Pith. sign in

REVIEW 4 major objections 5 minor 89 references

Who Decides How Knowing Becomes Doing? Redistributing Authority in Human-AI Music Co-Creation

T0 review · 4 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Making AI reasoning editable, not just visible, lets music creators contest, redirect, and safeguard their aesthetic choices.

desk verdict A well-framed and genuinely ambitious study of contestable reasoning surfaces for music co-creation, but missing statistics, an underspecified system, and partially circular survey items leave the central authority-redistribution claim under-supported. read the letter →

arxiv 2509.10331 v1 pith:TZXXGVHF submitted 2025-09-12 cs.HC cs.CY

Zhejing Hu , Yan Liu , Zhi Zhang , Gong Chen , Bruce X.B. Yu , Jiannong Cao This is my paper · ORCID
classification cs.HCcs.CY
keywords human-AIco-creationinterpretiveauthoritycontestabilityagencypluralitymusicgenerationgenerativeAIcriticalcomputing
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that the real bottleneck in human–AI music co-creation is not the technical difficulty of generating music but who holds interpretive authority—the power to decide how a user's intention becomes a concrete output. To address the hidden interpretations and homogenizing defaults of prompt-and-output systems, the authors propose three principles—contestability, agency, and plurality—and implement them in a prototype called Mutelier that shows the model's reasoning as an editable list of symbolic steps. In a six-week crossover study with 180 music practitioners at three experience levels, they find that beginners learn to dispute the system's definition of their intentions, intermediates can redirect outputs that previously drifted toward mainstream conventions, and advanced users can defend minority aesthetics such as open fifths and sliding ornaments. The study concludes that authority over the knowing-doing passage is not fixed but redistributable through design, turning creative collaboration from a model-controlled pipeline into a negotiated medium.

What carries the argument

The central mechanism is the editable reasoning surface: an externalized, symbolic trace of the model's translation of a user's intent into musical choices (e.g., 'unstable = add suspension (resolved)'). The paper's intervention, Mutelier, renders this trace as a list of steps that users can inspect, delete, or annotate, and pairs it with drift notices and branching options that preserve non-mainstream directions. The comparison of visible-but-uneditable versus editable reasoning isolates the work this surface does: visibility alone raised awareness but did not change who could act; editability converted the knowing-doing passage into a site of negotiation.

What would settle it

A controlled test where users delete one randomly selected reasoning step and the musical output's relevant feature is measured across many trials: if deletion produces no systematic, directionally correct change in the output (as the paper's own example hints), the central claim that editing reasoning redistributes authority would fail. Alternatively, instrument the prototype to log the true internal gradient or token attribution for each displayed step and check whether edited steps correspond to the model's actual decision points; a mismatch would indicate the surface is a fabrication.

Watch

Extended reading notes

Core claim

The paper's central discovery is that interpretive authority in the knowing-doing cycle can be redistributed from the model to the human by externalizing the model's reasoning as an inspectable and editable surface. The authors identify three structural failures of the dominant prompt-output paradigm: an interpretive monopoly (hidden reasoning cannot be contested), a loss of agency (users are limited to rephrasing prompts), and homogenization (defaults regress to majority conventions). Their prototype, Mutelier, surfaces reasoning as editable steps—for example, 'gentle = slow tempo; unstable = add suspension'—lets users delete or annotate those steps to redirect generation, and issues drift

Load-bearing premise

The load-bearing premise is that the reasoning steps shown to users are a faithful externalization of the model's actual interpretation, so that deleting or editing a displayed step genuinely redirects the generative trajectory; if the surface does not reflect the model's real logic, the measured effects are reactions to a UI, not a redistribution of authority.

Editorial extensions

If this is right

  • If the paper is right, transparency without contestability is insufficient: tools that show reasoning but forbid intervention leave interpretive authority intact.
  • Beginners, who are most vulnerable to interpretive monopoly, gain a concrete means to reject the model's definition of their intentions, changing their role from prompt-guessers to empowered participants.
  • Intermediates can stabilize non-mainstream practices against incremental normalization—a capability they lacked under prompt-only and transparency-only conditions.
  • Advanced practitioners can block the system from recoding minority aesthetics as errors, making plurality a matter of legitimacy rather than stylistic variance.
  • Authority over the knowing-doing passage becomes a designable property: redistributing it does not require changing the model, only how its reasoning is surfaced and acted upon.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The same three principles could be tested in other generative domains where tacit intention is compressed into prompts—visual art, writing, code—with the prediction that editable reasoning surfaces would similarly redistribute authority, though the paper itself leaves this as future work.
  • A cleaner experiment would separate the plurality safeguards (drift notices and branches) from editability, since Mutelier bundles both; if plurality alone—without editable reasoning—can protect minority aesthetics, then the three principles may act independently, not only in combination.
  • The paper's own example admits imperfect propagation ('The next version did not fully respect this choice'); a testable corollary is that making propagation failures visible (e.g., marking which edited steps were or were not honored) could further shift authority by making the model's resistance itself contestable.
  • A long-term implication: if redistribution persists beyond the study, repeated contestation could train users to externalize and refine their tacit intentions, potentially changing how they brief any generative tool, not just Mutelier.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper reframes the knowing–doing gap in human–AI co-creation as a question of interpretive authority. It proposes three principles—contestability, agency, and plurality—and operationalizes them in Mutelier, a music-generation system with an editable reasoning surface, drift notices, and branching safeguards, compared against a prompt–output baseline and a transparency-only condition. 180 practitioners completed a three-condition crossover (two weeks per condition with one-week washouts); data include a 15-item Likert questionnaire, interaction logs, and six follow-up interviews. The authors report a consistent gradient across conditions and claim that authority in the knowing–doing cycle is redistributable, transforming creative collaboration from a model-controlled pipeline into a negotiated medium.

