pith. sign in

arxiv: 2601.15671 · v2 · submitted 2026-01-22 · 💻 cs.HC · cs.AI

StreetDesignAI: A Multi-Persona Evaluation System for Inclusive Infrastructure Design

Ziyi Wang , Yilong Dai , Duanya Lyu , Mateo Nader , Sihan Chen , Wanghao Ye , Zjian Ding , Xiang Yan This is my paper

Pith reviewed 2026-05-16 12:38 UTC · model grok-4.3

classification 💻 cs.HC cs.AI
keywords cycling infrastructurepersona-based evaluationmulti-perspective feedbackAI-assisted designinclusive infrastructuretransportation planninghuman-AI interactiondesign trade-offs
0
0 comments X p. Extension

The pith

Structured feedback from multiple AI-simulated cyclist personas helps transportation designers identify and address conflicting user needs more effectively than general-purpose chatbots.

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

The paper introduces StreetDesignAI, an interactive system that grounds design evaluation in real street imagery and map data while delivering parallel feedback from simulated personas ranging from confident to cautious cyclists. These personas are built from crowdsourced bikeability assessments. A within-subjects study with 26 professionals found that the system significantly improves designers' grasp of varied perspectives, their skill at spotting diverse needs, and their confidence in turning those needs into concrete decisions. Participants also showed higher satisfaction and stronger plans to adopt the tool professionally. The work frames explicit disagreement across perspectives as a core interaction primitive that shifts design from single-view optimization toward deliberate trade-off reasoning.

Core claim

StreetDesignAI lets designers ground evaluations in actual street contexts via imagery and maps, receive simultaneous feedback from AI-simulated personas that span confident to cautious cyclists, and iteratively revise designs while the system highlights conflicts between those perspectives. In a controlled comparison against a general-purpose AI chatbot, the multi-persona approach produced measurable gains in understanding diverse cyclist experiences, recognizing persona-specific requirements, and feeling confident about incorporating them into design choices, along with greater overall satisfaction and intent to use the system in practice.

What carries the argument

The multi-persona evaluation engine that generates parallel feedback from AI-simulated cyclist personas derived from crowdsourced assessments and surfaces explicit conflicts during iterative design changes.

If this is right

  • Design exploration shifts from optimizing for one viewpoint to explicit reasoning about trade-offs between user groups.
  • Designers report higher satisfaction with the process and stronger willingness to use the tool in professional work.
  • Qualitative evidence shows that surfacing conflicts makes competing needs visible early enough to influence decisions.
  • The approach supports more inclusive infrastructure by making experiential differences between user groups legible during iteration.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Similar persona-based conflict surfacing could be tested in related domains such as pedestrian path or public transit station design.
  • If the personas prove stable across cities, the system might serve as a low-cost way to audit existing streets for equity gaps before physical changes occur.
  • Over time, repeated use could train designers to anticipate conflicts without the tool, provided the underlying persona data remains updated.

Load-bearing premise

That the AI-simulated cyclist personas, built from crowdsourced assessments, accurately represent real experiential differences among diverse users in actual street environments.

What would settle it

A side-by-side comparison where real cyclists from the same demographic groups as the simulated personas ride the evaluated streets and rate the same design features; large mismatches between their ratings and the AI persona outputs would undermine the claim.

Figures

Figures reproduced from arXiv: 2601.15671 by Duanya Lyu, Mateo Nader, Sihan Chen, Wanghao Ye, Xiang Yan, Yilong Dai, Ziyi Wang, Zjian Ding.

Figure 1
Figure 1. Figure 1: Overview of StreetDesignAI. (A) Users input coordinates to load Street View imagery, which is analyzed using OpenStreetMap [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: System workflow of StreetDesignAI: The system consists of four main modules: (A) Evaluation Generation collects street-level [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Study workflow: participants completed five phases: (1) pre-study survey on design confidence; (2-3) two design tasks using [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Distribution of participant ratings for four key system functions (N=26). All functions rated above neutral midpoint. [PITH_FULL_IMAGE:figures/full_fig_p013_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Frequency of design parameter selections across 48 design scenarios: (a) lane width, (b) lane color, (c) buffer type, (d) buffer [PITH_FULL_IMAGE:figures/full_fig_p015_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Distribution of safety, comfort, and overall suitability scores across four cyclist personas (N=78 evaluations from 26 sessions). [PITH_FULL_IMAGE:figures/full_fig_p016_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Mean safety and comfort scores by persona and scenario type (evaluation vs. design). Color intensity indicates score magnitude [PITH_FULL_IMAGE:figures/full_fig_p017_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Distribution of overall suitability scores by design parameter choice and persona: (a) lane width, (b) lane color, (c) buffer type, [PITH_FULL_IMAGE:figures/full_fig_p020_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Survey Interface 1: Immersive 360-degree Google Street View for bikeability assessment. [PITH_FULL_IMAGE:figures/full_fig_p033_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Survey Interface 2: Rating for augmented image. [PITH_FULL_IMAGE:figures/full_fig_p033_10.png] view at source ↗
read the original abstract

