Reflections on the FAccT 2026 Review Process

Welcome to the 9th Annual ACM Conference on Fairness, Accountability, and Transparency! We look forward to gathering with you in Montréal soon! As Program Chairs, we wanted to explain how papers were reviewed and selected, so that the community has transparency about the process and to reflect on possible improvements for future years. The biggest innovation this year was the addition of a revise and resubmit process. This raised new opportunities and challenges as outlined below. Two other major changes were the requirement of a new Generative AI Usage Statement and the adoption of the OpenReview platform for submission and reviews.

First, the numbers: This year FAccT received a record 985 full paper submissions, up 21.3% from the 812 submissions in 2025. Reviewing this many papers required the collective effort of 640 reviewers and 126 Area Chairs (ACs). In total, 325 papers were accepted for publication, for an acceptance rate of 33%. We view the increase in accepted papers over previous years as a reflection of the high quality of paper submissions and the benefits of the revise and resubmit process. We are excited about the work that will be shared in Montréal and the dialogue and discussion that will emerge from these papers.

FAccT’s strength arises from its deep interdisciplinarity, with participation from researchers, practitioners, and policymakers from many different fields contributing to advancing research on responsible, safe, ethical, and trustworthy computing. As Program Chairs from a variety of disciplines who have done interdisciplinary work before, we were committed to the vision for a “big tent” laid out in the FAccT Strategic Plan. The primary objective of the FAccT review process was to assemble a high-quality interdisciplinary program, while also providing authors with a fair process. At the same time, as prior Program Chairs have noted, creating a cohesive, inclusive, and consistent review process across disciplines continues to be an ongoing challenge for our community each year.

The process proceeded as follows:

Submissions and desk rejects: Authors submitted papers by January 13, 2026. The first step in the review process was to desk reject papers that did not comply with submission requirements, such as anonymization and length limitations. At this stage 52 papers were desk rejected.

Paper assignments: 933 papers were then assigned to ACs and reviewers through a process that weighted bids on papers from reviewers and ACs, paper (and reviewer) focus areas, and reviewer expertise based on publications accessible via their OpenReview profile. The average review load was 4.4 papers per reviewer and 7.4 papers per AC.

First review: Each submission was assigned to an Area Chair (AC) who oversaw three independent and anonymous reviewers. At this stage, reviewers submitted reviews with initial recommendations. Preliminary reviews were sent to authors on February 20.

Rebuttal: After initial reviews were posted, authors were given three days to write up to a 5,000 character rebuttal to raise any critical concerns or factual errors with the reviews and clarify key misunderstandings of reviewers. Reviewers and ACs then discussed the paper and rebuttal amongst each other (in a forum not visible to authors) to adjust their reviews and scores as necessary.

First Round Decisions: ACs then considered the reviews, rebuttals, and discussion, and wrote meta-reviews explaining their decisions to either accept (N=109), revise-and-resubmit (N=304), or reject. Papers with a revise and resubmit decision had three and half weeks to submit revised papers to address reviewers’ and ACs’ feedback.

Re-reviews: Revised papers were allowed an extra page of content to meet the meta-review’s recommendations, as well as the opportunity to post a cover letter explaining changes. The review team had slightly over a week to review revised drafts and make recommendations. Considering the reviewers’ discussion, and in collaboration with PCs where requested, ACs posted the 2nd round of accept or reject decisions on April 15, 2026.

As we reflect on this review process, there are several aspects of the review process we believe went well:

Addition of revise and resubmit: This was the first year FAccT included a revise and resubmit option in the review process. The goal of revise and resubmit was to recognize submissions that showed promise but required further development or clarification that could be accomplished within a 3-week revision period. This goal was met, as many papers improved significantly with the additional guidance and time, and authors expressed appreciation for the opportunity. 304 papers received a revise and resubmit decision, and of those, 71% (N=216) were ultimately accepted. We recognize that authors whose papers were ultimately rejected may have felt confused or disappointed that their extra work did not yield an acceptance. As noted below, we have further suggestions for setting realistic expectations about this phase of the review process in future years.

Detailed guides and responsive communication: We strove to provide detailed instructions about the review process and to keep ACs, reviewers, and authors updated at each step of the process. Building on the work of the 2025 PC, the FAccT website was updated with detailed instructions about submission requirements and roles of responsibilities of authors, reviewers, and ACs. Emails to all the stakeholders were sent regularly with clear directions about expectations and reminders about deadlines. In addition, we recognized that many authors were using platforms and templates that were new to them, and thus we posted guidance on a blog, links to sample submissions and templates, and responded to numerous author emails seeking clarification.

