arXiv / cs.HC / 2511.04081
EN ZH
Research Paper / Accepted at ASIS&T 2026 / Annual Meeting of the Association for Information Science and Technology

Everyone Else Does It

The Rise of Preprinting Culture in Computing Disciplines

How AI & HCI researchers stopped waiting for peer review — based on 15 in-depth interviews with academics from master's student to full professor.

1University of Texas at San Antonio
2Illinois Urbana-Champaign
3University of Wisconsin–Madison
4Northeastern University
5Emory University
Field Human–Computer Interaction
Method Semi-structured interviews
Posted November 4, 2025
The take

Researchers in AI and HCI stopped waiting for peer review.

We sat down with fifteen of them to ask why — and what's being lost in the deal.

15
academics interviewed — master's student to full professor, 2 to 30+ years.
77%
of CS preprints on arXiv eventually appear in peer-reviewed venues.
+
of CS reviewers admit to searching online for the papers they've been assigned.
Abstract

Preprinting has become a norm in fast-paced computing fields such as artificial intelligence (AI) and human–computer interaction (HCI). In this paper, we conducted semi-structured interviews with 15 academics in these fields to reveal their motivations and perceptions of preprinting. The results found a close relationship between preprinting and characteristics of the fields, including the huge number of papers, competitiveness in career advancement, prevalence of scooping, and an imperfect peer review system — preprinting comes to the rescue in one way or another for the participants. Based on the results, we reflect on the role of preprinting in subverting the traditional publication mode and outline possibilities of a better publication ecosystem.

Our study contributes by inspecting the community aspects of preprinting practices through talking to academics.

Index Terms Preprint Culture Computer Science AI HCI
§ 1 · Q01
Q.

What did this paper actually find?

A.

Three things. The rest of the page elaborates each — but if you read nothing else, take these.

01

Preprinting is now the default.

Posting moved from after acceptance to before submission. ACL dropped its anonymity window — the behavior preceded the policy.

02

The motivations are structural, not ideological.

Not open-science enthusiasm. Slow review, ruthless competition, scooping risk, career pressure — preprinting is a workaround for a system that stopped working.

03

Everyone sees the cost. No one is steering it.

“There's a lot of garbage on arXiv,” a senior professor told us. She keeps posting anyway. The field is improvising a replacement system in plain sight.

Clip · P12 P12master's student · HCI
Heard from a master's student
Famous papers did not get accepted to conferences — but preprints became the actual papers.
P12 On what gets remembered
§ 2 · Q02
Q.

So why did the field stop waiting?

A.

Five themes recurred across the fifteen interviews — none mutually exclusive, most participants raised three or more. We'll start with one of them, then list the rest.

Take P10 — about to start as an assistant professor in HCI and AI. He spoke for an hour about machine learning, tenure track, and the politics of his first faculty year. Halfway through, he said this:

Tape 01 P10incoming assistant professor · HCI & AI 58:14
The moment
You start stacking up the citation.
@ 54:02 On the politics of tenure-track posting
How we got there
Q

When did the pace of the field start to feel — different?

P10

I don't know how to recalibrate after everything that has been happening with machine learning. And ChatGPT, and the volume at which they are putting papers on arXiv… there are lots of factors at play.

Q

And that's changed how you think about posting your own work?

P10

Especially now that I will be starting as an assistant professor — I know that there are a lot of politics involved as well. Annual reviews, tenure track. You do want to get your paper out somewhere, whether it's arXiv or not. You start stacking up the citation.

His isn't an outlier story. Five themes emerged across all fifteen interviews — none mutually exclusive; most participants raised three or more.

01 · Norm
9/15

The default.

Absorbed from advisors, collaborators, the rhythm of the field. “Machine learning culture,” one PhD student called it.

It seems much more common and much more like the norm to just post on arXiv. — associate professor of HCI
02 · Visibility
13/15

A way to be seen.

Modern conferences are floods. Work that isn't actively promoted disappears.

Every year that you get delayed, you lose a lot of citations. arXiv is kind of like insurance. — PhD student in AI & HCI
03 · Career
7/15

Career capital.

