FAccT Author Response
Use this after FAccT reviews are released. FAccT has two distinct speaking turns, and conflating
them is a common mistake: a short rebuttal for factual corrections, and — if you receive a
Revise decision (new for the 2026 edition) — a full revision plus response that addresses
Area-Chair-prioritized concerns and is re-reviewed. Both must respect mutual anonymity: nothing
you write may reveal authors, institutions, field sites, or a repository owner.
Triage (both turns)
- Answer what affects the decision: relevance and framing, correctness, depth of exposition, honest
treatment of strengths and limitations, and — often decisive at FAccT — the reality of the
harm and the care taken with affected people.
- Answer each reviewer on the axis they raised, including the out-of-field reviewer. A legal or
qualitative objection is not noise to be waved off with a metric.
- Correct factual misreadings first; a reviewer who misread a subgroup table or a consent procedure
is often persuadable.
- Keep every word anonymous — do not name your system's real name, your institution, your field
site, your IRB, or a private repository, even to strengthen a point.
Turn 1 — the short rebuttal
Short and strictly for factual corrections and misunderstandings, plus signposting minor edits.
It is not a venue to re-argue substantive disagreement or to promise a new study — that is the
Revise round's job. Concede what is true, correct what is misread, and point to exactly where the
answer already lives in the submitted paper or supplementary material. Do not paste new results the
reviewers cannot verify in the current PDF.
Turn 2 — the Revise-and-resubmit response (the distinctive FAccT-2026 move)
A Revise is a genuine revise-and-resubmit. The Area Chair summarizes and prioritizes the
concerns; your job is to address that prioritized list. The response is a change ledger: for
every prioritized concern, either make the change and show it, or decline it with a reason.
[AC.1] Prioritized concern (quoted briefly)
-> Action: DONE | added disaggregation by <group> (Table 3); reported the utility cost
-> Where: §5.2, Table 3; analysis/rq2.ipynb (anonymized)
[AC.2] Prioritized concern (from the legal reviewer)
-> Action: DONE | corrected the doctrinal framing per §2; the "right" is now cited to <origin>
[AC.3] Prioritized concern
-> Action: DECLINED (with reason) | the requested field study cannot be run in the window
without re-consenting participants; added a bounded limitation instead (§6) and an
Adverse Impacts note
The rule that turns a Revise into an acceptance: no unaddressed prioritized concern. A concern
that is neither done nor explicitly and reasonably declined is what the re-review punishes.
Reviewer pushback patterns
| Pushback |
What it signals |
FAccT-ready response |
| "Results are aggregate; where is the subgroup harm?" |
Disaggregation gap |
Add per-group metrics with uncertainty; report the utility trade-off |
| "The proxy for the protected attribute is not valid" |
Construct-validity doubt |
Validate the proxy, report its error, bound the claim |
| "You speak for a community you did not study" |
Standpoint/rigor objection |
Add the community's data or scope the claim; note it in Positionality (at acceptance) |
| "The legal claim misreads the doctrine" |
Cross-lane correctness |
Fix the framing, cite the doctrine correctly, or narrow the legal claim |
| "This could be misused to justify deployment" |
Adverse-impact concern |
Strengthen the Adverse Impacts statement and the paper's caveats |
Anonymity in the response (easy to slip)
- Refer to your own prior work in the third person, as in the paper.
- Describe changes to released code/data without linking to an identity-revealing repository; use
the anonymized location.
- Do not thank a named collaborator, partner organization, or funder inside the response, and do not
reveal the field site.
Calibration
- Respond to the criterion the reviewer actually raised, not the one you would rather defend.
- The rebuttal is short and factual; the substantive work is the Revise round — do not burn the
rebuttal trying to do the revision's job.
- Length and format norms for both turns, and the very existence of the Revise round, are
2026-new — confirm the current instructions before sending.
