive, if their name was attached to their critique. Authors would receive the gift of actual, actionable feedback instead of a generic paragraph that reads like it was generated by a template. This is the same principle that makes Unidentified Individual's Data Leaks Spark Debate on Online Anonymity so relevant: anonymity has a cost, and in the academic review process, that cost is often paid in the currency of rigor and fairness. We are not suggesting that every review should be a public spectacle, but a simple, searchable archive of past reviews for non-accepted work would be a powerful tool for demystifying the process.

Our take is straightforward: the community is mature enough to handle this level of transparency. The fear is not that bad reviews will be exposed, but that we will continue to accept a system where a single, unaccountable opinion can stall a research direction for months. If we want to avoid the feeling that we are just feeding numbers into a machine, we need to demand that the machine gives something back. A concrete step would be for the next major conference to run a pilot program, publishing anonymized reviews for a random sample of rejected papers. It would not solve everything, but it would be a start. The question is not whether we can handle the truth about our submissions; it is whether we are brave enough to demand it. We should be.

From Machine Learning

Recently submitted my abstract and the submission number is 32xxx. With still a day to go, I just wonder where are we heading.

Hope these conferences at least start making the reviews and names public for the withdrawn/rejected papers. So that people atleast take that accountability

Read the original at Machine Learning