Job applications, grants, tenure files all count visible work. For early-career researchers, even one preprint can matter.

Even if some work is still in submission, employers are still interested. Even one paper really matters. — PhD student in NLP
04 · Scooping
13/15

Defense against scooping.

Parallel discovery is the rule. One PhD student saw two arXiv papers, time-stamped two days apart, arrive at the same conclusion.

If you don't arXiv on time, other researchers may scoop your idea. Your research has to be on time. — PhD student in NLP
05 · Workaround
11/15

Workaround for a broken system.

Too slow. Too inconsistent. Too punitive. An associate professor waited two years for minor revisions. Another waited two more between acceptance and publication.

The peer review system suppresses novelty. Reviewers tend to reject papers that are too novel. — PhD student in AI & HCI
Tape 02 P1PhD student · NLP & embodied AI 64:47
The moment
But now we sort of have to look for some low-hanging fruit.
@ 31:14 On the cost of the new pace
How we got there
Q

Does the pace ever feel like too much?

P1

It's fast. Sometimes you feel tired, and sometimes you have to work seriously to fight against FOMO — fear of missing out. People are aware that there's a lot of hype on social media. It really takes some effort and mental health to get used to this kind of pace.

Q

What's the cost of that pace, in your own work?

P1

In the past, we could take our time, spend one year on a good problem, and really define it, and work it, and make it solid, and submit. But now we sort of have to look for some low-hanging fruit. You either give them up — or do it fast.

Clip · P3 P3PhD student · AI & HCI
Heard from a doctoral student
In some fast-moving fields, every year that you get delayed, you will lose a lot of citations. arXiv, basically, is kind of like an insurance.
P3 On preprinting as insurance against time
Tape 03 P8PhD student · NLP / AI 22:31
The moment
Either paper would have been scooped if they were not preprinted.
@ 12:08 Two papers, time-stamped two days apart
How we got there
Q

Have you seen scooping happen?

P8

There were two arXiv papers I remember — time-stamped two days apart — that were highly similar. Either paper would have been scooped if they were not preprinted. Since everything comes down to deep learning, people are thinking about similar things all the time.

Q

And what changed in your own posting habits after that?

P8

A year or two ago, we would archive once the paper is accepted. But now — my field is NLP, our conference is ACL — they got rid of the anonymity requirement. It used to be, before you submitted, you needed to be anonymous for two months. No preprinting, no tweeting. They got rid of this.

The paradox

Most of them post to arXiv every week.
Two-thirds also flag the platform's quality.

Both positions are held by many of the same researchers. 10 of 15 raised quality control — one senior professor called it “a lot of garbage.” She keeps posting anyway. The workaround has a cost; they see it.

§ 3 · Q03
Q.

But what's lost in the trade-off?

A.

Three concrete costs. Each surfaced repeatedly in the interviews — named, in some cases, by the same researchers who keep posting anyway.

01 · Recognition
5/15

Preprints don't count.

Tenure committees don't count them. Grant reviewers don't either. Google Scholar quietly favors the arXiv version, making the formal record harder to find.

Why am I obligated to cite your preprint that would have been rejected nine times out of ten? — assistant professor of HCI
02 · Quality
10/15

The gatekeeper has left.

Anyone uploads anything. The ease compounds with volume. Generative AI now writes some of it. The field's collective taste filter is weakening.

A lot of overclaiming, a lot of incomplete work in progress. We're just there. — PhD student in NLP
03 · Anonymity
3/15

The blindfold is off.

A preprint indexed by Google undoes double-blind review on day one. Less-known institutions suffer the most.

Some relatively less privileged research groups might suffer from fairness issues. — PhD student in NLP
Tape 04 P2full professor · AI · 30+ years in the field 51:09
The moment
I think there's a lot of garbage on arXiv.
@ 18:45 On the absent gatekeeper
How we got there
Q

What worries you about the trend?

P2

I think there's a lot of garbage on arXiv. Sometimes people use it for, like, here's a paper, I need to put it somewhere. It's not getting accepted, or, for whatever reason, they're not submitting it somewhere.

Q

And the indexing situation?

