Conference rejects submissions with undeclared Conflicts of Interest (COI) automatically found on DBLP instead of using them to choose reviewers. Why?

Question

I read some email about some paper submission:

Here are potential COIs [Conflicts of Interest] found by our automated tool, which uses DBLP to detect recent coauthorships: [Alice, Bob]. Per [conference name] submission guidelines, it is the full responsibility of all authors of a paper to identify and declare all COIs, and we reserve the right to desk-reject submissions with undeclared conflicts or spurious conflicts.

Why would a conference desk-reject submissions with undeclared conflicts based on COIs automatically found on DBLP instead of using these COIs to choose proper reviewers (e.g., as CVPR 2024 does)?

(The dblp.org website is a database of authorship information for computer science publications.)

Answer 1

You/They did not follow the rules and got sanctioned for it.

Answer 2

While the answer by Thomas Schwarz is correct, I guess a deeper question is why the conference would even ask the authors about COIs if it has a way to automatically extract them.

The reason, quite simply, is that not all COIs are visible from DBLP. DBLP does not index all papers, and there are other types of COIs that are not directly observable from co-authored papers. In practice, it's almost impossible to get around authors self-declaring COIs, and it's crucial to get them to do so carefully and truthfully. I suspect the conference's somewhat draconian response is a reaction to that - authors failing to declare obvious COIs is not a light matter, as it calls into question which other COIs they failed to declare (which are not visible from DBLP).

Answer 3

A conference needs to proceed in a timely manner. The conflict of interest are needed to select reviewers. If a reviewer is asked to review (at a top conference they might have already agreed to review N papers at this time) and finds a conflict of interest, the reviewer will inform the chair of the program committee and the paper needs to be reassigned to another reviewer. What is easy for a small conference with less than one hundred submissions becomes an administrative nightmare with more than a thousand submissions.

Even with a small conference, there is no time to go out and start recruiting reviewers after submissions. That is why there are program committees, which might or not might give discretion to assigned reviewers to assign to sub-reviewers (in their research group) or recruit other sub-reviewers outside. Usually, shortly after the submission deadline, reviewers are invited to bid on papers (in which they declare their confidence in being able to do a good job at reviewing for a particular paper). Papers with a conflict of interest are usually already excluded for a particular reviewer from bidding. During this phase, reviewers can also state conflicts of interest. After the bidding phase (usually a few days), a typically automatic assignment of papers to reviewers is made, giving reviewers a relatively short time window to prepare the reviews. For conferences with a very large number of submissions, this process needs to be modified. Instead of bidding on papers, the conference might choose a different mechanism, such as assigning papers and reviewers to sub-topics.

Not declaring conflicts of interests interferes with this process. It is not the task of the conference organizers to do the conflict of interest declaration. Automatic tools are actually dangerous. People publish under different names, change affiliations, and there are even people with the same or similar names in the same field. The current process has enough informal safe-guards to deal with difficulties arising from the difficulties of identifying conflicts of interests.

There is also a concern in the community about various forms of manipulations. There have been cases of groups of scientists helping each other to get published by manipulating the peer review process. Openly declaring conflicts of interests is our duty as submitters in a peer-reviewed conference.

Answer 4

To me, this sounds like the conference doesn't trust their own reviewers to indicate conflicts of interest.

The automated system is going to be terrible at detecting legitimate conflicts of interests because conflicts based on relationships or shared non-publishing interests will not appear. This will be particularly problematic for conflicts based on rivalries or disputes, because professional enemies are unlikely to publish together.

If the field is one in which multi-author papers are common, the authors are also likely to resent the burden placed on them. I feel annoyed enough when I have to put hundreds of co-author names on an NSF conflict on interest forms and would certainly not want to do so for a conference.

The only potentially good thing that I see is that "we reserve the right to reject" is not the same as "we will reject", so it is possible that the conference is using it as a detector for egregious cases rather than an obnoxious demand for perfection.

Bottom line: the conference appears to distrust its reviewers honesty.