Significance. The question is timely and important: mainstream prompt–output tools obscure interpretive decisions, and a design probe that makes reasoning contestable could be a meaningful contribution to critical HCI and creativity-support research. The study's scale (N=180), crossover structure, mixed methods, and attention to differential effects across expertise levels are real strengths. If the evidence were analytically complete, the paper would be a noteworthy advance. However, as reported, the quantitative results cannot be independently checked, several outcome items restate the manipulated affordances, and the implementation of the editable reasoning surface is underspecified. The central empirical claim that interpretive authority is genuinely redistributed is therefore not yet established.

major comments (4)
  1. [§4.2.2, Fig. 3, §3.3] No descriptive or inferential statistics are reported. The claim that contestability/agency/plurality scores show a 'clear gradient' is supported only by radar plots of means; there are no standard deviations, confidence intervals, per-item values, test statistics, or effect sizes. Moreover, although the design is described as a counterbalanced crossover (§3.3), order and carryover effects are never analyzed. A repeated-measures model with condition order and expertise group as factors is needed; without it the gradient could reflect learning, fatigue, or sequence effects. Please provide a table of means/SD per item and condition, mixed-model results, and pairwise comparisons with effect sizes.
  2. [Table 4] Several questionnaire items restate the intervention's affordances almost verbatim, making the measured gradient partly by construction. C1 ('I could see how the system interpreted my intentions') is impossible in Baseline and trivially present in Transparency-only and Mutelier; G4 ('I did not feel restricted to prompt retries') describes only Mutelier; P2 ('The system helped me avoid unwanted regression to common defaults') refers directly to Mutelier's drift notices. These are manipulation checks, not outcome measures of authority redistribution. The paper should either validate the questionnaire as measuring latent constructs separate from surface affordances (e.g., perceived legitimacy, authorship, ability to affect final outcomes) or re-analyze the data excluding such items and show that the gradient remains for non-affordance items.
  3. [§3.2, §4.2.2] The faithfulness of the editable reasoning surface is unsubstantiated. No implementation details are given for how GPT-4o's reasoning is extracted, how user edits propagate to generation, how drift is detected, or how branches are constructed. The only traced example (B1) is a single qualitative narrative, and advanced user A2 explicitly reports that 'The next version did not fully respect this choice.' Without evidence—such as log-based fidelity metrics or ablations—that editing displayed reasoning causally changes the generative trajectory, the results may demonstrate perceived control over a UI facade rather than a genuine redistribution of interpretive authority. Please provide technical specifications and a validation of edit-to-output propagation, or substantially weaken the theoretical conclusion.
  4. [§3.2, Fig. 1] The conditions are confounded. Mutelier differs from Transparency-only not only in editability but also in the presence of drift notices and plurality-preserving branches, so the relative effect cannot be attributed to contestability, agency, and plurality as isolated principles. The observed differences could be driven by the additional UI cues or by the 'safeguard' framing rather than by editability per se. A factorial manipulation, or at least a log analysis separating the contributions of editing from branch/drift features, is needed to support the claim that the three principles individually reshape authority relations.
minor comments (5)
  1. [Table 2] The Beginner row appears malformed: '20.20.6±0.3' should likely read '20.2 ± 0.6' or similar. Please correct the formatting and check all mean/SD values.
  2. [Fig. 3] The Likert scale is 1–7, but the radar plots appear to use a 1–6 axis. Please clarify the scale, add error bars or confidence bands, and report the underlying numeric values.
  3. [References] There are duplicate references for IntentTuner ([84] and [90]), and the ACM Reference Format still contains placeholder text ('Conference acronym ’XX', 2018, DOI). These need to be updated.
  4. [§4.2.3, Fig. 1 caption] Minor language issues: 'a authority practice' should be 'an authority practice,' and 'Toget her' in the Fig. 1 caption should be 'Together.'
  5. [§5.3] The ethics statement says local regulations do not require formal ethics review but does not state whether institutional approval was nonetheless sought or waived. Please clarify the review status and consent process.

