Journals for certification, conferences for rapid dissemination

Joseph Y. Halpern, David C. ParkesCommunications of the ACM 2011 · DOI

Authors: Joseph Y. Halpern and David C. Parkes (2010)

Abstract

In the field of Computer Science (CS), selective conferences have historically served as the primary venues for both prestige and the rapid dissemination of research. However, this unique reliance on conferences for "certification"—the process of quality control and peer validation—has created systemic pressures that may be hindering the quality and reach of the discipline. This viewpoint paper argues for a fundamental shift in CS publication culture: decoupling the role of journals as the primary certification authorities from the role of conferences as venues for networking, publicity, and rapid community feedback.

The Anomaly of Computer Science Publishing

Most scientific disciplines view journals as the "gold standard" for research, with conferences serving as informal gatherings for preliminary work. Computer Science is a notable outlier. Born from a need for rapid sharing in a fast-moving field, CS elevated selective conferences to a status of high prestige. While this model served the community well during its formative years, Halpern and Parkes argue that it has reached a breaking point, creating a "deadline-driven" culture that prioritizes speed over depth.

The Systemic Issues of Conference-Centricity

The authors identify three primary problems stemming from the current conference-first model:

  1. The Deadline Trap: When conferences are the only path to prestige, research becomes driven by submission dates rather than the natural completion of a project. This leads to rushed results, incomplete proofs, and a lack of iterative refinement.
  2. Fragmentation: To accommodate the sheer volume of submissions, the field has splintered into numerous niche conferences. This prevents these venues from acting as broad community forums and isolates researchers in silos.
  3. Interdisciplinary Friction: In fields like Biology or Economics, conference papers are not viewed as "final." This creates a "double-jeopardy" for CS researchers working at the intersection of disciplines; they often cannot publish their conference-proven work in top-tier journals of other fields because those journals consider the work to be previously published.

Decoupling Certification and Publicity

The core of the authors' proposal is the separation of two distinct functions: certification (the rigorous signal of competence and prestige) and publicity (reaching the wider community).

Under this framework, journals would handle the slow, rigorous certification process—checking proofs and allowing for iterative revisions. Conferences would be repurposed to focus on publicity, networking, and community building. This shift would allow for "continuous submission," eliminating the mad dash for conference deadlines and allowing for more polished, thoughtful research.

Learning from Emerging Models

The paper highlights two existing "experiments" within the CS community that offer a path forward:

  • The TOG/SIGGRAPH Model: The ACM Transactions on Graphics (TOG) publishes conference papers as special issues. This ensures that work presented at the conference has undergone journal-level review.
  • The PVLDB Model: The Proceedings of the VLDB Endowment (PVLDB) utilizes a "journal-first" approach where authors submit to a journal year-round. If accepted, they are automatically invited to present at the VLDB conference. This model removes the artificial pressure of a single annual deadline.

A Roadmap for a Sustainable Future

Halpern and Parkes outline a multi-step vision to transition the CS community toward a more sustainable publishing ecosystem:

  1. Public Archives First: All research should be hosted on managed archives like CoRR (arXiv) for immediate availability, ensuring that "rapid dissemination" is not tied to a specific event.
  2. Journals as Primary Certifiers: Journals should become the ultimate authorities for quality, operating without strict page limits to allow for deep technical discussion.
  3. Accelerating the Review Cycle: To make journals viable in a fast-moving field, the review process must be streamlined. This could include "fast-track" page limits for shorter papers and the sharing of reviews between venues to avoid redundant labor.
  4. Reviewer Reciprocity: To address the chronic shortage of qualified reviewers, the authors suggest a "cost" for submission: authors must commit to reviewing a proportional number of papers in return.

Conclusion

While the authors acknowledge the "Churchill view" of peer review—that it is the worst form of certification, except for all the others—they maintain that a journal-centric model is superior to metrics like citation counts, which are often influenced by fads. Shifting a decades-old culture will require significant experimentation and community-wide debate, but the authors argue it is a necessary evolution to ensure the long-term health and interdisciplinary impact of Computer Science research.

Summary updated

BibTeX
@article{halpern2011journals,
  author = {Joseph Y. Halpern and David C. Parkes},
  title = {{Journals for certification, conferences for rapid dissemination}},
  journal = {Communications of the ACM},
  year = {2011},
  doi = {10.1145/1978542.1978555},
}
Figure