pith. sign in

arxiv: 2605.16273 · v1 · pith:DRYBFXFPnew · submitted 2026-04-05 · 💻 cs.HC

AI-Generated 3D Environments as Speculative Mediators in More-Than-Human Design: An Exploratory Study

Aung Pyae This is my paper

Pith reviewed 2026-05-21 10:30 UTC · model grok-4.3

classification 💻 cs.HC
keywords more-than-human designgenerative AI3D environmentsspeculative mediationreflection-in-actionanthropocentric assumptionsdesign explorationAI-generated worlds
0
0 comments X

The pith

Navigating AI-generated 3D environments supports reflection-in-action distinct from static evaluation and lets designers oscillate between treating outputs as provocations and as authoritative representations.

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

The paper explores how a text-to-3D generation tool can create navigable environments based on traces from non-human entities and serve as a mediator in more-than-human design. Participants engaged with these worlds in an exploratory study and the work finds that real-time navigation prompts a form of reflection and iteration different from reviewing fixed images. Designers shift between using the outputs to spark ideas and accepting them as factual depictions. This matters because more-than-human design requires questioning human-centered priorities without direct access to other entities experiences.

Core claim

The central claim is that instant exploration of AI-generated navigable 3D environments derived from non-human traces shapes reflection, iteration, and provisional treatment of outputs. Designers epistemic stances oscillate between viewing the results as generative provocations and as authoritative representations. The study proposes technologically-amplified backtalk and productive provisionality as lenses for how such environments can surface anthropocentric assumptions in design practice.

What carries the argument

The text-to-3D world generation platform that produces instantly navigable environments from non-human traces, acting as the speculative mediator that enables real-time exploration and stance oscillation during design.

Load-bearing premise

The AI-generated 3D outputs derived from non-human traces can reliably surface anthropocentric assumptions without the generative model itself introducing unexamined biases that shape participant reflections.

What would settle it

If a follow-up study found no measurable difference in reflection depth or type between participants who navigated the 3D environments and those who only viewed static images of the same outputs, the claim of distinct reflection-in-action from navigation would not hold.

Figures

Figures reproduced from arXiv: 2605.16273 by Aung Pyae.

Figure 1
Figure 1. Figure 1: Navigable AI-generated environments surfacing anthropocentric assumptions in more-than-human design More-than-human design challenges anthropocentric assumptions by foregrounding non-human entities as stakeholders, yet designers face an epistemic boundary: they cannot directly access non-human experience. We present an exploratory study examining how generative AI—specifically a text-to-3D world generation… view at source ↗
read the original abstract

More-than-human design challenges anthropocentric assumptions by foregrounding non-human entities as stakeholders, yet designers face an epistemic boundary: they cannot directly access non-human experience. We present an exploratory study examining how generative AI -- specifically a text-to-3D world generation platform producing navigable environments -- may function as a speculative mediator in more-than-human design. Through a qualitative study with five participants from engineering and sustainability backgrounds engaging with AI-generated worlds derived from non-human traces, we investigate how instant exploration -- navigating generated environments within seconds -- shapes reflection, iteration, and provisional treatment of outputs. Our findings suggest that navigating AI-generated environments supports reflection-in-action distinct from evaluating static representations, while designers' epistemic stances oscillate between treating outputs as generative provocations and as authoritative representations. We propose technologically-amplified backtalk and productive provisionality as preliminary lenses for understanding how navigable AI-generated 3D environments can surface anthropocentric assumptions in more-than-human design.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript presents an exploratory qualitative study with five participants from engineering and sustainability backgrounds. It investigates a text-to-3D AI platform that generates navigable environments from non-human traces as a speculative mediator for more-than-human design. The central claims are that instant navigation supports reflection-in-action distinct from evaluating static representations, that designers' epistemic stances oscillate between treating outputs as generative provocations versus authoritative representations, and that this process surfaces anthropocentric assumptions via the proposed lenses of technologically-amplified backtalk and productive provisionality.