Designing cycling infrastructure requires balancing the competing needs of diverse user groups, yet designers often struggle to anticipate how different cyclists experience the same street environment. We investigate how persona-based evaluation can support cycling infrastructure design by making experiential conflicts explicit during the design process. Informed by a formative study with 12 domain experts and crowdsourced bikeability assessments from 427 cyclists, we present StreetDesignAI, an interactive system that enables designers to (1) ground evaluation in real street context through imagery and map data, (2) receive parallel feedback from simulated cyclist personas spanning confident to cautious users, and (3) iteratively modify designs while the system surfaces conflicts across perspectives. A within-subjects study with 26 transportation professionals comparing StreetDesignAI against a general-purpose AI chatbot demonstrates that structured multi-perspective feedback significantly Broaden designers' understanding of various cyclists' perspectives, ability to identify diverse persona needs, and confidence in translating those needs into design decisions. Participants also reported significantly higher overall satisfaction and stronger intention to use the system in professional practice. Qualitative findings further illuminate how explicit conflict surfacing transforms design exploration from single-perspective optimization toward deliberate trade-off reasoning. We discuss implications for AI-assisted tools that scaffold persona-aware design through disagreement as an interaction primitive.

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 paper presents StreetDesignAI, an interactive system for cycling infrastructure design that generates parallel feedback from AI-simulated cyclist personas (spanning confident to cautious users) derived from crowdsourced bikeability ratings of 427 cyclists and a formative study with 12 experts. The system grounds evaluations in real street imagery and map data, surfaces perspective conflicts, and supports iterative design. A within-subjects study with 26 transportation professionals reports that the system significantly improves designers' understanding of diverse perspectives, ability to identify persona needs, confidence in design decisions, overall satisfaction, and intention to use compared to a general-purpose AI chatbot; qualitative results highlight a shift toward trade-off reasoning.

Significance. If the personas reliably proxy real experiential differences, the work offers a concrete advance in AI-supported inclusive design by treating explicit disagreement across user groups as an interaction primitive rather than a post-hoc concern. The within-subjects comparison against a baseline provides initial evidence that structured multi-perspective scaffolding can change professional design exploration, with potential implications for other domains requiring multi-stakeholder trade-offs.

major comments (2)
  1. [Persona construction] Persona construction (formative study and crowdsourced data section): The personas are built from 427 cyclists' bikeability ratings plus 12-expert input, yet the manuscript reports no direct validation such as a side-by-side comparison of persona outputs versus real cyclists evaluating identical street images; without this, measured gains in the user study could stem from interface structure alone rather than accurate representation of experiential differences.
  2. [Evaluation] User study analysis (evaluation section): The abstract states significant improvements in understanding, need identification, and design confidence, but provides no details on the statistical tests, effect sizes, controls for order effects, AI hallucination mitigation, or how persona feedback accuracy was assessed, leaving the central claim only partially supported.
minor comments (2)
  1. [Abstract] Abstract contains a capitalization error: 'significantly Broaden' should be 'significantly broadens'.
  2. [System description] The paper would benefit from explicit discussion of how conflicts across personas are quantified and surfaced in the interface.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback. We address each major comment below, providing our response and indicating the revisions we will incorporate to strengthen the manuscript.

read point-by-point responses

  1. Referee: [Persona construction] Persona construction (formative study and crowdsourced data section): The personas are built from 427 cyclists' bikeability ratings plus 12-expert input, yet the manuscript reports no direct validation such as a side-by-side comparison of persona outputs versus real cyclists evaluating identical street images; without this, measured gains in the user study could stem from interface structure alone rather than accurate representation of experiential differences.

