Queueing to publish in AI and CS
3 hours ago
- #publication model
- #acceptance rate
- #CS conferences
- The common CS conference publication model with a fixed low acceptance rate is analyzed using a toy model.
- In an ideal scenario where authors resubmit indefinitely, the number of accepted papers remains constant regardless of the acceptance rate.
- When authors give up after a certain number of submissions, lower acceptance rates increase the pool size and reviewing load, disproportionately affecting average-quality papers.
- Lower acceptance rates lead to more abandoned papers, especially bad ones, but also significantly impact average papers due to bad luck.
- The model suggests that higher acceptance rates could reduce the pool of unaccepted papers without dramatically increasing absolute acceptances.
- Proposed changes include community feedback, faster reviewing experiments, mindful resource use, and openness to tested changes.
- The effective acceptance rate is higher than a single conference's rate due to resubmissions, and the quality of accepted papers improves with lower acceptance rates.