Circularity Check

1 steps flagged · score 4.0 of 10

Quantitative gradient is partly a manipulation check: key questionnaire items restate Mutelier's affordances, though interviews and logs give independent evidence.

  1. self definitional [Table 4 (items C1, G4, P2) vs §3.2 'Comparable Conditions' and Figure 1]
    "Mutelier ... externalized reasoning as an editable surface, enabled users to redirect the generative process, and issued drift notices with plurality-preserving branches when regression to defaults was detected. [Table 4:] C1 'I could see how the system interpreted my intentions.' G4 'I did not feel restricted to prompt retries but could actively intervene in the process.' P2 'The system helped me avoid unwanted regression to common defaults (e.g., C major, 4/4).'"

    The independent variable (Mutelier) is defined by an editable reasoning surface, process redirection, and drift notices; the outcome items C1, G4, and P2 ask participants to rate exactly those affordances. The between-condition differences on these items are therefore entailed by the condition definitions rather than being independent evidence that 'authority relations' were reshaped. The paper presents the questionnaire gradient as support for the central claim, so part of the quantitative demonstration reduces to a manipulation check. The interview excerpts and interaction logs provide some non-circular evidence, which is why the circularity is partial.

full rationale

The paper's central claim is that contestability, agency, and plurality reshape human-AI authority relations. The strongest circularity concern is measurement: several Likert items in Table 4 (C1, G4, P2) restate the defining affordances of the Mutelier condition (visible reasoning, editable process, drift protection). Scoring higher on those items is close to tautological. However, the study also includes questionnaire items about felt legitimacy and fairness (C2, C5, P5), plus interview quotes and interaction logs that go beyond mere affordance recognition. There is no load-bearing self-citation chain, no imported uniqueness theorem, and no fitted parameter renamed as a prediction. The unverified faithfulness of the reasoning surface is a missing-evidence limitation, not a circular step. Overall, the quantitative gradient is partly by construction, but the paper retains independent qualitative and behavioral evidence, so the circularity is moderate (4/10) rather than total.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

No numeric free parameters apply to this empirical HCI study; the design-level choices that play the role of adjustable inputs are captured in the axioms list. The most consequential is the unverified fidelity of the reasoning surface to GPT-4o's actual interpretation logic, followed by the assumption that the conditions isolate affordances, the validity of the self-report measures, the adequacy of the crossover design controls, and the representativeness of the sample. No invented entities: Mutelier is a designed prototype, not a postulated theoretical entity.

assumptions (5)
  • domain assumption GPT-4o's internal interpretation can be faithfully externalized as an editable surface of symbolic steps, and editing those steps redirects subsequent output.
    Mutelier's core mechanism (§3.2). No implementation detail is given; the paper's own example shows imperfect propagation ('The next version did not fully respect this choice', §4.2.2), so full fidelity is unverified.
  • domain assumption The three conditions differ only in interaction affordances, not in model capability.
    Stated in §3.2 ('the difference lies only in how reasoning is surfaced and acted upon'), but Mutelier adds drift notices and plurality-preserving branches, i.e., extra detection and branching machinery the other conditions lack.
  • domain assumption Fifteen Likert items grouped into three dimensions validly measure redistribution of interpretive authority.
    Table 4 shows items C1, G1, G4, P2 that restate the intervention's affordances; no reliability or construct-validity analysis is reported.
  • domain assumption A three-condition, two-week-per-condition crossover with one-week washout controls carryover and order effects.
    §3.3 states the design 'reduced carryover effects', but no order, period, or carryover analysis is reported; attrition across six weeks is not discussed.
  • domain assumption The recruited 180 practitioners are representative of music practitioners generally.
    §3.4: recruitment ran through one music-technology company's networks; eligibility required active music-making within the past month, but sampling bias and compensation are not discussed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Who Decides How Knowing Becomes Doing? Redistributing Authority in Human-AI Music Co-Creation." pith.science (2026). https://pith.science/paper/TZXXGVHF

@misc{pith2026250910331,
  author       = {Pith},
  title        = {Pith review of: Who Decides How Knowing Becomes Doing? Redistributing Authority in Human-AI Music Co-Creation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TZXXGVHF}},
  note         = {Machine review of arXiv:2509.10331}
}
read the original abstract

In the era of human-AI co-creation, the maxim "knowing is easy, doing is hard" is redefined. AI has the potential to ease execution, yet the essence of "hard" lies in who governs the translation from knowing to doing. Mainstream tools often centralize interpretive authority and homogenize expression, suppressing marginal voices. To address these challenges, we introduce the first systematic framework for redistributing authority in the knowing-doing cycle, built on three principles, namely contestability, agency, and plurality. Through interactive studies with 180 music practitioners, complemented by in-depth interviews, we demonstrate that these principles reshape human-AI authority relations and reactivate human creative expression. The findings establish a new paradigm for critical computing and human-AI co-creation that advances from critique to practice.