Significance. If the observations hold, the work could provide HCI and design researchers with a practical method for using generative AI to engage non-human perspectives in speculative design, extending reflection-in-action concepts to AI-mediated environments. The exploratory findings on epistemic oscillation offer preliminary conceptual tools, though the small sample and interpretive nature limit immediate applicability beyond opening avenues for further controlled studies.

major comments (2)
  1. [§4 (Study Design and Participants)] §4 (Study Design and Participants): The qualitative study with five participants includes no control conditions (e.g., static renders, text-only prompts, or non-AI 3D navigation). This is load-bearing for the central claim that navigability itself produces distinct reflection-in-action and epistemic oscillation, as effects may instead arise from the generative model's training data, prompt artifacts, or general novelty of 3D tools.
  2. [§5 (Findings on Reflection and Epistemic Stances)] §5 (Findings on Reflection and Epistemic Stances): The distinction between navigation-enabled reflection-in-action and static evaluation, along with the reported oscillation, rests on thematic analysis of participant quotes without reported inter-rater reliability, detailed coding scheme, or direct within-subject comparisons to non-navigable conditions, weakening support for the proposed lenses as specific to AI-generated navigable environments.
minor comments (2)
  1. [Abstract] The abstract and §3 could more explicitly acknowledge the preliminary status and small sample as limiting generalizability to broader more-than-human design practices.
  2. [§6 (Discussion)] §6 (Discussion) introduces 'technologically-amplified backtalk' and 'productive provisionality' with limited connections to prior HCI literature on reflection-in-action (e.g., Schön) or AI bias in generative tools; adding targeted citations would strengthen framing.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their insightful comments, which have helped us improve the clarity and rigor of our manuscript. We address each of the major comments below, agreeing with the need to better contextualize our exploratory findings and to enhance methodological transparency. We have made revisions accordingly.

read point-by-point responses

  1. Referee: [§4 (Study Design and Participants)] §4 (Study Design and Participants): The qualitative study with five participants includes no control conditions (e.g., static renders, text-only prompts, or non-AI 3D navigation). This is load-bearing for the central claim that navigability itself produces distinct reflection-in-action and epistemic oscillation, as effects may instead arise from the generative model's training data, prompt artifacts, or general novelty of 3D tools.

    Authors: We recognize the value of control conditions for isolating the specific contribution of navigability. However, given the exploratory and qualitative nature of this study, our primary aim was to surface rich, contextual insights into how designers engage with AI-generated navigable environments rather than to establish causal effects through controlled comparisons. We have added a new paragraph in the Limitations section acknowledging that effects could stem from novelty or other factors, and we explicitly recommend future work incorporating control conditions such as static renders or non-navigable interfaces to test these distinctions more rigorously. This revision clarifies the scope of our claims without overclaiming causality. revision: partial

  2. Referee: [§5 (Findings on Reflection and Epistemic Stances)] §5 (Findings on Reflection and Epistemic Stances): The distinction between navigation-enabled reflection-in-action and static evaluation, along with the reported oscillation, rests on thematic analysis of participant quotes without reported inter-rater reliability, detailed coding scheme, or direct within-subject comparisons to non-navigable conditions, weakening support for the proposed lenses as specific to AI-generated navigable environments.

    Authors: Our qualitative analysis employed a reflexive thematic analysis approach, which prioritizes interpretive depth over quantitative reliability metrics. We have expanded the Methods section to include a more detailed description of our coding process, including how themes were iteratively developed from the data and examples from our codebook. We did not conduct inter-rater reliability checks as this is not standard for reflexive approaches with small samples; instead, we relied on reflexive discussion among the research team. Regarding within-subject comparisons, the study was designed as an open exploration rather than a comparative experiment. We have added this as an explicit limitation and note that the proposed lenses are preliminary conceptual tools intended to guide future research, including studies with direct comparisons. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical findings from direct participant study

full rationale

The paper reports an exploratory qualitative study with five participants who directly engaged with AI-generated navigable 3D environments derived from non-human traces. Claims regarding reflection-in-action, epistemic oscillation, technologically-amplified backtalk, and productive provisionality are presented as emerging from observed participant interactions and interpretive coding of their reflections. No mathematical derivations, equations, fitted parameters, or predictive models are described that could reduce outputs to inputs by construction. No self-citation chains or uniqueness theorems are invoked as load-bearing premises. The derivation chain is therefore self-contained in the empirical data collection and analysis process rather than tautological.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption that non-human experience is inaccessible and that AI-generated environments can function as effective speculative mediators without model-specific artifacts dominating the outcomes.

