Chi Topic Selection

CHI Topic Selection

CHI accepts almost every topic that touches people and computing, so "is this in scope?" is rarely the real question. The decisions that actually shape the outcome are: which contribution type is this paper, and which subcommittee should read it. Both are declared at submission time in PCS, and both are strategy, not paperwork — CHI 2026 rejected roughly three quarters of 6,730 completed submissions, and a large share of early rejections trace to papers pitched at the wrong reviewer community inside CHI itself.

Start with the human, not the system

Ask: who changes what they do because this paper exists? A designer picks a different pattern, a researcher runs a study differently, a developer builds an interface differently, a policy team understands a user population differently. If no answer survives scrutiny, the project is an engineering result wearing an HCI costume, and reviewers will say exactly that. A system with no study can still be CHI-worthy — but only when the system itself embodies a claim about interaction that others can build on.

Name the contribution type before the venue

CHI's own guidance enumerates the contribution types the conference publishes. Force the project into exactly one primary type; hybrids are fine but one must lead:

Source: the "Contributions to CHI" pages carried across recent cycles (chi2026.acm.org/contributions-to-chi/, checked 2026-07-08).

The subcommittee decision

PCS asks you to designate up to two subcommittees, and the official guidance recommends naming two. In the vast majority of cases one of your two choices reviews the paper; chairs may transfer it only when another subcommittee would clearly review it more thoroughly. Practical consequences:

• The subcommittee determines your 1AC, 2AC, and reviewer pool — effectively your jury. A quantitative-methods paper sent to a design-led subcommittee gets design criticism, and vice versa.
• Read the current cycle's "Selecting a Subcommittee" page and match against the posted scope sentences, not the subcommittee titles. The CHI 2027 list was not yet posted at check time (待核实); the CHI 2026 list is the working draft of what to expect.
• Pick the second subcommittee as a genuine alternative reading of the paper, not a duplicate of the first.

Subcommittee fit test (run per candidate subcommittee):
1. Do ≥3 papers you cite belong to this community?        yes / no
2. Would its typical reviewer accept your method as rigorous
   *by that community's standards*?                        yes / no
3. Does your contribution statement use that community's
   vocabulary without translation?                         yes / no
3× yes = designate. 2 = possible second choice. ≤1 = do not list it.

When CHI is the wrong room

CHI remains right when the audience is the broad HCI community: the paper's lesson should matter to people who will never build your exact system. Note the calendar asymmetry: CHI runs one cycle per year (papers due early September), so a missed fit costs twelve months — the sibling venues above fill the gap between cycles.

Anti-signals before you commit a year

• Your related-work section cites no CHI, CSCW, UIST, or TOCHI work — you are writing for a community you have not read. (Insufficient contextualization, "ADR-Context," was the most-cited assisted desk-reject reason at CHI 2026.)
• The user study exists to decorate a systems result you would publish anyway.
• The contribution is real but 5,000 words would carry it — that is not a defect (short papers are welcome), but plan the paper as short instead of inflating it.
• Every plausible reviewer you can imagine works in one narrow specialty — a specialist venue will review it better and cite it more.

Output format

[Human stake] who changes what: <one sentence>
[Contribution type] primary: <type> / secondary: <type or none>
[Subcommittee designations] 1st: <name + fit-test score> / 2nd: <name + score>
[Venue verdict] CHI-first / CHI-viable / route to <UIST|CSCW|DIS|IUI|ASSETS|IMWUT|TOCHI>
[Length plan] short (<5k words) / standard (5-8k) / justify-longer
[Blocking gap] <the one thing to fix before drafting>