Figures

Figures reproduced from arXiv: 2509.10331 by the authors.

Figure 1
Figure 1. Comparable conditions. (a) Baseline prompt–output hides reasoning and limits users to prompt retries. (b) [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Study procedure with three conditions in counterbalanced order. Each condition lasted two weeks, followed by a [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Radar plots of mean ratings for three evaluative dimensions aligned with the research questions: Contestability [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

89 extracted references · 5 linked inside Pith

  1. [1]

    Ian Arawjo, Chelse Swoopes, Priyan Vaithilingam, Martin Wattenberg, and Elena L Glassman. 2024. Chainforge: A visual toolkit for prompt engineering and llm hypothesis testing. InProceedings of the 2024 CHI Conference on Human Factors in Computing Systems. 1–18

    2024

  2. [2]

    Shaowen Bardzell. 2010. Feminist HCI: taking stock and outlining an agenda for design. InProceedings of the SIGCHI conference on human factors in computing systems. 1301–1310

    2010

  3. [3]

    Maciej Besta, Torsten Hoefler, and et al. 2024. Graph of Thoughts: Solving Elaborate Problems with Large Language Models. InAAAI. https://ojs.aaai.org/ index.php/AAAI/article/view/29918

    2024

  4. [4]

    Pierre Bourdieu. 2018. Distinction a social critique of the judgement of taste. In Inequality. Routledge, 287–318

    2018

  5. [5]

    Vincent Cavez, Catherine Letondal, Caroline Appert, and Emmanuel Pietriga

  6. [6]

    DaEun Choi, Kihoon Son, Hyunjoon Jung, and Juho Kim. 2025. Expandora: Broadening Design Exploration with Text-to-Image Model. InExtended Abstracts of the 2025 CHI Conference on Human Factors in Computing Systems (CHI EA ’25). ACM. doi:10.1145/3706599.3720189

    arXiv 2025

  7. [7]

    Youjin Choi, JaeYoung Moon, JinYoung Yoo, and Jin-Hyuk Hong. 2025. Under- standing the Potentials and Limitations of Prompt-based Music Generative AI. In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems. 1–15

    2025

  8. [8]

    Yunjae J Choi, Minha Lee, and Sangsu Lee. 2023. Toward a multilingual conver- sational agent: Challenges and expectations of code-mixing multilingual users. InProceedings of the 2023 CHI conference on human factors in computing systems. 1–17

    2023

Show all 89 references
  1. [9]

    2020.Design justice: Community-led practices to build the worlds we need

    Sasha Costanza-Chock. 2020.Design justice: Community-led practices to build the worlds we need. The MIT Press

    2020

  2. [10]

    2014.Aristotle: nicomachean ethics

    Roger Crisp. 2014.Aristotle: nicomachean ethics. Cambridge University Press

    2014

  3. [11]

    Hai Dang, Lukas Mecke, and Daniel Buschek. 2022. Ganslider: How users control generative models for images using multiple sliders with and without feedforward information. InProceedings of the 2022 CHI Conference on Human Factors in Computing Systems. 1–15

    2022

  4. [12]

    Nicholas Davis and Janet Rafner. 2025. AI Drawing Partner: Co-Creative Drawing Agent and Research Platform to Model Co-Creation.arXiv preprint arXiv:2501.06607(2025)

    2025 arXiv

  5. [13]

    John Dewey. 2024. Art as experience. InAnthropology of the Arts. Routledge, 37–45

    2024

  6. [15]

    Upol Ehsan, Q Vera Liao, Michael Muller, Mark O Riedl, and Justin D Weisz

  7. [16]

    Will Epperson, Gagan Bansal, Victor C Dibia, Adam Fourney, Jack Gerrits, Erkang Zhu, and Saleema Amershi. 2025. Interactive debugging and steering of multi- agent ai systems. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems. 1–15

    2025

  8. [17]

    Motahhare Eslami, Karrie Karahalios, Christian Sandvig, Kristen Vaccaro, Aimee Rickman, Kevin Hamilton, and Alex Kirlik. 2016. First I" like" it, then I hide it: Folk Theories of Social Feeds. InProceedings of the 2016 cHI conference on human factors in computing systems. 2371–2382

    2016

  9. [18]

    1999.The Beatles as musicians: Revolver through the Anthology

    Walter Everett. 1999.The Beatles as musicians: Revolver through the Anthology. Oxford University Press, USA

    1999

  10. [19]

    Noyan Evirgen and Xiang’Anthony Chen. 2023. Ganravel: User-driven direction disentanglement in generative adversarial networks. InProceedings of the 2023 CHI Conference on Human Factors in Computing Systems. 1–15