axioms (1)
  • domain assumption Designers face an epistemic boundary because they cannot directly access non-human experience.
    Explicitly stated in the opening of the abstract as the core challenge that the AI mediator is meant to address.

pith-pipeline@v0.9.0 · 5697 in / 1304 out tokens · 53881 ms · 2026-05-21T10:30:47.983110+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

20 extracted references · 20 canonical work pages

  1. [1]

    Duke University Press, Durham, NC

    Vibrant Matter: A Political Ecology of Things. Duke University Press, Durham, NC. DOI: https://doi.org/10.1215/9780822391623. 9 Mark Blythe and Peter Wright

    work page doi:10.1215/9780822391623

  2. [2]

    Interacting with Compu ters 18, 5 (2006), 1139–1164

    Pastiche scenarios: Fiction as a resource for user centred design. Interacting with Compu ters 18, 5 (2006), 1139–1164. DOI: https://doi.org/10.1016/j.intcom.2006.02.001 Rosi Braidotti

    work page doi:10.1016/j.intcom.2006.02.001 2006

  3. [3]

    Design Issues 8, 2 (1992), 5 –21

    Wicked problems in design thinking. Design Issues 8, 2 (1992), 5 –21. DOI: https://doi.org/10.2307/1511637 Rachel Clarke, Sara Heitlinger, Marcus Foth, Carl DiSalvo, Ann Light, and Laura Forlano

    work page doi:10.2307/1511637 1992

  4. [4]

    https://doi.org/10.1145/3210604.3210641 Paul Coulton and Joseph G

    ACM, New York, NY, Article 34, 1–4. https://doi.org/10.1145/3210604.3210641 Paul Coulton and Joseph G. Lindley

    work page doi:10.1145/3210604.3210641

  5. [5]

    The Design Journal 22, 4 (2019), 463 –481

    More -than human centred design: considering other things. The Design Journal 22, 4 (2019), 463 –481. DOI: https://doi.org/10.1080/14606925.2019.1614320 Paul Coulton, Joseph Lindley, Miriam Sturdee, and Mike Stead

    work page doi:10.1080/14606925.2019.1614320 2019

  6. [6]

    In Proceedings of the 3rd Biennial Research Through Design Conference (RTD '17)

    Design fiction as world building. In Proceedings of the 3rd Biennial Research Through Design Conference (RTD '17). https://doi.org/10.6084/m9.figshare.4746964 Kees Dorst and Nigel Cross

    work page doi:10.6084/m9.figshare.4746964

  7. [7]

    Design Studies 22, 5 (2001), 425 –437

    Creativity in the design process: co-evolution of problem –solution. Design Studies 22, 5 (2001), 425 –437. DOI: https://doi.org/10.1016/S0142-694X(01)00009-6 Anil R. Doshi and Oliver P. Hauser

    work page doi:10.1016/s0142-694x(01)00009-6 2001

  8. [8]

    Generative AI enhances individual creativity but reduces the collective diversity of novel content

    Generative AI enhances individual creativity but reduces the collective diversity o f novel content. Science Advances 10, 28 (2024), eadn5290. DOI: https://doi.org/10.1126/sciadv.adn5290 Anthony Dunne and Fiona Raby

    work page doi:10.1126/sciadv.adn5290 2024

  9. [9]

    she ji: The Journal of Design, Economics, and Innovation 3, 1 (2017), 16 –29

    Posthumanism and design. she ji: The Journal of Design, Economics, and Innovation 3, 1 (2017), 16 –29. DOI: https://doi.org/10.1016/j.sheji.2017.08.001 Verena Fuchsberger and Christopher Frauenberger

    work page doi:10.1016/j.sheji.2017.08.001 2017

  10. [10]

    Human–Computer Interaction 40, 1–4 (2023), 195–220

    Doing responsibilities in entangled worlds. Human–Computer Interaction 40, 1 –4 (2023), 65–88. DOI: https://doi.org/10.1080/07370024.2023.2269934 Elisa Giaccardi, Johan Redström, and Iohanna Nicenboim

    work page doi:10.1080/07370024.2023.2269934 2023

  11. [11]

