12 - Why CHI rejected your paper (w/ Phoebe Toups Dugas)

This podcast episode marks the relaunch of my dormant How to Write CHI Papers podcast, now broadening into How to Write Research Papers. My guest is associate professor at Monash University and one of the CHI 2025 Papers Chairs, Phoebe Toups Dugas, a longtime friend and co-author. The conversation focuses on three threads: what removing CHI’s page limits did to paper quality, how the conference’s massive federated review system actually works, and what generative AI means for authors who sign their names to archival publications.

On the publishing side, Toups Dugas explains the death of page limits (4-page notes → 10-page papers → unlimited) as the biggest change of her career: it accommodates legitimate needs like qualitative quotes and large figures, but also lets under-polished 20-page drafts through. As Papers Chair, she explains the system behind it — Technical Program Chairs appoint Papers Chairs, who recruit 48 subcommittee chairs, who recruit Associate Chairs (the closest analogue to journal editors) — and details desk rejects: broken anonymization (grant numbers, institutions, metadata), grossly broken formatting, non-research objects, and the new-for-2025 criterion that papers must be situated in scholarly/CHI literature (prompted by submissions citing only webpages and encyclopedias). She also explains why CHI hasn’t followed CSCW and CHI Play into a PACM journal: a new journal needs ~5 years of citation data before it can be ranked (e.g., Australia’s CORE rankings), which would strand CHI’s 4,000+ annual submissions in an unrankable venue.

On craft and AI, her “secret sauce” is structure: reflexive thematic analysis and substantive positionality statements (which she’d extend to quantitative work too), research questions placed in introductions with subsections — even when co-authors object — and treating your own draft like qualitative data, splitting and merging sections as concepts emerge so readers always have guideposts. On generative AI she draws a bright line: never feed others’ unpublished work into an LLM; polishing a review is murkier, but she’d rather read point-form bullets than ChatGPT paragraphs anyway. Her deepest concern is archival responsibility — “you’re responsible for anything that has your name on it,” for readers far beyond the present toolkit, since you can’t know an LLM’s sources or biases. Her closing tip for junior writers: think carefully about structure, because headers tell readers where they are, what they should already know, and what’s coming.

Topics

* Restarting the podcast — Lennart opens by saying How to Write CHI Papers is coming back after years of silence and will become How to Write Research Papers, with a CHI special still attached, and that more general writing advice is welcome because listeners sit all along the writing spectrum.

* What changed in a decade and a half of CHI publishing — Toups Dugas names paper length as the largest shift: the old 4-page notes and 10-page papers counted references, so graduate students abbreviated names to squeeze citations; references stopped counting; page limits then disappeared.

* The polish-versus-space tradeoff — Unlimited length lets qualitative work keep giant figures and quoted material, and it also lets authors submit drafts that never got tightened; she has reviewed 20-page papers that did not need to be 20 pages, while “how long they need to be” stays an open question.

* Which sections grew — Asked whether introductions, related work, methods, or discussions took the extra pages, she suspects the data and results portions more than backgrounds; backgrounds that list papers instead of synthesizing them remain a frustration; a longer discussion is the growth she would actually want.

* IMRAD, related work, and room for creativity — Nacke cites a paper from this year’s CHI that criticizes the hundred-year normalization of the IMRAD (introduction–methods–results–discussion) scientific article, plus CHI’s extra related-work section, and asks whether writing tools plus no page limit produce more sameness or more invention.

* Experimental layouts versus guardrails — Toups Dugas has used the extra space for more creative layouts; as Papers Chair she mostly sees problematic papers, not the experimental ones; qualitative, observational, and argument papers follow IMRAD less strictly and then get hit with “where’s your methodology?” on research-synthesis work.

* Updating Contributions to CHI — Papers chairs reviewed and updated the Contributions to CHI document so it would match their sense of what CHI should look like, as the conference tries to include more types of work.

* What a CHI Papers Chair actually does — The CHI job is management, not reading papers: Technical Program Chairs appoint Papers Chairs; at least one Papers Chair rolls over (the reason she is doing 2025); they recruit 48 subcommittee chairs (at least two per subcommittee, some three or four), who recruit Associate Chairs; ACs pick reviewers and may serve on only one subcommittee.

* Subcommittees and the AC–editor analogy — Subcommittees are concept-driven: Games and Play; three Understanding People tracks (statistical, qualitative, mixed methods). The AC is the closest analogue to a journal associate editor. Nacke points listeners who want the rest of the pipeline to chicourse.com and the YouTube comments rather than replaying his course on air.