    2023

  11. [20]

    Michel Foucault. 2020. Power/knowledge. InThe new social theory reader. Routledge, 73–79

    2020

  12. [22]

    GPT-4o. 2025. Official website. https://chatgpt.com/. Accessed: 2025-09-01

    2025

  13. [23]

    Ben Green and Yiling Chen. 2019. Disparate interactions: An algorithm-in-the- loop analysis of fairness in risk assessments. InProceedings of the conference on fairness, accountability, and transparency. 90–99

    2019

  14. [24]

    Jessica He, Stephanie Houde, and Justin D. Weisz. 2025. Which Contributions Deserve Credit? Perceptions of Attribution in Human–AI Co-Creation. InPro- ceedings of the 2025 CHI Conference on Human Factors in Computing Systems (CHI ’25). ACM. doi:10.1145/3706598.3713522

    2025

  15. [25]

    Twelve-Tone Method

    Martha M Hyde. 1985. Musical Form and the Development of Schoenberg’s" Twelve-Tone Method".Journal of Music Theory29, 1 (1985), 85–143

    1985

  16. [26]

    Alon Jacovi, Gregor Geigle, and et al. 2024. REVEAL: A Benchmark for Verifiers of Reasoning Chains. InACL. https://aclanthology.org/2024.acl-long.254/

    2024

  17. [28]

    Maurice Jakesch, Advait Bhat, Daniel Buschek, Lior Zalmanson, and Mor Naaman

  18. [29]

    Minsuk Kahng and et al. 2024. LLM Comparator: Visual Analytics for Side-by-Side Evaluation of Large Language Models. InCHI. https://arxiv.org/abs/2402.10524

    2024 arXiv

  19. [30]

    2000.Critique of the Power of Judgment

    Immanuel Kant. 2000.Critique of the Power of Judgment. Cambridge University Press

    2000

  20. [31]

    Shivani Kapania, Alex S Taylor, and Ding Wang. 2023. A hunt for the snark: Annotator diversity in data practices. InProceedings of the 2023 CHI Conference on Human Factors in Computing Systems. 1–15

    2023

  21. [32]

    Abidullah Khan, Atefeh Shokrizadeh, and Jinghui Cheng. 2025. Beyond Automa- tion: How UI/UX Designers Perceive AI as a Creative Partner in the Divergent Thinking Stages. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems (CHI ’25). ACM

    2025

  22. [33]

    Chae-Young Kim, Soojeong Kim, Jong-Yun Jang, and Sang-Hyeok Choi. 2025. Bridging Generations Using AI-Supported Co-Creative Activities. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems (CHI ’25). ACM. doi:10.1145/3706598.3713718

    2025

  23. [35]

    Faye Kollig, Jessica Pater, Fayika Farhat Nova, and Casey Fiesler. 2025. Fictional Failures and Real-World Lessons: Ethical Speculation Through Design Fiction on Emotional Support Conversational AI. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems. 1–15

    2025

  24. [36]

    male nurse

    Simone Kopeinik, Martina Mara, Linda Ratz, Klara Krieg, Markus Schedl, and Navid Rekabsaz. 2023. Show me a" male nurse"! how gender bias is reflected in the query formulation of search engine users. InProceedings of the 2023 CHI Conference on Human Factors in Computing Systems. 1–15

    2023

  25. [37]

    Stephen James Krol, Maria Teresa Llano Rodriguez, and Miguel J Loor Paredes

  26. [38]

    ChungHa Lee, DaeHo Lee, and Jin-Hyuk Hong. 2025. MVPrompt: Building Music- Visual Prompts for AI Artists to Craft Music Video Mise-en-scène. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems. 1–21

    2025

  27. [39]

    Min Kyung Lee, Daniel Kusbit, Anson Kahng, Ji Tae Kim, Xinran Yuan, Allissa Chan, Daniel See, Ritesh Noothigattu, Siheon Lee, Alexandros Psomas, et al. 2019. WeBuildAI: Participatory framework for algorithmic governance.Proceedings of the ACM on human-computer interaction3, CS...

    2019

  28. [40]

    artificial intelli- gence as a service

    Kornel Lewicki, Michelle Seng Ah Lee, Jennifer Cobbe, and Jatinder Singh. 2023. Out of context: Investigating the bias and fairness concerns of “artificial intelli- gence as a service”. InProceedings of the 2023 CHI Conference on Human Factors in Computing Systems. 1–17

    2023

  29. [41]

    InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems

    Exploring the Needs of Practising Musicians in Co-Creative AI Through Co- Design. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems. 1–13

    2025

  30. [42]

    Susan Lin, Jeremy Warner, J. D. Zamfirescu-Pereira, Matthew G. Lee, Sauhard Jain, Shanqing Cai, Piyawat Lertvittayakumjorn, Michael Huang, Shumin Zhai, Björn Hartmann, and Can Liu. 2024. Rambler: Supporting Writing with Speech via LLM- Assisted Gist Manipulation. InProceedings...