    Authors: We acknowledge that the original manuscript did not include a direct side-by-side validation comparing persona-generated feedback against evaluations from real cyclists on identical street images. The personas were constructed from a substantial crowdsourced dataset of 427 cyclists' bikeability ratings combined with input from the 12-expert formative study, which provides an empirical basis for capturing experiential variation. However, to strengthen the claim that the observed benefits arise from accurate representation of differences rather than interface structure alone, we will add a dedicated validation subsection in the revised manuscript. This will report a new comparison using a held-out set of street images, including quantitative agreement metrics (e.g., Pearson correlations on bikeability scores) and qualitative alignment between persona outputs and real cyclist responses. revision: yes

  2. Referee: [Evaluation] User study analysis (evaluation section): The abstract states significant improvements in understanding, need identification, and design confidence, but provides no details on the statistical tests, effect sizes, controls for order effects, AI hallucination mitigation, or how persona feedback accuracy was assessed, leaving the central claim only partially supported.

    Authors: We agree that the evaluation section in the submitted manuscript lacked sufficient detail on the quantitative analysis. The within-subjects study employed paired t-tests for the Likert-scale measures, with effect sizes reported as Cohen's d; order effects were controlled via counterbalancing of the two conditions across participants. Hallucination mitigation relied on grounding all persona feedback in real street imagery and map data plus structured prompting; persona accuracy was cross-checked through expert review during the formative study. In the revision we will expand the Evaluation section with a new subsection explicitly detailing the statistical tests, effect sizes, counterbalancing procedure, hallucination controls, and accuracy assessment. We will also verify that the abstract claims are precisely supported by these expanded results. revision: yes

Circularity Check

0 steps flagged

No significant circularity; claims rest on independent user study and external crowdsourced data

full rationale

The paper describes a system whose personas are constructed from an external formative study (12 experts) and crowdsourced bikeability ratings (427 cyclists), then evaluates the system via a separate within-subjects study with 26 transportation professionals. No equations, fitted parameters, or derivations appear. The central claims about broadened understanding and design confidence are measured directly from participant responses in the user study rather than reducing to self-referential definitions or self-citation chains. The evaluation is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim depends on the assumption that AI can faithfully simulate diverse cyclist experiences from limited crowdsourced data; no free parameters are explicitly fitted in the abstract, but the persona model itself functions as an unvalidated constructed entity.

axioms (1)
  • domain assumption Simulated personas derived from crowdsourced bikeability assessments can stand in for real cyclists' experiential perspectives
    Invoked in the system description and evaluation claims without independent validation metrics provided in the abstract.
invented entities (1)
  • StreetDesignAI multi-persona evaluation system no independent evidence
    purpose: To surface conflicts across cyclist perspectives during design iteration
    New system introduced in the paper; independent evidence would require external validation of persona fidelity.

pith-pipeline@v0.9.0 · 5541 in / 1325 out tokens · 51374 ms · 2026-05-16T12:38:11.214719+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

74 extracted references · 74 canonical work pages · 1 internal anchor

  1. [1]

    Rachel Aldred. 2017. Cycling policy in the UK: A historical and thematic overview.Cycling Futures(2017), 23–44

    work page 2017

  2. [2]

    Lisa P Argyle, Ethan C Busby, Nancy Fulda, Joshua R Gubler, Christopher Rytting, and David Wingate. 2023. Out of one, many: Using language models to simulate human samples.Political Analysis31, 3 (2023), 337–351

    work page 2023

  3. [3]

    Ciro Beneduce, Bruno Lepri, and Massimiliano Luca. 2025. Urban Safety Perception Through the Lens of Large Multimodal Models: A Persona-based Approach. arXiv:2503.00610 [cs.CY]

    work page arXiv 2025

  4. [4]

    Hugh Beyer and Karen Holtzblatt. 1999. Contextual design.interactions6, 1 (1999), 32–42

    work page 1999

  5. [5]

    Karen Bickerstaff, Rodney Tolley, and Gordon Walker. 2002. Transport planning and participation: the rhetoric and realities of public involvement. Journal of Transport Geography10, 1 (2002), 61–73

    work page 2002

  6. [6]