* Desk rejects after killing quick rejects — Desk rejects and quick rejects arrived a couple of years earlier; this past year they kept only desk rejects because a later “fail fast” felt like an extra hurdle. Official criteria live on the CHI website; her list is mechanical: failed anonymization (grant numbers, institution, names in metadata), grossly broken formatting (all-bold caps on three pages), and objects that are not research papers (a patent).

* The 2025 situatedness bullet — The controversial addition for 2025 is whether the work is situated in scholarly literature and/or the CHI literature; the extreme case is a paper that cites five webpages, two encyclopedias, and a book and nothing from CHI. Subcommittee chairs can judge that; until this year there was no desk-reject hook for it.

* Why CHI stays a conference while siblings became journals — Computer science used fast annual conferences as the publication venue and built heavy review around them; CSCW moved first to a journal that later invites authors to present; CHI Play and most other spin-offs followed into PACM so careers would be legible to tenure committees that read journals.

* The five-year ranking trap — CHI keeps revisiting the journal question in steering conversations and at a Future of SIGCHI summit a couple of months before the taping. A new journal needs about five years of citation data before it can be ranked; cutting off CHI for that interval would strand 4,000-plus submissions in an unrankable venue, so the community still uses acceptance rate and process descriptions instead.

* CORE in Australia — Beside SCImago journal quartiles, Australia uses CORE conference ranks. Toups Dugas has sat the CORE panel: citation data, global versus regional reach, who attends and how often, prestigious authors. Ranks run A, A+, B, C; CHI is A+ for HCI; she would put CHI Play at A, except CHI Play is now a journal and therefore no longer CORE-ranked.

* Writing craft: positionality and structure — Asked for the move that made her papers succeed, she jokes about the field’s love of a rejection badge, then talks reflexive thematic analysis and positionality statements that go past a boilerplate paragraph into how the researcher entered the work. She has been putting research questions in the introduction and adding introduction subsections, which enraged a co-author; she reads the chunking as a neurodivergent need for guideposts, and Nacke recalls work showing more subsections than five or six years ago, now that the 10-page crush is gone.

* Organic outlines and multiple backgrounds — She drops an outline into the template near the start, then splits and merges as she writes—sometimes several background sections (game design; historical ways of knowing) so a later unexpected word, including language about transgender experience, can point back. She describes that as a qualitative analysis of her own draft.

* Generative AI in reviews and manuscripts — She has not been able to prove a review was ChatGPT; a handful were flagged and could not be separated from ordinary prose, including prose that looks “AI” to neurotypical readers. Feeding unpublished submissions into a large language model is, for her as Papers Chair, ethically off-limits; running a finished review through an LLM to make paragraphs is closer to the same problem. She would rather receive a point-form list of concerns than ChatGPT paragraphs built from that list.

* The “inflowful” failure mode — The writing-aid tell she has seen is twisting a field-specific term as if it were ordinary English; the invented example is Csikszentmihalyi’s flow theory becoming something like “inflowful theory.”

* Policy, water, and archival time — Nacke teaches a how-to-use-AI-productively course, cites a CHI keynote that compared each query to pouring a bottle of water on the ground for cooling and energy, and asks whether backlash or toolkit-embrace wins. Toups Dugas treats embrace as inevitable unless something drastic happens, then shifts the burden: a name on an archival CHI paper is a claim aimed at readers far beyond the present toolkit, and extracted model text does not tell you its sources or the company’s intentions.

* Closing tip — For a junior writer she repeats structure: headers as guideposts so a reader knows what they should already know and what is still coming. Nacke thanks her and ends the taping.

Highlights

“It’s very possible to submit papers that are kind of of a lower level of polish.” — Phoebe Toups Dugas

Context. First answer of the interview, after Nacke asked what had changed across a decade and a half and why CHI dropped page limits.

What surfaced. Unlimited length is sold as accommodation for qualitative figures and quoted evidence; the same rule is also a permission structure for drafts that never got tightened. She will not quantify “how long they need to be,” and she later suspects the extra pages landed in results more than in discussion—the one section she would have wanted to grow. The same tension sits under the 2025 situatedness desk-reject: a paper can be a real paper and still be the wrong length, or the wrong literature, for the venue.

“Have I been successful in my rejected author position?” — Phoebe Toups Dugas

Context. Around 21 minutes, after the organizational tour, Nacke asked for the one structure or approach that had made her publications stand apart.