    Human–Computer Interaction 40, 1–4 (2025), 1–16

    The making(s) of more -than-human design: Introduction to the special issue on more - than-human design and HCI. Human–Computer Interaction 40, 1–4 (2025), 1–16. DOI: https://doi.org/10.1080/07370024.2024.2353357 Elisa Giaccardi, Chris Speed, Nazli Cila, and Melissa L. Caldwell

    work page doi:10.1080/07370024.2024.2353357 2025

  12. [12]

    Design Issues 19, 1 (2003), 72 –88

    The backtalk of self -generated sketches. Design Issues 19, 1 (2003), 72 –88. DOI: https://doi.org/10.1162/074793603762667728 Donna J. Haraway

    work page doi:10.1162/074793603762667728 2003

  13. [13]

    Liesen, Z

    Reassembling the Social: An Introduction to Actor -Network-Theory. Oxford University Press, Oxford. DOI: https://doi.org/10.1093/oso/9780199256044.001.0001 Joseph Lindley, Jesse Josua Benjamin, David Philip Green, Glenn McGarry, Franziska Pilling, Laura Dudek, Andy Crabtree, and P aul Coulton

    work page doi:10.1093/oso/9780199256044.001.0001

  14. [14]

    Human–Computer Interaction 40, 1–4 (2023), 195–220

    Productive oscillation as a strategy for doing more -than-human design research. Human –Computer Interaction 40, 1 –4 (2023), 117 –142. DOI: https://doi.org/10.1080/07370024.2023.2276393 Vivian Liu, Jo Vermeulen, George Fitzmaurice, and Justin Matejka

    work page doi:10.1080/07370024.2023.2276393 2023

  15. [15]

    In Proceedings of the 2023 ACM Designing Interactive Systems Conference (DIS '23)

    3DALL -E: integrating text-to-image AI in 3D design workflows. In Proceedings of the 2023 ACM Designing Interactive Systems Conference (DIS '23). ACM, New York, NY, 1955 –1977. https://doi.org/10.1145/3563657.3596098 Clara Mancini

    work page doi:10.1145/3563657.3596098 2023

  16. [16]

    Interactions 18, 4 (2011), 69 –73

    Animal-computer interaction: A manifesto. Interactions 18, 4 (2011), 69 –73. DOI: https://doi.org/10.1145/1978822.1978836 Iohanna Nicenboim, Doenja Oogjes, Heidi Biggs, and Seowoo Nam

    work page doi:10.1145/1978822.1978836 2011

  17. [17]

    Human–Computer Interaction 40, 1–4 (2023), 195–220

    Decentering through design: Bridging posthuman theory wi th more-than-human design practices. Human–Computer Interaction 40, 1–4 (2023), 195–220. DOI: https://doi.org/10.1080/07370024.2023.2283535 Iohanna Nicenboim, Joseph Lindley, and Johan Redström

    work page doi:10.1080/07370024.2023.2283535 2023

  18. [18]

    More -than-human design and AI: exploring the space between theory and practice. In C. Gray, E. Ciliotta Chehade, P. Hekkert, L. Forlano, P. Ciuccarelli, and P. Lloyd (Eds.), DRS2024: Boston, 23 –28 June, Boston, USA. Design Research Society. DOI: https://doi.org/10.21606/drs.2024.948 Raja Parasuraman and Victor Riley

    work page doi:10.21606/drs.2024.948 2024

  19. [19]

    Human Factors 39, 2 (1997), 230 –253

    Humans and automation: use, misuse, disuse, abuse. Human Factors 39, 2 (1997), 230 –253. DOI: https://doi.org/10.1518/001872097778543886 Donald A. Schön

    work page doi:10.1518/001872097778543886 1997

  20. [20]

    In Proceedings of the Third International Conference on Animal -Computer Interaction (ACI '16)

    Becoming with: Towards the inclusion of animals as participants in design processes. In Proceedings of the Third International Conference on Animal -Computer Interaction (ACI '16). ACM, New York, NY, Article 1, 1 –10. https://doi.org/10.1145/2995257.2995392

    work page doi:10.1145/2995257.2995392