P2

For one thing — once something's on arXiv, the actual published versions become harder to find. Google Scholar favors arXiv. I would prefer that when someone finds my papers that have been peer-reviewed, they find them together with that information about peer review.

§ 4 · Q04
Q.

So where does this leave the field?

A.

Three proposals surfaced — none of them radical. Reasonable adjustments to a publication mode that is straining at its seams.

Tape 05 composite — P5 (asst. professor) · P8 (PhD student) · P2 (senior professor) edit
The moment
Work on problems that no one else is in a hurry to scoop you on.
P2 Thirty-plus years' counsel
Three answers to one question
Q

If you could redesign the system, what would you change?

P5

I'd like research magazines. Venues with shorter formats, faster turnarounds, and the capacity to circulate research with policy urgency. The current system is set up for archival prestige — not for the pace that AI governance actually needs.

P8

A decentralized scientific communication network. Built on small social ties rather than venue prestige. So under-promoted work can still find an audience.

P2

(after a long pause) Or — the most sustainable strategy of all. Work on problems that no one else is in a hurry to scoop you on.

Three proposals surfaced. None radical — reasonable adjustments to a publication mode that is straining at its seams.

  1. 01 Research magazines — shorter formats, faster turnarounds. Suggested by an assistant professor working on AI governance.
  2. 02 Decentralized networks — built on social ties, not venue prestige. Suggested by an NLP PhD student.
  3. 03 Work on unscoopable problems — the most sustainable strategy of all, offered by a full professor with thirty years in the field.
Work on problems that no one else is in a hurry to scoop you on. — full professor of AI · 30+ years

Peer review and paper quality have long traveled together. Our participants, with care and ambivalence, have begun to unbundle them.

§ 5 · Methodology

How this paper was reported.

For the curious or the skeptical: how fifteen interviews became five themes.

This study is built from 15 semi-structured interviews conducted under approved IRB protocol with academics in Computer Science and Information Sciences whose research is in AI, HCI, or the intersection of the two. Recruitment was via direct email and a posting on X. Each conversation lasted approximately one hour, was audio-recorded with consent, transcribed with Otter.ai, and manually corrected.

Two of the authors performed thematic analysis on the transcripts — open coding first, then hierarchical clustering of emerging themes in XMind, then consensus. Four major themes emerged: practices, motivations, drawbacks, and preprints vs. publications. The full interview script appears as Appendix A in the underlying paper.

Table 1 · Participants (N = 15) 4 HCI · 8 AI · 3 HCI∩AI · 8M · 7F
IDRoleYrsField
P1PhD Student5AI
P2Full Professor30+AI
P3PhD Student6AI · HCI
P4Postdoc10AI
P5Assistant Professor19AI
P6Assistant Professor10HCI
P7Associate Professor11HCI
P8PhD Student7AI
P9PhD Student3HCI · AI
P10Asst. Professor7HCI · AI
P11PhD Student5HCI
P12Master's Student3HCI
P13Software Engineer2AI
P14PhD Student2AI
P15PhD Student5AI
§ 6 · Cite this work

If this study is useful to you.

A BibTeX entry below — or a plain-text citation, for prose. Both note that the paper is forthcoming at ASIS&T 2026.

BibTeX
@misc{zhou2025everyone,
  title         = {“Everyone Else Does It”: The Rise of Preprinting Culture
                  in Computing Disciplines},
  author        = {Zhou, Kyrie Zhixuan and Chen, Justin Eric and Zheng, Xiang and
                  Qian, Yaoyao and Xiao, Yunpeng and Shu, Kai},
  year          = {2025},
  eprint        = {2511.04081},
  archivePrefix = {arXiv},
  primaryClass  = {cs.HC},
  note          = {To appear in Proceedings of ASIS&T 2026}
}
Plain text
Zhou, K. Z., Chen, J. E., Zheng, X., Qian, Y., Xiao, Y., & Shu, K. (2026).
"Everyone Else Does It": The Rise of Preprinting Culture in Computing
Disciplines. In Proceedings of ASIS&T 2026.
arXiv:2511.04081 [cs.HC].
Copied to clipboard