Student or educator? Sign up with your school email and get 6 months of Pro free.

The DOI matches its Crossref record; 75 of the 88 references it cites were found in a registry or an index, and 4 are not checked yet; 32 of the 49 sources its claims cite were checked for support.

Where Do I 'Add the Egg'?: Exploring Agency and Ownership in AI Creative Co-Writing Systems

Open access
Carrera, Dashiel, Thomas-Mitchell, Jeb, Wigdor, Daniel · 2026 · Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems

Abstract

AI co-writing systems challenge long held ideals about agency and ownership in the creative process, thereby hindering widespread adoption. To address this, we investigate conceptions of agency and ownership in AI creative co-writing. Drawing on insights from a review of commercial systems, we developed three co-writing systems with similar functionality but differing interface metaphors: agentic, tool-like, and magical. Through interviews with creative writers (n=18), we explored the role of these metaphors in participants’ sense of control and authorship. Our analysis resulted in a taxonomy of agency and ownership subtypes and underscored how metaphorical framings afforded participants different conceptions of agency and ownership. We conclude with recommendations for the design of AI co-writing systems, emphasizing the role of metaphors in participants’ creative practice.

Verified 2026-10-05 · 4 references not checked yet

Overview

CoChat Summary

TL;DR

The paper investigates how interface metaphors in AI creative co-writing systems shape writers’ conceptions of agency and ownership. Through a review of 12 commercial systems and interviews with 18 writers using three prototype metaphors, it identifies three agency subtypes and three ownership subtypes, arguing that metaphors frame expectations of control and authorship.

Key findings

  • A review of 12 commercial AI creative co-writing systems identified three recurring interface metaphor archetypes: agent-like (7 systems), tool-like (10 of 12 systems), and magic-like (9 systems); sparkles appeared in 8 of 12 systems.
  • Interviews with 18 writers (9 professional, 9 non-professional) produced a taxonomy of agency—explicit control, implicit control, and writing process control—and ownership—stylistic, conceptual, and effort-based.
  • Tool-like metaphors shifted writers’ expected points of control toward explicit interface elements such as dropdowns, sliders, and buttons, even though those parameters did not actually modify the AI output; some participants noticed this and felt less control (e.g., P17).
  • Agent-like metaphors foregrounded conceptual contributions: participants attributed ownership to ideas, plot points, character sketches, and setting, even though the system was only capable of generating more text.

Methods

  • In May 2024, the authors reviewed commercial AI co-writing systems: a preliminary search found 45 systems, non-creative-writing systems were excluded, and 12 were reviewed in depth (ChatGPT, DreamGen, Fictionary, Google Gemini in Google Docs, Grammarly, NovelAI, NovelCraft, Novelcrafter, Quillbot, Squibler, Sudowrite, and Writesonic), with hour-long explorations of each.
  • Three prototype systems with near-identical functionality were built: Wordsworth (agent-like), Autowrite (tool-like), and Magic Quill (magic-like). Completions used GPT-3.5-turbo, limited to 250 tokens (approximately 100 words).
  • The exploratory study used a between-subjects design: 18 participants (9 professional creative writers, 9 non-professional writers) completed a 15-minute writing task with one randomly assigned prototype, followed by semi-structured interviews; sessions lasted 60–90 minutes. Analysis used open coding followed by reflexive thematic analysis by two coders, with professional and non-professional themes initially analyzed separately and then merged.

Limitations

  • Stated: the review focused on commercial systems available at the time, and new systems may introduce new metaphors; longitudinal studies of a week or more may provide more reliable results; findings may be culturally specific to North American, English-language creative writers, so future work is needed in other cultures, languages, and genres.
  • Evident from the text: the prototypes used deception to maintain identical functionality—Autowrite’s sliders and dropdowns did not modify AI output, and Wordsworth’s chat responses were scripted—so participants’ expectations of control were tested against non-functional controls.
  • Evident from the text: the study used a short 15-minute writing task, a small sample of 18 participants, and a between-subjects design, which limits generalizability and prevents within-participant comparison of metaphors.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1145/3772318.3791952

Cites

75 of 88 references verified0 citation errors · 9 not checkable

Statistics

Nothing to recompute.

Figures

(a) Agent-like Commercial Reference Example: ChatGPT
(a) Agent-like Commercial Reference Example: ChatGPTFrom the PDF
(a) Tool-like Commercial Reference Example: Squibler
(a) Tool-like Commercial Reference Example: SquiblerFrom the PDF
Fig. 3 (a) A reference example of a magic-like interface: Sudowrite's Write button menu used a sparkle icon, the text "Autocomplete plus magic", and withheld other interface controls to invoke magic as a metaphor for creative generation. (b) The Magic Quill interface featured a single point of control, animated sparkles around the quill button, and the text "magic quill."
Fig. 3 (a) A reference example of a magic-like interface: Sudowrite's Write button menu used a sparkle icon, the text "Autocomplete plus magic", and withheld other interface controls to invoke magic as a metaphor for creative generation. (b) The Magic Quill interface featured a single point of control, animated sparkles around the quill button, and the text "magic quill."From the PDF

Verification

Claims the paper makes

25 Valid · 2 Not valid · 10 Not checked · 59 Own results and interpretations

39 more claims · 2 with their cited sources checked

Create a free account to see all 96 claims and every source verdict.

See all claims — free account
  • It may state two separate claims as one. 22
  • Probably not a substantial claim. 40

Create a free account to see these claims.

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

2 of 2

Citations

Cites

Show all 81 references

Cited by

See all 4

Metadata

Authors
Carrera, D., Thomas-Mitchell, J., & Wigdor, D.
Affiliations
University of Toronto (CA)
Venue
Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems
Publisher
ACM
Published
2026-04-13
Pages
1-16
OpenAlex
W4414695472
Cited by
3
References
81
Citations by year
2026: 3
FWCI
27.60
Type
conference paper
Language
en
Open access
Locations
  • CHI Conference on Human Factors in Computing Systems (publishedVersion, cc-by)
  • arXiv (Cornell University) (submittedVersion)
  • arXiv (Cornell University) (cc-by)
Topics
  • Physical Sciences › Computer Science › Human-Computer Interaction › Innovative Human-Technology Interaction
  • Physical Sciences › Computer Science › Human-Computer Interaction › Usability and User Interface Design
  • Physical Sciences › Computer Science › Computer Graphics and Computer-Aided Design › Digital Media and Visual Art
Keywords
creative agency, authorship, ownership, interface metaphors, perceived control, AI-assisted writing, AI system design

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-09-30, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-09-30 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Where Do I 'Add the Egg'?: Exploring Agency and Ownership in AI Creative Co-Writing Systems. https://app.cochat.ai/papers/10.1145/3772318.3791952
The paper
Carrera, D., Thomas-Mitchell, J., & Wigdor, D. (2026). Where Do I 'Add the Egg'?: Exploring Agency and Ownership in AI Creative Co-Writing Systems. Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems. https://doi.org/10.1145/3772318.3791952

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.