Interdisciplinary inclusion: We are pleased that the selected papers continue FAccT’s tradition of representing a range of disciplines across the six focus areas: evaluations and evaluation practices (115 papers listing this as their primary focus area) experiences and interactions (78 papers); law and policy (30 papers); normative foundations and implications (30 papers); power and practice (40 papers); and system development and deployment (32 papers). We took steps this year to circulate the call for papers to new audiences. Submission requirements allowed for both LateX and Word document submissions. Papers could be submitted as archival and non-archival (allowing authors to present papers at FAccT without publishing in the conference proceedings if their field requires other publication venues). Reviewers and ACs were invited and selected across all 6 focus areas. As PCs, we also represent different fields, and each of us work in interdisciplinary contexts. This diversity allowed us to jointly make policy decisions as they arose with a broad range of stakeholder interests in mind. We worked collaboratively with each other and with ACs and reviewers, and we devoted many hours to overseeing the process, answering questions, and coordinating with ACs.

There are also aspects of the review process where we believe there is room for improvement:

Need for a longer timeline: This year’s timeline was compressed to include the time for a revise and resubmit cycle, and to have an earlier final acceptance decision to allow ample time for visa applications. The call for papers was issued in mid-October. Submissions were due on January 13, and final decisions were issued on April 15. For reviewers, there were two rounds of reviews with quick turnarounds. These were heavy workloads in short time periods, and sometimes, the reviews had insufficient detail to provide ACs and authors with adequate feedback. Some authors were confused by positive reviewer scores that were followed by AC recommendations to reject. ACs were empowered to make accept or reject recommendations based on their own expertise and assessment, incorporating reviewers’ scores and recommendations, but not solely basing their recommendations on them. However, additional time might have allowed for more discussion between the reviewers and Area Chairs with additional input from the PCs, in addition to the oversight and calibration we engaged in. As PCs, we strove to give all stakeholders reasonable time for their tasks, but we were mindful that decisions needed to be issued with adequate time to allow authors of selected papers to obtain visas and other travel documentation as necessary.

Larger pool of reviewers and ACs: This year, we were able to recruit 640 reviewers and 126 ACs. To build these pools, we reached out to prior reviewers and ACs, required authors of paper submissions to sign-up to serve in a review capacity, circulated the call for reviewers and ACs on the FAccT website and in social media postings, and reached out individually to qualified colleagues. Still, we came up shorter than we had hoped, and the reviewer pool did not grow commensurate with the growth in submissions (a challenge that other conferences are facing now as well). With 933 papers to review after desk rejects (i.e., nearly 2,800 unique reviews needed), this resulted in a heavier workload for reviewers and ACs than we had hoped, given that everyone is very busy with pre-existing obligations. We proactively notified reviewers and ACs in our invitation emails that the review load may be up to 6-8 papers (for reviewers) and up to 10 papers (for ACs). Unfortunately, a small, but integral, group of reviewers and ACs disengaged with the process at various stages, leaving the Program Chairs to spend time tracking them down and, when that was unsuccessful, to re-allocate their workloads to emergency reviewers or ACs. Thus, we recommend that next year’s Program Chairs begin recruiting early and often to build a larger pool of reviewers and ACs in order to cap paper review obligations to no more than four or five papers per person. And, we strongly encourage all members of the FAccT community to sign up to review or AC for next year. As a growing community (with over a 20% increase in submissions last year!) it is up to all of us to help contribute to the growth and health of this field by volunteering to review or Area Chair if you are able to.

Expectations around revise and resubmit: The revise and resubmit process was new, and largely successful. Still, some authors were disappointed and frustrated when they revised papers in line with the meta-review’s feedback but were ultimately rejected. We recommend that next year’s PCs and Area Chairs clarify that meeting the requested revisions is a necessary, but not sufficient condition for acceptance. In other words, revise and resubmit is not simply a conditional acceptance, conditioned on making the required changes. It may sometimes be the case that papers do what was asked of them, but this results in a worse paper overall (e.g., pre-existing issues are brought into starker relief by the changes made; or new additions lead to a less coherent paper overall, and so forth). Moreover, at this review stage, the reviewers and AC must make another independent judgment of the revised paper based on their expert opinions to evaluate whether the revised paper meets the bar for quality to publish at FAccT. Not all revised papers will meet FAccT standards despite authors’ best efforts.

Systematic approach to suspected Generative AI usage: During the review process, some reviewers and ACs suspected that LLMs were used in preparing the manuscript or drafting text without disclosure. Under the ACM policies at the time of the submission deadline, LLMs are not permitted at all for writing text, but they are permitted, with attribution, for tasks such as grammar and style editing, summarization, and figure or table structuring. Academia is in a period of flux with regard to LLM usage, and norms and values are shifting quickly; the bar is high for a conference and field focused on societal impacts of algorithmic systems to address this in a responsible way. This raised challenging issues in the review process, as some reviewers and ACs had to spend time enforcing the AI policy rather than engaging with the substance of the paper. This is a difficult and unpleasant role. We recommend that future Program Chairs continue to require AI usage disclosure statements, as well as developing and enforcing a systematic policy for how to handle potential cases of undisclosed or unauthorized AI usage.

Despite these challenges of a growing field, we know you will be inspired and informed by this year’s papers, keynote speakers, and plenary panels. We selected keynote speakers and panels at the cutting edge of issues important to the FAccT community. We welcome your feedback at the town hall and are already coordinating with next year’s PCs to ensure another successful FAccT conference in 2027. See you in Montréal!