    Virginia Braun and Victoria Clarke. 2006. Using thematic analysis in psychology.Qualitative Research in Psychology3, 2 (Jan. 2006), 77–101. https://doi.org/10.1191/1478088706qp063oa

    work page doi:10.1191/1478088706qp063oa 2006

  7. [7]

    Tim Brown. 2008. Design thinking.Harvard Business Review86, 6 (2008), 84–92

    work page 2008

  8. [8]

    Jiangjie Chen, Xintao Wang, Rui Xu, Siyu Yuan, Yikai Zhang, Wei Shi, Jian Xie, Shuang Li, Ruihan Yang, Tinghui Zhu, Aili Chen, Nianqi Li, Lida Chen, Caiyu Hu, Siye Wu, Scott Ren, Ziquan Fu, and Yanghua Xiao. 2024. From Persona to Personalization: A Survey on Role-Playing Language Agents.Transactions on Machine Learning Research(2024). https://openreview.n...

    work page 2024

  9. [10]

    Yilong Dai, Luyu Liu, Kaiyue Wang, Meiqing Li, and Xiang Yan. 2025. Using computer vision and street view images to assess bus stop amenities. Computers, Environment and Urban Systems117 (2025), 102254. Manuscript submitted to ACM StreetDesignAI 25

    work page 2025

  10. [11]

    Yilong Dai, Ziyi Wang, Chenguang Wang, Kexin Zhou, Yiheng Qian, Susu Xu, and Xiang Yan. 2026. Persona-aware and Explainable Bikeability Assessment: A Vision-Language Model Approach.arXiv preprint arXiv:2601.03534(2026)

    work page arXiv 2026

  11. [12]

    Boyang Deng, Richard Tucker, Zhengqi Li, et al . 2024. Streetscapes: Large-scale consistent street view generation using autoregressive video diffusion. InACM SIGGRAPH 2024 Conference Papers. 1–11

    work page 2024

  12. [13]

    Jennifer Dill and Nathan McNeil. 2013. Four Types of Cyclists?: Examination of Typology for Better Understanding of Bicycling Behavior and Potential.Transportation Research Record2387, 1 (2013), 129–138. https://doi.org/10.3141/2387-15

    work page doi:10.3141/2387-15 2013

  13. [14]

    Jennifer Dill and Nathan McNeil. 2016. Revisiting the Four Types of Cyclists: Findings from a National Survey.Transportation Research Record2587, 1 (2016), 90–99. https://doi.org/10.3141/2587-11

    work page doi:10.3141/2587-11 2016

  14. [15]

    Ferenchak and Wesley E

    Nicholas N. Ferenchak and Wesley E. Marshall. 2025. The Link between Low-Stress Bicycle Infrastructure and Bicycle Commuting.Journal of Transport Geography118 (2025), 104098

    work page 2025

  15. [16]

    1957.A Theory of Cognitive Dissonance

    Leon Festinger. 1957.A Theory of Cognitive Dissonance. Stanford University Press

    work page 1957

  16. [17]

    Fowler, Bradley L

    Sara L. Fowler, Bradley L. Beall, and Danielle W. Derr. 2017. Perceptions of cycling among potential cyclists.Transportation Research Part F: Traffic Psychology and Behaviour49 (2017), 474–486

    work page 2017

  17. [18]

    Froehlich, Alexander J

    Jon E. Froehlich, Alexander J. Fiannaca, Nimer M Jaber, Victor Tsaran, and Shaun K. Kane. 2025. StreetViewAI: Making Street View Accessible Using Context-Aware Multimodal AI. InProceedings of the 38th Annual ACM Symposium on User Interface Software and Technology (UIST ’25). Association for Computing Machinery, New York, NY, USA, Article 43, 22 pages. htt...

    work page doi:10.1145/3746059.3747756 2025

  18. [19]

    Peter G Furth, Maaza C Mekuria, and Hilary Nixon. 2013. Network connectivity for low-stress bicycling.Transportation Research Record2387, 1 (2013), 144–154

    work page 2013

  19. [20]

    2006.Four Types of Cyclists

    Roger Geller. 2006.Four Types of Cyclists. Technical Report. Portland Bureau of Transportation, Portland, OR

    work page 2006

  20. [21]

    Google. 2025. Street View — Maps JavaScript API. https://developers.google.com/maps/documentation/javascript/streetview. Accessed 2025-01-03

    work page 2025

  21. [22]