What surfaced. The invited “secret sauce” question hits a Papers Chair who still wears the rejected-author badge the field jokes about. What follows is not a template; it is positionality, introduction subsections that angered a co-author, and a refusal to treat qualitative craft as optional garnish. The joke and the later advice to junior writers (headers as guideposts) are the same person: success, in her telling, is a relationship to structure, not a win rate.

“It takes, I think, at least five years before you can get ranked as a journal, because you essentially need five years of data on citations.” — Phoebe Toups Dugas

Context. Mid-interview, after Nacke asked why CHI had not followed CSCW and CHI Play into a PACM-style journal, and after he floated the political fear that merging would cost CHI its “prime time status.”

What surfaced. The blocker she names is not pride; it is a bootstrap. CHI would have to stop being a rankable conference for the interval it takes a journal to grow a citation series, while 4,000-plus people still need a venue they can show a dean. That is why acceptance rate and process talk remain the currency, and why Australia’s CORE panel—which she has sat—can still call CHI A+ and must mark CHI Play “journal published” instead of A. The steering committee and the SIGCHI summit keep reopening a question this arithmetic has not solved.

“I’m not entirely convinced that everyone shouldn’t just be writing positionality statements for all of their work.” — Phoebe Toups Dugas

Context. After she described reflexive thematic analysis and the way the researcher enters qualitative work, and after she admitted she has been “grumping at people” who skip positionality in qualitative papers.

What surfaced. She then applies the same demand to statistical papers: the choice of which numbers to gather is already a position. Nacke tries to treat that as settled community consensus and asks again for the “Phoebe sauce”; she pivots to experimental structure instead of walking the claim back. Whether CHI will police positionality the way it now polices situatedness is left hanging.

“It’s generally not okay to put other people’s unpublished work into large language models because they shouldn’t be learning from unpublished work.” — Phoebe Toups Dugas

Context. Near 28 minutes, answering whether she has seen ChatGPT in the CHI review process; she had just said she could not prove any review was generated.

What surfaced. The bright line is training-and-prompting on a submission the reviewer has no right to donate. Polishing a review you already wrote sits beside that line—she calls it “maybe less ethically questionable” and then undercuts it by saying she would rather read a point-form list than ChatGPT paragraphs built from one. The later archival-time argument is the same ethic pointed at authors instead of reviewers.

“Humans have spent time and attention on them and have asserted that there is long term, for some definition of long term, ten to hundreds of thousands of years worth of value.” — Phoebe Toups Dugas

Context. After Nacke described the CHI keynote’s bottle-of-water image and asked whether policy backlash or toolkit-embrace would win; she had already called embrace “kind of inevitable.”

What surfaced. Inevitability is not permission. A name on a CHI paper is, in her framing, a claim about the scholarly record, not a claim about productivity. The worry she describes is not cheaper labour but not knowing the sources or the intentions inside the model.

Books

* Flow: The Psychology of Optimal Experience — Mihaly Csikszentmihalyi’s flow theory, the term Toups Dugas used to show how an LLM can mangle a field-specific phrase.

Institutions/tools

* How to Write CHI Papers — the podcast being restarted; also listed as How to Write Research Papers.

* chicourse.com — Nacke’s writing course; the full CHI review pipeline.

* CHI 2025 — flagship ACM conference; Toups Dugas is a Papers Chair for this cycle.

* SIGCHI — parent society; Nacke refers to SIGCHI conferences

* CHI Play — smaller SIGCHI conference; fewer layers of federation; now a PACM journal.

* CSCW — first of the sibling conferences she remembers moving to a journal-plus-presentation model.

* Proceedings of the ACM on Human-Computer Interaction (PACMHCI) — the “PACM” journal Nacke mentions as the home for those special issues.

* ACM — publisher of CHI proceedings and PACM.

* Contributions to CHI — the document Papers Chairs reviewed and updated.

* CHI 2025 Papers review process / desk rejects — official desk-reject list she told juniors to treat as the last word.

* CHI 2025 subcommittees

* CORE / ICORE conference rankings — Australian conference ranks; CHI is A+; CHI Play is marked journal-published.

* SCImago — journal quartile index Nacke names for career strategy.

* ChatGPT — the generated-review and generated-prose system they keep returning to.

* Claude — named with ChatGPT in Nacke’s “uprising of the robots” setup.

* Microsoft Word — Nacke’s comparison case for a writing technology the field already absorbed.

* IMRAD — the introduction–methods–results–discussion template in the CHI paper Nacke cites.

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit lennartnacke.substack.com/subscribe