    2024

  31. [43]

    Vivian Liu. 2023. Beyond Text-to-Image: Multimodal Prompts to Explore Gener- ative AI. InExtended Abstracts of the 2023 CHI Conference on Human Factors in Computing Systems (CHI EA ’23). ACM, 482:1–482:6. doi:10.1145/3544549.3577043

    2023

  32. [44]

    Vivian Liu and Lydia B Chilton. 2022. Design guidelines for prompt engineering text-to-image generative models. InProceedings of the 2022 CHI conference on human factors in computing systems. 1–23

    2022

  33. [45]

    Haichuan Lin, Yilin Ye, Jiazhi Xia, and Wei Zeng. 2025. SketchFlex: Facilitating Spatial–Semantic Coherence in Text-to-Image Generation with Region-Based Sketches. InProceedings of the 2025 CHI Conference on Human Factors in Comput- ing Systems (CHI ’25). ACM. doi:10.1145/3706...

    2025

  34. [46]

    Atefeh Mahdavi Goloujeh, Anne Sullivan, and Brian Magerko. 2024. Is it AI or is it me? Understanding users’ prompt journey with text-to-image generative AI tools. InProceedings of the 2024 CHI Conference on Human Factors in Computing Systems. 1–13

    2024

  35. [47]

    Sullivan, and Brian Magerko

    Atefeh Mahdavi Goloujeh, Anne M. Sullivan, and Brian Magerko. 2024. The Social Construction of Generative AI Prompts. InExtended Abstracts of the 2024 CHI Conference on Human Factors in Computing Systems (CHI EA ’24). ACM, 320:1–320:7. doi:10.1145/3613905.3650947

    2024

  36. [48]

    Martin E Malandro. 2024. Composer’s Assistant 2: Interactive Multi-Track MIDI Infilling with Fine-Grained User Control.arXiv preprint arXiv:2407.14700(2024)

    2024 arXiv

  37. [49]

    Henrietta Lyons, Eduardo Velloso, and Tim Miller. 2021. Conceptualising con- testability: Perspectives on contesting algorithmic decisions.Proceedings of the ACM on Human-Computer Interaction5, CSCW1 (2021), 1–25

    2021

  38. [50]

    Damien Masson, Sylvain Malacria, Géry Casiez, and Daniel Vogel. 2024. Direct- GPT: A Direct Manipulation Interface to Interact with Large Language Models. InProceedings of the 2024 CHI Conference on Human Factors in Computing Systems (CHI ’24). ACM. doi:10.1145/3613904.3642462

    2024

  39. [51]

    1954.Emotion and meaning in music

    Leonard B Meyer. 1954.Emotion and meaning in music. Ph. D. Dissertation. The University of Chicago

    1954

  40. [52]

    Midjourney. 2025. Official website. https://www.midjourney.com/. Accessed: 2025-08-01

    2025

  41. [53]

    2016.Schoenberg and Hollywood Modernism

    Kenneth H Marcus. 2016.Schoenberg and Hollywood Modernism. Cambridge University Press

    2016

  42. [55]

    Xiaohan Peng, Janin Koch, and Wendy E. Mackay. 2025. FusAIn: Composing Generative AI Visual Prompts Using Pen-based Interaction. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems (CHI ’25). ACM. doi:10.1145/3706598.3714027

    2025

  43. [56]

    Julien Porquet, Sitong Wang, and Lydia B Chilton. 2025. Copying style, Extracting value: Illustrators’ Perception of AI Style Transfer and its Impact on Creative Labor. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems. 1–16

    2025

  44. [57]

    Xingang Pan, Ayush Tewari, Thomas Leimkühler, Lingjie Liu, Abhimitra Meka, and Christian Theobalt. 2023. Drag your gan: Interactive point-based manip- ulation on the generative image manifold. InACM SIGGRAPH 2023 conference proceedings. 1–11

    2023

  45. [58]

    Nina Rajcic, Maria Teresa Llano Rodriguez, and Jon McCormack. 2024. Towards a Diffractive Analysis of Prompt-Based Generative AI. InProceedings of the 2024 CHI Conference on Human Factors in Computing Systems (CHI ’24). ACM, 844:1–844:15. doi:10.1145/3613904.3641971

    2024

  46. [59]

    2013.The politics of aesthetics

    Jacques Rancière. 2013.The politics of aesthetics. A&C Black

    2013

  47. [60]

    Mohi Reza, Nathan Laundry, Ilya Musabirov, Peter Dushniku, Zhi Yuan, Kashish Mittal, Tovi Grossman, Michael Liut, Anastasia Kuzminykh, and Joseph Jay Williams. 2024. ABScribe: Rapid Exploration & Organization of Multiple Writing Variations in Human–AI Co-Writing Tasks Using La...