    Eddie Harmon-Jones and Judson Mills. 1999. An introduction to cognitive dissonance theory and an overview of current perspectives on the theory. (01 1999). https://doi.org/10.1037/10318-001

    work page doi:10.1037/10318-001 1999

  22. [23]

    Mingyi He, Yuebing Liang, Shenhao Wang, et al. 2025. Generative AI for urban design: A stepwise approach integrating human expertise with multimodal diffusion models.arXiv preprint arXiv:2505.24260(2025)

    work page arXiv 2025

  23. [24]

    Yu-Kai Hung, Yun-Chien Huang, Ting-Yu Su, Yen-Ting Lin, Lung-Pan Cheng, Bryan Wang, and Shao-Hua Sun. 2025. SimTube: Simulating Audience Feedback on Videos using Generative AI and User Personas. InProceedings of the 30th International Conference on Intelligent User Interfaces (IUI ’25). Association for Computing Machinery, New York, NY, USA, 1256–1271. ht...

    work page doi:10.1145/3708359.3712146 2025

  24. [25]

    Samuel Jang et al. 2024. PersonaGym: Evaluating Persona Agents and LLMs.arXiv preprint arXiv:2407.18416(2024)

    work page arXiv 2024

  25. [26]

    Azem, João M

    Ilkka Kaate, Joni Salminen, Soon-Gyo Jung, Trang Thi Thu Xuan, Jinan Y. Azem, João M. Santos, and Bernard J Jansen. 2025. When Personas Talk to You: Evaluating the Evolution of User Personas from Static Profiles to Conversational User Interfaces. InProceedings of the 2025 ACM Designing Interactive Systems Conference (DIS ’25). ACM, 2350–2372. https://doi....

    work page doi:10.1145/3715336.3735676 2025

  26. [27]

    Janin Koch, Andrés Lucero, Lena Hegemann, and Antti Oulasvirta. 2019. May AI? Design ideation with cooperative contextual bandits. InProceedings of the 2019 CHI Conference on Human Factors in Computing Systems. ACM, 1–12

    work page 2019

  27. [28]

    Emma R Lawlor, Kate Ellis, Jean Adams, et al . 2023. Stakeholders’ experiences of what works in planning and implementing environmental interventions to promote active travel: a systematic review and qualitative synthesis.Transport Reviews43, 3 (2023), 478–501

    work page 2023

  28. [29]

    Chu Li, Zhihan Zhang, Michael Saugstad, Esteban Safranchik, Chaitanyashareef Kulkarni, Xiaoyu Huang, Shwetak Patel, Vikram Iyer, Tim Althoff, and Jon E Froehlich. 2024. LabelAId: Just-in-time AI interventions for improving human labeling quality and domain knowledge in crowdsourcing systems. InProceedings of the 2024 CHI Conference on Human Factors in Com...

    work page 2024

  29. [30]

    Yongming Li, Hangyue Zhang, Andrea Yaoyun Cui, Zisong Ma, Yunpeng Song, Zhongmin Cai, and Yun Huang. 2025. EyeSee: Enhancing Art Appreciation through Anthropomorphic Interpretations from Multiple Perspectives. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems (CHI ’25). ACM, Article 660. https://doi.org/10.1145/3706598.3714042

    work page doi:10.1145/3706598.3714042 2025

  30. [31]

    Ke Liu, Tan Yigitcanlar, Will Browne, and Yanjie Fu. 2025. Prompts for planning-AI integration: LLM prompt design for supporting sustainable urban development.Journal of Open Innovation: Technology, Market, and Complexity11 (2025), 100666

    work page 2025

  31. [32]

    Yiren Liu, Pranav Sharma, Mehul Oswal, Haijun Xia, and Yun Huang. 2025. PersonaFlow: Designing LLM-Simulated Expert Perspectives for Enhanced Research Ideation. InProceedings of the 2025 ACM Designing Interactive Systems Conference (DIS ’25). ACM, 506–534. https://doi.org/10. 1145/3715336.3735789

    work page arXiv 2025

  32. [33]

    Yuxuan Lu, Bingsheng Yao, Hansu Gu, Jing Huang, Zheshen Jessie Wang, Yang Li, Jiri Gesi, Qi He, Toby Jia-Jun Li, and Dakuo Wang. 2025. UXAgent: An LLM Agent-Based Usability Testing Framework for Web Design. InProceedings of the Extended Abstracts of the CHI Conference on Human Factors in Computing Systems (CHI EA ’25). ACM, 1–12. https://doi.org/10.1145/3...

