--- name: brycewang-stanford/chi-submission source: https://app.decimal.ai/s/brycewang-stanford-chi-submission@1/SKILL.md source_sha256: 6ee64b2aedc3 --- # CHI Submission Run this audit before uploading to PCS (Precision Conference Solutions) — CHI does not use OpenReview or CMT. Every date and cap below was read from the CHI 2027 Papers pages on 2026-07-08 through search-engine renderings (direct fetches of `chi2027.acm.org` were blocked in the verification environment); reopen the live pages before acting on any number. ## What the CHI 2027 cycle posted - Full Papers deadline: **Thursday, September 10, 2026**, on PCS, *including videos and supplementary materials* — CHI collects everything on one date, unlike venues with a later supplement window. No separate abstract-registration deadline was found for 2027; confirm on the live page rather than assuming (待核实). - Review format: the **single-column** ACM submission template (CHI has used single-column reviewing since 2021). The two-column look happens later, in TAPS. - Length is governed by words, not pages. The posted norm: roughly **5,000–8,000 words** for a typical paper (excluding references), submissions at or above **12,000 words** face extra scrutiny and desk rejection when the length is not justified, and papers of about 5,000 words or fewer are welcomed as short papers reviewed with expectations commensurate with their size. CHI's standing phrase is that a paper should be the length appropriate to its contribution. - The submission form requires you to designate a reviewing **subcommittee** (up to two choices) — a strategic decision, not paperwork; see `chi-topic-selection`. - Anonymization is mandatory and extends beyond the PDF: supplemental materials and **external links to datasets or code repositories** that reveal identity are posted desk-reject grounds. ## The anonymization trap specific to CHI CHI's policy is anonymized review of an HCI paper, which creates leak channels that ML venues rarely have: consent-form headers naming the lab, screenshots of a deployed system showing a university logo, video figures with a narrator saying "at our institution," OSF or GitHub links registered to a named account, and app-store pages for the studied prototype. Cite your own prior studies in the third person and keep the citation; scrub everything else. ```bash # Mechanical pre-upload sweep on the PCS package pdfinfo paper.pdf | grep -Ei 'author|creator|producer' # PDF metadata identity pdftotext paper.pdf - | grep -nEi 'github|osf\.io|our (lab|university|institution)' | head texcount -inc -sum main.tex | tail -3 # word count vs the norms ffprobe -v error -show_entries format=duration,size video-figure.mp4 # video sanity unzip -l supplement.zip | grep -Ei 'consent|irb|ethics|\.git/' | head # named forms ``` Treat consent forms and ethics-approval letters as the highest-risk files: they carry institution names by design, so upload templated, de-identified versions. ## Desk-reject triage, CHI edition CHI 2026 formalized an *assisted desk reject* step run jointly by subcommittee chairs and ACs, with expanded criteria — desk screening is now a real stage, not a formality. | Finding at audit time | CHI severity | Realistic fix | |---|---|---| | Wrong template or altered format | Desk reject | Recompile on the current ACM single-column class | | Identity leak in PDF, supplement, or link | Desk reject | De-identify; replace links with anonymized archives | | ~12,000+ words without justification | Desk reject risk | Cut or explicitly justify length in the submission | | No plausible HCI contribution for any subcommittee | Desk reject at SC/PC screening | Re-frame or re-route before the deadline | | Video figure missing captions | Review-stage damage | Add a .srt/.sbv file (see `chi-supplementary`) | | Weak subcommittee match | Quiet mis-review | Change the designation, not the paper, if the work is sound | ## Deadline-week order of operations 1. Lock the subcommittee pair first; it determines who reads everything else. 2. Freeze the manuscript at a defensible word count and record the `texcount` output. 3. Render and caption the video figure early — encoding and captioning always take longer than the writing team budgets. 4. Assemble supplements from a clean export, then run the mechanical sweep above on the exact files going into PCS. 5. Fill every PCS field (title, abstract, authors, subcommittees, contribution statement fields if asked) a day before; PCS has no post-deadline mercy. 6. Re-download the submitted package and open it cold, the way a 1AC will. ## Length is an argument, not a quota Because CHI reviews words against contribution, padding is a visible defect: a 9,500 word paper whose study supports 7,000 words of claims reads as inflated, and reviewers say so. Cut the third restatement of the findings, move instrument details to the supplement, and keep the method section exactly as long as replication requires — no longer. ## Cycle-volatile items — reverify every year - The September date, the no-abstract-deadline reading, and all word-count wording. - Subcommittee list and count for the live cycle (the 2026 list is posted on chi2026.acm.org; the 2027 list was not yet published at check time — 待核实). - Desk-reject criteria wording, which changed as recently as August 2025. - Whether PCS asks new form questions (AI-use disclosure, data-availability fields). ## Output format ```text [CHI submission status] ready / blocked / needs work [Word count] words vs posted norms; justification needed? yes/no [Subcommittees] first: / second: [Anonymity sweep] clean / leaks: [Desk-reject exposure] [Fix queue] ```