    2024

  48. [61]

    Rida Qadri, Piotr Mirowski, and Remi Denton. 2025. AI and Non-Western Art Worlds: Reimagining Critical AI Futures through Artistic Inquiry and Situated Di- alogue. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems. 1–17

    2025

  49. [62]

    male code

    Katie Seaborn, Shruti Chandra, and Thibault Fabre. 2023. Transcending the “male code”: implicit masculine biases in NLP contexts. InProceedings of the 2023 CHI Conference on Human Factors in Computing Systems. 1–19

    2023

  50. [63]

    Phoebe Sengers, Kirsten Boehner, Shay David, and Joseph’Jofish’ Kaye. 2005. Re- flective design. InProceedings of the 4th decennial conference on Critical computing: between sense and sensibility. 49–58

    2005

  51. [64]

    Zekai Shao, Siyu Yuan, Lin Gao, Yixuan He, Deqing Yang, and Siming Chen. 2025. Unlocking Scientific Concepts: How Effective Are LLM-Generated Analogies for Student Understanding and Classroom Practice?. InProceedings of the 2025 CHI Conference on Human Factors in Computing Sys...

    2025

  52. [65]

    1984.Style and Idea: selected writings of Arnold Schoenberg

    Arnold Schoenberg and Leonard Stein. 1984.Style and Idea: selected writings of Arnold Schoenberg. Univ of California Press

    1984

  53. [66]

    Noah Shinn, Federico Cassano, Ashwin Gopinath, Karthik Narasimhan, and Shunyu Yao. 2023. Reflexion: Language Agents with Verbal Reinforcement Learning. InNeurIPS. https://papers.nips.cc/paper_files/paper/2023/hash/ 1b44b878bb782e6954cd888628510e90-Abstract-Conference.html

    2023

  54. [67]

    Momin N Siddiqui, Roy D Pea, and Hari Subramonyam. 2025. Script&Shift: A layered interface paradigm for integrating content development and rhetorical strategy with llm writing assistants. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems. 1–19

    2025

  55. [68]

    2012.The Beatles: the biography

    Bob Spitz. 2012.The Beatles: the biography. Hachette UK

    2012

  56. [69]

    Yujun Shi, Chuhui Xue, Jun Hao Liew, Jiachun Pan, Hanshu Yan, Wenqing Zhang, Vincent YF Tan, and Song Bai. 2024. Dragdiffusion: Harnessing diffusion mod- els for interactive point-based image editing. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Reco...

    2024

  57. [70]

    Suno. 2025. Official website. https://suno.com. Accessed: 2025-08-01

    2025

  58. [71]

    Ethics is not neutral

    Tiera Tanksley, Angela DR Smith, Saloni Sharma, and Earl W Huff Jr. 2025. " Ethics is not neutral": Understanding Ethical and Responsible AI Design from the Lenses of Black Youth. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems. 1–20

    2025

  59. [72]

    Kristen Vaccaro, Karrie Karahalios, Deirdre K Mulligan, Daniel Kluttz, and Tad Hirsch. 2019. Contestability in algorithmic systems. InCompanion Publication of the 2019 Conference on Computer Supported Cooperative Work and Social Comput- ing. 523–527

    2019

  60. [73]

    2007.Human-machine reconfigurations: Plans and situated actions

    Lucille Alice Suchman. 2007.Human-machine reconfigurations: Plans and situated actions. Cambridge university press

    2007

  61. [74]

    Wen-Fan Wang, Chien-Ting Lu, Nil Ponsa Campanyà, Bing-Yu Chen, and Mike Y. Chen. 2025. AIdeation: Designing a Human–AI Collaborative Ideation System for Concept Designers. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems (CHI ’25). ACM. doi:10.114...

    2025

  62. [75]

    Xuezhi Wang, Jason Wei, Dale Schuurmans, and et al. 2023. Self-Consistency Improves Chain of Thought Reasoning in Language Models. InICLR. https: //openreview.net/forum?id=1PL1NIMMrw

    2023

  63. [76]

    Zhijie Wang, Yuheng Huang, Da Song, Lei Ma, and Tianyi Zhang. 2024. PromptCharm: Text-to-Image Generation through Multi-modal Prompting and Refinement. InProceedings of the 2024 CHI Conference on Human Factors in Com- puting Systems (CHI ’24). ACM, 185:1–185:21. doi:10.1145/36...