    work page doi:10.1145/3706599.3719729 2025

  33. [34]

    Wo Meijer, Tilman Dingler, and Gerd Kortuem. 2025. D360: a Tool for Supporting Rapid, Iterative, and Collaborative Analysis of 360 Video. In Proceedings of the 2025 ACM Designing Interactive Systems Conference (DIS ’25). ACM, 1615–1627. https://doi.org/10.1145/3715336.3735793

    work page doi:10.1145/3715336.3735793 2025

  34. [35]

    2012.Low-Stress Bicycling and Network Connectivity

    Maaza C Mekuria, Peter G Furth, and Hilary Nixon. 2012.Low-Stress Bicycling and Network Connectivity. Technical Report CA-MTI-12-1005. Mineta Transportation Institute

    work page 2012

  35. [36]

    Maaza C Mekuria, Peter G Furth, and Hilary Nixon. 2012. Low-stress bicycling and network connectivity. (2012)

    work page 2012

  36. [37]

    National Association of City Transportation Officials. 2011. Urban Bikeway Design Guide

    work page 2011

  37. [38]

    2014.Urban Bikeway Design Guide(2nd ed.)

    National Association of City Transportation Officials. 2014.Urban Bikeway Design Guide(2nd ed.). Island Press, Washington, DC. Manuscript submitted to ACM 26 Wang, et al

    work page 2014

  38. [39]

    Task Force on Geometric Design

    Transportation Officials. Task Force on Geometric Design. 1999.Guide for the development of bicycle facilities. American Association of State Highway & Transportation Officials

    work page 1999

  39. [40]

    OpenAI. 2025. GPT-4.1. https://openai.com/index/gpt-4-1/. Accessed 2025-01-03

    work page 2025

  40. [41]

    OpenAI. 2025. Introducing 4o image generation. https://openai.com/index/introducing-4o-image-generation/

    work page 2025

  41. [42]

    Overpass API Development Team. 2025. Overpass API Documentation: Preface. https://dev.overpass-api.de/overpass-doc/en/preface/preface.html. Accessed 2025-01-03

    work page 2025

  42. [43]

    Joon Sung Park, Joseph C O’Brien, Carrie J Cai, Meredith Ringel Morris, Percy Liang, and Michael S Bernstein. 2023. Generative agents: Interactive simulacra of human behavior. InProceedings of the 36th Annual ACM Symposium on User Interface Software and Technology. 1–22

    work page 2023

  43. [44]

    LLM Agents Grounded in Self-Reports Enable General-Purpose Simulation of Individuals

    Joon Sung Park, Carolyn Q. Zou, Aaron Shaw, Benjamin Mako Hill, Carrie Cai, Meredith Ringel Morris, Robb Willer, Percy Liang, and Michael S. Bernstein. 2024. Generative Agent Simulations of 1,000 People. arXiv:2411.10109 [cs.AI] https://arxiv.org/abs/2411.10109

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  44. [45]

    John Pucher and Ralph Buehler. 2010. Walking and cycling for healthy cities.Built Environment36, 4 (2010), 391–414

    work page 2010

  45. [46]

    John Pucher and Ralph Buehler. 2010. Walking and Cycling in Western Europe and the United States: Trends, Policies, and Lessons.TR News280 (2010), 34–42

    work page 2010

  46. [47]

    Steven Jige Quan, James Park, Athanassios Economou, and Sugie Lee. 2019. Artificial intelligence-aided design: Smart design for sustainable city development.Environment and Planning B: Urban Analytics and City Science46, 8 (2019), 1581–1599

    work page 2019

  47. [48]

    Manaswi Saha, Michael Saugstad, Hanuma Teja Maddali, Aileen Zeng, Ryan Holland, Steven Bower, Aditya Dash, Sage Chen, Anthony Li, Kotaro Hara, et al. 2019. Project sidewalk: A web-based crowdsourcing tool for collecting sidewalk accessibility data at scale. InProceedings of the 2019 CHI Conference on Human Factors in Computing Systems. 1–14

    work page 2019

  48. [49]