Output format
[Turn] short rebuttal / revise-and-resubmit response
[Priority issue] <reviewer or AC concern>
[Decision dimension] relevance / correctness / depth / strengths-and-limits / harm-and-ethics / discipline
[Change ledger] <prioritized concern -> DONE/PARTIAL/DECLINED + where in paper/artifact>
[Anonymity check] <no identity leak, no field-site/positionality reveal: passed/issues>
1---2name: facct-author-response3description: Use when drafting ACM FAccT author responses — the short factual-correction rebuttal after preliminary reviews, and, distinctively for the 2026-new process, the Revise-and-resubmit round where you address Area-Chair-prioritized concerns from a mixed CS+law+social-science panel, map each to a concrete change, and survive a re-review while keeping mutual anonymity.4---56# FAccT Author Response78Use this after FAccT reviews are released. FAccT has **two distinct speaking turns**, and conflating9them is a common mistake: a **short rebuttal** for factual corrections, and — if you receive a10**Revise** decision (new for the 2026 edition) — a full **revision plus response** that addresses11Area-Chair-prioritized concerns and is **re-reviewed**. Both must respect mutual anonymity: nothing12you write may reveal authors, institutions, field sites, or a repository owner.1314## Triage (both turns)1516- Answer what affects the decision: relevance and framing, correctness, depth of exposition, honest17 treatment of strengths **and** limitations, and — often decisive at FAccT — the reality of the18 harm and the care taken with affected people.19- Answer **each reviewer on the axis they raised**, including the out-of-field reviewer. A legal or20 qualitative objection is not noise to be waved off with a metric.21- Correct factual misreadings first; a reviewer who misread a subgroup table or a consent procedure22 is often persuadable.23- Keep every word anonymous — do not name your system's real name, your institution, your field24 site, your IRB, or a private repository, even to strengthen a point.2526## Turn 1 — the short rebuttal2728Short and strictly for **factual corrections and misunderstandings**, plus signposting minor edits.29It is **not** a venue to re-argue substantive disagreement or to promise a new study — that is the30Revise round's job. Concede what is true, correct what is misread, and point to exactly where the31answer already lives in the submitted paper or supplementary material. Do not paste new results the32reviewers cannot verify in the current PDF.3334## Turn 2 — the Revise-and-resubmit response (the distinctive FAccT-2026 move)3536A Revise is a genuine revise-and-resubmit. The Area Chair **summarizes and prioritizes** the37concerns; your job is to address that prioritized list. The response is a **change ledger**: for38every prioritized concern, either make the change and show it, or decline it with a reason.3940```text41[AC.1] Prioritized concern (quoted briefly)42 -> Action: DONE | added disaggregation by <group> (Table 3); reported the utility cost43 -> Where: §5.2, Table 3; analysis/rq2.ipynb (anonymized)44[AC.2] Prioritized concern (from the legal reviewer)45 -> Action: DONE | corrected the doctrinal framing per §2; the "right" is now cited to <origin>46[AC.3] Prioritized concern47 -> Action: DECLINED (with reason) | the requested field study cannot be run in the window48 without re-consenting participants; added a bounded limitation instead (§6) and an49 Adverse Impacts note50```5152The rule that turns a Revise into an acceptance: **no unaddressed prioritized concern.** A concern53that is neither done nor explicitly and reasonably declined is what the re-review punishes.5455## Reviewer pushback patterns5657| Pushback | What it signals | FAccT-ready response |58|---|---|---|59| "Results are aggregate; where is the subgroup harm?" | Disaggregation gap | Add per-group metrics with uncertainty; report the utility trade-off |60| "The proxy for the protected attribute is not valid" | Construct-validity doubt | Validate the proxy, report its error, bound the claim |61| "You speak for a community you did not study" | Standpoint/rigor objection | Add the community's data or scope the claim; note it in Positionality (at acceptance) |62| "The legal claim misreads the doctrine" | Cross-lane correctness | Fix the framing, cite the doctrine correctly, or narrow the legal claim |63| "This could be misused to justify deployment" | Adverse-impact concern | Strengthen the Adverse Impacts statement and the paper's caveats |6465## Anonymity in the response (easy to slip)6667- Refer to your own prior work in the third person, as in the paper.68- Describe changes to released code/data without linking to an identity-revealing repository; use69 the anonymized location.70- Do not thank a named collaborator, partner organization, or funder inside the response, and do not71 reveal the field site.7273## Calibration7475- Respond to the criterion the reviewer actually raised, not the one you would rather defend.76- The rebuttal is short and factual; the substantive work is the Revise round — do not burn the77 rebuttal trying to do the revision's job.78- Length and format norms for both turns, and the very existence of the Revise round, are79 2026-new — confirm the current instructions before sending.8081## Output format8283```text84[Turn] short rebuttal / revise-and-resubmit response85[Priority issue] <reviewer or AC concern>86[Decision dimension] relevance / correctness / depth / strengths-and-limits / harm-and-ethics / discipline87[Change ledger] <prioritized concern -> DONE/PARTIAL/DECLINED + where in paper/artifact>88[Anonymity check] <no identity leak, no field-site/positionality reveal: passed/issues>89```
brycewang-stanford