Ad Arch Studies - Research Proposal
Summer Sun
12.17.2025
Abstract
Can androids dream of
electric sheep grazing
on an open plain?
This project begins with a simple but persistent question: when AI gener-
ates a world, is it imagining in the same way humans do—or is it follow-
ing an entirely different logic that only looks similar on the surface?
Rather than focusing on the visual realism or technical sophistication of
AI-generated environments, this research examines the underlying logic
of worldbuilding. Human designers—such as architects, set designers,
game designers, or lmmakers—do not construct worlds by assembling
isolated objects. Instead, they think in terms of atmosphere, spatial re-
lationships, narrative consistency, and how a world affords movement,
behavior, and meaning. Worldbuilding, for humans, is a process of imag-
ining how a space works before deciding how it looks.
This research explores how contemporary AI world-generation tools
operate by analyzing the internal logics of existing AI-driven world or
scene generation systems, and comparing them with the workows and
reasoning processes of human set designers. Through this comparison, the
project asks where AI worldbuilding diverges from human imagination: Is
AI organizing space through rules, patterns, and optimization rather than
lived intuition? Does it “understand” a world as a system of relations, or
merely reproduce statistical coherence?
Finally, the project looks forward to a speculative question: if imagina-
tion itself follows certain structural steps, could AI be trained not just to
generate worlds, but to imagine them in ways that more closely resemble
human world-thinking? In doing so, the research seeks to clarify the fun-
damental differences—and possible convergences—between human and
machine imagination.
Why This Question Matters Now
Comparison between
human and computa-
tional imagination
In recent years, a growing number of AI tools have emerged that promise
to automatically generate or assist in the construction of worlds, scenes,
and environments. From text-to-image models that produce cinematic
landscapes to AI-driven systems that assemble spatial assets or propose
entire world settings, imagination itself increasingly appears to be some-
thing that can be outsourced, accelerated, or automated. These tools are
often framed as creative partners, yet their underlying logics remain large-
ly opaque. While their outputs may appear coherent or visually convinc-
ing, it is rarely clear how these systems organize a world—or what kind
of “imagination” they are performing.
This question is particularly urgent because human worldbuilding has
never been only about producing images. Designers, lmmakers, and set
designers imagine worlds by reasoning through relationships: how spaces
connect, how movement unfolds, how atmosphere supports narrative, and
how a world sustains internal consistency over time. When AI systems
generate worlds without explicitly modeling these relational processes,
the result may resemble imagination while operating through fundamen-
tally different mechanisms.
As AI-assisted worldbuilding tools continue to shape creative practice, it
becomes necessary to ask not whether AI can generate worlds, but how
it does so—and whether its internal logic can meaningfully approach
the way humans think through and construct worlds. By examining the
underlying structures of contemporary AI world-generation systems, this
research positions the comparison between human and machine imag-
ination not as a question of output quality, but as a deeper inquiry into
process, structure, and the future training of imaginative systems.
Schedule and Research Plan
This project unfolds in three stages: understanding existing AI imagina-
tion systems, examining human imaginative processes, and exploring
how computational imagination might be improved by learning from
human worldbuilding logic. The nal outcome will be a written research
paper.
Week 1 – Week 3
Survey and test contemporary AI world- and scene-generation tools.
Analyze the underlying logic of each system and diagram their generative
pipelines.
Week 4 – Week 6
Interview and observe designers (set designers, worldbuilders) to under-
stand how humans approach imagination and world construction. Dia-
gram human imagination workows.
Week 7 – Week 9
Compare human and computational imagination procedures at the level
of structure and process. Begin drafting the comparative analysis section
of the paper.
Week 10 –12
Research theoretical and speculative approaches to improving AI imagi-
nation based on insights from human worldbuilding logic.
Week 13–15
Finalize the paper, integrating analysis, diagrams, and conclusions.
References
1. Lu, T., Shu, T., Xiao, J., Ye, L., Wang, J., Peng, C., … & Chen, J. (2024). GenEx: Generating an Explorable World. arXiv preprint.
https://arxiv.org/abs/2412.09624
2. Mo, S., Leng, Z., Liu, L., Wang, W., He, H., & Zhou, B. (2025). Dreamland: Controllable World Creation with Simulator and Generative Models. arXiv preprint.
https://arxiv.org/abs/2506.08006
3. Goslin, A. (2025). Terrain Diffusion: A Diffusion-Based Successor to Perlin Noise in Innite, Real-Time Terrain Generation. arXiv preprint.
https://arxiv.org/abs/2512.08309
4. Duan, Y., Zou, Z., Gu, T., Jia, W., Zhao, Z., Xu, L., … & Qiu, S. (2025). LatticeWorld: A Multimodal LLM-Empowered Framework for Interactive Complex World Generation. arXiv preprint.
https://arxiv.org/abs/2509.05263
5. Procedural Content Generation via Generative Articial Intelligence. (2024). arXiv survey paper on generative methods in content and environment creation.
https://arxiv.org/html/2407.09013v1
6. Alharthi, S. A. (2025). Generative AI in Game Design: Enhancing Creativity or … PMC article on generative AI for creative environments.
https://pmc.ncbi.nlm.nih.gov/articles/PMC12193870/
7. World and Human Action Models towards gameplay ideation (2025). Nature article introducing WHAM generative model.
https://www.nature.com/articles/s41586-025-08600-3
8. Davies, J. (2020). Articial Intelligence and Imagination. In The Cambridge Handbook of the Imagination (pp. 162-172). Cambridge University Press.
https://doi.org/10.1017/9781108580298.011
9. Waghmare, S. S. K. (2025). AI as an Artist: Exploring the Impact of Generative Algorithms on Creativity. PhilArchive preprint.
https://philarchive.org/archive/SAKAAA-6
10. Conceptual Blending (Theory of cognition relevant to imagination). Wikipedia article.
https://en.wikipedia.org/wiki/Conceptual_blending
11. Marble: A Multimodal World Model (World Labs ofcial blog).
https://www.worldlabs.ai/blog/marble-world-model
12. Dogra, S. (2025). New Marble AI Creates Entire 3D Worlds from Text. Analytics Vidhya blog.
https://www.analyticsvidhya.com/blog/2025/11/marble-world-ai-creates-3d-worlds-from-text/
13. Marble world model opens up new worlds for AI (Substack commentary on Marble & spatial intelligence).
https://patmcguinness.substack.com/p/marble-world-model-opens-up-new-worlds
14. Li, F.-F. (2025). Fei-Fei Li on AI World Models & Spatial Intelligence. Financial Times article (discussion of world models like Marble).
https://www.ft.com/content/d8fec7b5-f64a-4c5b-8439-6b8fe557be95
15. Zhang, Q. X. (2024). Creativity and Innovation at the Intersection of AI, Computer Graphics and Design (SIGGRAPH discussion).
https://s2025.siggraph.org/creativity-and-innovation-at-the-intersection-of-ai-computer-graph ics-and-design-2/
16. Bello, J. M. (2025). Generative AI Models and Creativity: Redening Human-Machine Collaboration. ResearchGate.
https://www.researchgate.net/publication/391874759_Generative_AI_Models_and_Creativity_Redening_Hu man-Machine_Collaboration_in_the_Creative_Process
17. Srivastava, I. (2024). A Comparative Analysis of Generative Models for Terrain Generation in Open-World Video Games. Scholastica PDF.
https://jhss.scholasticahq.com/api/v1/articles/92856-a-comparative-analysis-of-genera tive- models-for-terrain-generation-in-open-world-video-games.pdf