    Jinming Su, Songen Gu, Yiting Duan, Xingyue Chen, and Junfeng Luo. 2024. Text2Street: Controllable Text-to-image Generation for Street Views. CoRRabs/2402.04504 (2024). https://doi.org/10.48550/arXiv.2402.04504

    work page doi:10.48550/arxiv.2402.04504 2024

  49. [50]

    Xinyu Tan, Qiwei Song, Xun Liu, and Waishan Qiu. 2025. Visual Perception-Informed Urban Design Toolkit: Computational Urban Morphology Optimisation to Inform Real-Time Perceived Safety.Journal of Urban Management(2025). https://doi.org/10.1016/j.jum.2025.09.005

    work page doi:10.1016/j.jum.2025.09.005 2025

  50. [51]

    Mathias Peter Verheijden and Mathias Funk. 2023. Collaborative Diffusion: Boosting Designerly Co-Creation with Generative AI. InExtended Abstracts of the 2023 CHI Conference on Human Factors in Computing Systems (CHI EA ’23). ACM, Article 73. https://doi.org/10.1145/3544549.3585680

    work page doi:10.1145/3544549.3585680 2023

  51. [52]

    Chenguang Wang, Xiang Yan, Yilong Dai, Ziyi Wang, and Susu Xu. 2025. From Image Generation to Infrastructure Design: a Multi-agent Pipeline for Street Design Generation. arXiv:2509.05469 [cs.AI] https://arxiv.org/abs/2509.05469

    work page arXiv 2025

  52. [53]

    Qingyi Wang, Yuebing Liang, Yunhan Zheng, Kaiyuan Xu, Jinhua Zhao, and Shenhao Wang. 2025. Generative AI for Urban Planning: Synthesizing Satellite Imagery via Diffusion Models.arXiv preprint arXiv:2505.08833(2025)

    work page arXiv 2025

  53. [54]

    Ziyi Wang, Ziwen Zeng, Yuan Li, and Zijian Ding. 2025. CareerPooler: AI-Powered Metaphorical Pool Simulation Improves Experience and Outcomes in Career Exploration. arXiv:2509.11461 [cs.HC]

    work page arXiv 2025

  54. [55]

    Jason Wu, Kashyap Todi, Joannes Chan, Brad A Myers, and Ben Lafreniere. 2024. FrameKit: A Tool for Authoring Adaptive UIs Using Keyframes. In Proceedings of the 29th International Conference on Intelligent User Interfaces (IUI ’24). ACM, 660–674. https://doi.org/10.1145/3640543.3645176

    work page doi:10.1145/3640543.3645176 2024

  55. [56]

    Xiao, Richard Patterson, David Ogilvie, Esther M.F

    Christina S. Xiao, Richard Patterson, David Ogilvie, Esther M.F. van Sluijs, Stephen J. Sharp, and Jenna Panter. 2023. Design effects of cycle infrastructure changes: An exploratory analysis of cycle levels.Transportation Research Interdisciplinary Perspectives22 (2023), 100949. https: //doi.org/10.1016/j.trip.2023.100949

    work page doi:10.1016/j.trip.2023.100949 2023

  56. [57]

    Hannah Younes and Yonah Freemark. 2024. Cycling infrastructure and road safety: A meta-analysis.Journal of Safety Research88 (2024), 100071

    work page 2024

  57. [58]

    Rzeszotarski

    Chao Zhang, Kexin Ju, Zhuolun Han, Yu-Chun Grace Yen, and Jeffrey M. Rzeszotarski. 2025. Synthia: Visually Interpreting and Synthesizing Feedback for Writing Revision. InProceedings of the 38th Annual ACM Symposium on User Interface Software and Technology (UIST ’25). Association for Computing Machinery, New York, NY, USA, Article 88, 16 pages. https://do...

    work page doi:10.1145/3746059.3747703 2025

  58. [59]

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. 2023. Adding conditional control to text-to-image diffusion models. InProceedings of the IEEE/CVF International Conference on Computer Vision. 3836–3847

    work page 2023

  59. [60]

    ${userMessage}

    Zhilun Zhou, Yuming Lin, Depeng Jin, and Yong Li. 2024. Large Language Model for Participatory Urban Planning. arXiv:2402.17161 [cs.AI] https://arxiv.org/abs/2402.17161 Manuscript submitted to ACM StreetDesignAI 27 9 APPENDIX 9.1 System Prompts Persona Agent – Single-Design Deep Analysis You are an independent evaluation agent representing the following p...