    2024

  64. [77]

    It Felt Like Having a Second Mind

    Qian Wan, Siying Hu, Yu Zhang, Piaohong Wang, Bo Wen, and Zhicong Lu. 2024. " It Felt Like Having a Second Mind": Investigating Human-AI Co-creativity in Prewriting with Large Language Models.Proceedings of the ACM on human- computer interaction8, CSCW1 (2024), 1–26

    2024

  65. [78]

    1966.Lectures & conversations on aesthetics, psychology, and religious belief

    Ludwig Wittgenstein. 1966.Lectures & conversations on aesthetics, psychology, and religious belief. Univ of California Press

    1966

  66. [79]

    Jundong Xu, Hao Fei, Liangming Pan, Qian Liu, Mong-Li Lee, and Wynne Hsu

  67. [80]

    Shunyu Yao, Jeffrey Zhao, Dian Yu, and et al. 2023. ReAct: Synergizing Reasoning and Acting in Language Models. InICLR. https://arxiv.org/abs/2210.03629

    2023 arXiv

  68. [81]

    Jason Wei, Xuezhi Wang, Dale Schuurmans, and et al. 2022. Chain- of-Thought Prompting Elicits Reasoning in Large Language Models. In NeurIPS. https://proceedings.neurips.cc/paper_files/paper/2022/hash/ 9d560961f0aa0a8a1e3953eacb5f63b0-Abstract-Conference.html

    2022

  69. [82]

    Mireia Yurrita, Tim Draws, Agathe Balayn, Dave Murray-Rust, Nava Tintarev, and Alessandro Bozzon. 2023. Disentangling fairness perceptions in algorithmic decision-making: the effects of explanations, human oversight, and contestability. InProceedings of the 2023 CHI Conference...

    2023

  70. [85]

    Chao Zhang, Cheng Yao, Jiayi Wu, Weijia Lin, Lijuan Liu, Ge Yan, and Fangtian Ying. 2022. StoryDrawer: a child–AI collaborative drawing system to support children’s creative visual storytelling. InProceedings of the 2022 CHI conference on human factors in computing systems. 1–15

    2022

  71. [86]

    Shunyu Yao, Jeffrey Zhao, Dian Yu, and et al. 2023. Tree of Thoughts: Deliberate Problem Solving with Large Language Models. InNeurIPS. https://arxiv.org/abs/ 2305.10601

    2023 arXiv

  72. [87]

    Ruochen Zhao, Xingxuan Li, Shafiq Joty, Chengwei Qin, and Lidong Bing. 2023. Verify-and-Edit: A Knowledge-Enhanced Chain-of-Thought Framework. InACL. https://aclanthology.org/2023.acl-long.320/

    2023

  73. [88]

    Xi Zheng, Zhuoyang Li, Xinning Gui, and Yuhan Luo. 2025. Customizing emo- tional support: How do individuals construct and interact with LLM-powered chatbots. InProceedings of the 2025 CHI Conference on Human Factors in Comput- ing Systems. 1–20. Who Decides How Knowing Become...

    2025

  74. [89]

    Ruican Zhong, Donghoon Shin, Rosemary Meza, Predrag Klasnja, Lucas Colusso, and Gary Hsieh. 2024. AI-assisted causal pathway diagram for human-centered design. InProceedings of the 2024 CHI Conference on Human Factors in Computing Systems. 1–19

    2024

  75. [90]

    Ruican Zhong, Donghoon Shin, Rosemary Meza, Predrag Klasnja, Lucas Colusso, and Gary Hsieh. 2024. AI-Assisted Causal Pathway Diagram for Human-Centered Design. InProceedings of the 2024 CHI Conference on Human Factors in Computing Systems (CHI ’24). ACM. doi:10.1145/3613904.3642179

    2024

  76. [91]

    Xuan Zhang, Chao Du, Tianyu Pang, and et al. 2024. Improving Chain-of-Thought Reasoning in LLMs via Chain-of-Preference Optimiza- tion. InNeurIPS. https://proceedings.neurips.cc/paper_files/paper/2024/file/ 00d80722b756de0166523a87805dd00f-Paper-Conference.pdf

    2024

  77. [96]

    Yao Zhou, Nathanael Schärli, Le Hou, and et al. 2023. Least-to-Most Prompt- ing Enables Complex Reasoning in Large Language Models. InICLR. https: //openreview.net/forum?id=WZH7099tgf

    2023

  78. [2021]

    In Proceedings of the 2021 CHI conference on human factors in computing systems

    Expanding explainability: Towards social transparency in ai systems. In Proceedings of the 2021 CHI conference on human factors in computing systems. 1–19

    2021

  79. [2023]

    In Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems (CHI ’23)

    Co-Writing with Opinionated Language Models Affects Users’ Views. In Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems (CHI ’23). ACM. doi:10.1145/3544548.3581196

    2023

  80. [2024]

    Faithful Logical Reasoning via Symbolic Chain-of-Thought. InACL. https: //aclanthology.org/2024.acl-long.720/

    2024

  81. [2025]

    In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems

    EuterPen: Unleashing Creative Expression in Music Score Writing. In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems. 1–16

    2025

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.