CHI Writing Style

A CHI paper is read by a jury that rarely shares one methodological culture: a quantitative experimenter, a design researcher, and an ethnographer may all review the same submission. The prose must therefore do explicit work that a single-culture venue lets authors skip — naming the contribution type, justifying the method choice to outsiders, and calibrating claims to human evidence. And because CHI reviews length against contribution ("papers should be commensurate with their contribution"; 5,000–8,000 words encouraged, per the CHI 2027 Papers page, read 2026-07-08), every paragraph is auditable: does this paragraph earn its words?

The contribution statement is load-bearing

CHI convention puts an explicit contribution list at the end of the introduction. Write it first, keep it honest, and let it govern the paper:

This paper contributes:
(1) <empirical finding>: the first interview study (n=24) of <population>
    doing <activity>, showing <the finding in one clause>;
(2) <artifact>: <system>, an open-source <what>, which embodies <the design claim>;
(3) <design implications>: five implications for <who>, grounded in (1) and
    tested against (2).

Each bullet must name its type (finding, artifact, method, theory — the taxonomy in chi-topic-selection), its evidence, and its beneficiary. A bullet you cannot phrase this way is not yet a contribution; cutting it now is cheaper than a reviewer doing it in November.

Section-by-section: what each reviewer culture checks

Participant voice without ventriloquism

Quotes are evidence at CHI, and their handling is a style signature:

• Attribute with stable IDs (P7), balance across participants, and never let one articulate participant carry three themes.
• Interpret after quoting; a quote is not self-explanatory.
• Report speech faithfully but humanely — mark elisions, do not correct grammar silently, do not mock dialect through excessive [sic].
• Describe people the way current bias-free-language guidance and the communities in question prefer (identity-first vs person-first varies by community — accessibility research has explicit norms; when studying a specific community, follow its stated preference and say so).

Calibration is the house register

HCI findings are situated. The credible CHI register scopes claims in the sentence that makes them, not only in the limitations section: "participants in our two-week deployment", "for information-seeking tasks like ours", "we observed, among these 24 U.S.-based teachers". Reviewers punish two opposite failures — universal claims from situated studies, and findings hedged into vapor. The discipline: one claim sentence per finding, scoped, in active voice, with the evidence pointer attached.

Write for the ear and the screen reader

Accessibility is a review-community value and a camera-ready requirement (chi-camera-ready), and it starts in drafting: expand every acronym at first use, keep sentences parseable when heard rather than seen, never encode meaning purely in color or spatial position ("the red bars", "as seen on the left"), and write figure references that work without the figure ("error rates halved (Fig. 3)" not "see Fig. 3"). These habits also serve the many CHI reviewers reading in their second or third language.

Length discipline in a no-page-limit venue

No maximum length means the justification is on you. Practical passes:

• Run texcount -inc -sum main.tex at every full draft; drift above ~8,000 words needs a reason you could state to a reviewer, and 12,000+ risks desk rejection (chi-submission).
• The three chronic bloat sites in CHI drafts: related work written as coverage proof (halve it by making every citation do positioning work), methods detail that belongs in the supplement (move instruments out, keep decisions in), and discussion sections that re-narrate findings (each discussion move must add an implication, comparison, or consequence).
• Short papers (<5,000 words) are a legitimate, reviewed-to-scale form — cutting to short is often the strongest move for a single clean finding.

Output format

[Contribution statement] <n> bullets, each typed+evidenced: yes/no
[Audience test] method justified for non-specialist reviewer: yes/no
[Claim calibration] unscoped universals found: <list or none>
[Voice] quote balance across participants: ok / dominated by <IDs>
[Accessible prose] acronyms, color-independence, hearable sentences: pass/fail
[Word count] <n> words — verdict: short / standard / justify / cut