    work page arXiv 2024

  60. [61]

    Analyze visual differences in the images relevant to your persona’s priorities

    work page

  61. [62]

    Score each design option from 0.0 to 1.0

    work page

  62. [63]

    Select a preferred design and explain trade-offs from your persona’s perspective

    work page

  63. [64]

    persona":

    List persona-specific deal-breakers. Respond with ONLY valid JSON: { "persona": "${personaName}", "scores": [ { "design_id": "<id>", "score": <0.0-1.0>, "rationale": "<1-2 sentences>" } ], "preferred_design": "<id>", "deal_breakers": ["<list>"] } Manuscript submitted to ACM 28 Wang, et al. Strong & Fearless Persona Agent Evaluation You are a Strong & Fear...

    work page

  64. [66]

    Ensure these white boundary lines strictly contain and distinctly outline the bike lane area

    Right boundary: a prominent, continuous solid white line. Ensure these white boundary lines strictly contain and distinctly outline the bike lane area. [If bufferType === ’standard’ && bufferLocation === ’moving-cars’:]

    work page

  65. [67]

    Do not apply any green paint within this buffer zone

    Left Boundary: A buffer zone adjacent to the bike lane on its left side, clearly marked with prominent diagonal white stripes, bounded on both sides by solid white lines. Do not apply any green paint within this buffer zone

    work page

  66. [68]

    [If bufferType === ’narrow-bollards’ && bufferLocation === ’moving-cars’:]

    Right Boundary: A prominent, continuous solid white line marking the right-hand edge of the bike lane. [If bufferType === ’narrow-bollards’ && bufferLocation === ’moving-cars’:]

    work page

  67. [69]

    This buffer zone should: - Be bounded on both sides by solid white lines

    Left Boundary: A narrow buffer zone adjacent to the bike lane on its left side. This buffer zone should: - Be bounded on both sides by solid white lines. - Be filled with prominent diagonal white stripes. - Include vertical red-and-white striped bollards placed at regular intervals, explicitly positioned in the center of the buffer zone. - Do not apply an...

    work page

  68. [70]

    [If bufferType === ’narrow-armadillo’ && bufferLocation === ’moving-cars’:]

    Right Boundary: A prominent, continuous solid white line. [If bufferType === ’narrow-armadillo’ && bufferLocation === ’moving-cars’:]

    work page

  69. [71]

    This buffer zone should: - Be bounded on both sides by solid white lines

    Left Boundary: A narrow buffer zone adjacent to the bike lane on its left side. This buffer zone should: - Be bounded on both sides by solid white lines. - Be filled with prominent diagonal white stripes. - Include rounded, semi-flexible rubber lane dividers (often called ‘armadillos’), evenly spaced along the center of the buffer zone. The dividers shoul...

    work page

  70. [72]

    [If bufferType === ’standard’ && bufferLocation === ’parked-cars’:]

    Right Boundary: A prominent, continuous solid white line. [If bufferType === ’standard’ && bufferLocation === ’parked-cars’:]

    work page

  71. [74]

    [If bufferType === ’narrow-bollards’ && bufferLocation === ’parked-cars’:]

    Right boundary: A clearly marked buffer zone adjacent to the bike lane, filled with prominent diagonal white stripes, and bounded on both sides by solid white lines. [If bufferType === ’narrow-bollards’ && bufferLocation === ’parked-cars’:]

    work page

  72. [76]

    This buffer zone should: - Be bounded on both sides by solid white lines

    Right boundary: A clearly marked narrow buffer zone immediately adjacent to the bike lane. This buffer zone should: - Be bounded on both sides by solid white lines. - Be filled with prominent diagonal white stripes. - Distinctly feature vertical red-and-white striped bollards placed at regular intervals. [If bufferType === ’narrow-armadillo’ && bufferLoca...

    work page

  73. [77]

    Left boundary: a prominent, continuous solid white line

    work page

  74. [78]

    This buffer zone should: - Be bounded on both sides by solid white lines

    Right boundary: narrow buffer zone adjacent to the bike lane. This buffer zone should: - Be bounded on both sides by solid white lines. - Be filled with prominent diagonal white stripes. - Within this buffer zone, clearly place individual black-and-white striped armadillo lane dividers, positioned as separate, regularly spaced units. Ensure the updated bi...

    work page