AI-Driven Deconstruction of Urban Regulatory Frameworks: Unveiling Social Sustainability Gaps in Santiago’s Communal Zoning

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The article 'AI-Driven Deconstruction of Urban Regulatory Frameworks: Unveiling Social Sustainability Gaps in Santiago’s Communal Zoning' represents pragmatic approach of review of a large body of planning regulatory documents and their comparison and evaluation with the help of AI functionalities.

The work is original and well informative, showcasing the extended possibilities of the artificial intelligence related tools, techniques and methods. The article is clearly structured and readable.

The methodological part has to be strengthened via the inclusion of more detailed explanation of what AI procedures, specific adjustments and choices were involved as that part is vague an not replicable. Supplementary files can support this improved transparency of the research design and smaller steps in between the bigger ones described. The style of writing is too much AI one and there should be restructuring of the bullet points into paragraphs. Additional statement about the support of AI tools in the writing should be made more explicit as it seems that not only the analysis but also the writing of the manuscript it supported by such.

More specific comments and questions can be addressed to the following parts and lines of the text:

• Line 261 - What about the disparities due to different development, evolution and legacy patterns, pressure and state, not only to the regulatory content, some more context before jumping in the results or in parallel to them is needed, at least with some examples and references to data, indicators or described phenomena, otherwise the text is too abstract;
• Lines 315-316 - in order to relate to the de facto model when addressing the de jure planned urban development model it is advisable to include one or more of the following maps which would allow more context and easier navigation and comparison among the results:

- base map 
- population density map
- other thematic maps visualizing the median zonal/municipal statistics (plus min and max) of or the more detailed status inside the communities in terms of ratio between primary roads / green areas / number or gross floor area of kindergardens/primary schools and the number of residents

• Line 329 - How the blind spots relate to national legislation and other political requirements, is it metropolitan or municipal gap, how necessary it is to include such content in these ordinances given their role in the planning system?
• Lines 362-371 - Do the planning system in Chile addresses these issues in municipal plans or for instance accessibility is a horizontal policy which should be addressed at the investment design level and climate extremes are issues addressed in other plans considering catchment areas, the city as a whole, etc.
• Lines 377-381 - The operational issues of implementation are important part of the planning cycle but how explicit the everyday communication with the residents should be in an ordinance?
• Lines 382-385 - This is being repeated from above and there are also other repeated observations. The final text should avoid such duplications
• Lines 386-392 - This information with part of the text following it should find its place in the methodological part
• Line 458 - The figure is over the text at the top
• Line 528 - Maybe the foundations are not for a form of automated urban planning but more of automated legislative impact assessment and planning support system which can be adjusted to provide monitoring given the transparency of the approved ordinances and other relevant content

 

Author Response

Reviewer 1

Observation 1: The methodological part has to be strengthened via the inclusion of more detailed explanation of what AI procedures, specific adjustments and choices were involved as that part is vague an not replicable. Supplementary files can support this improved transparency of the research design and smaller steps in between the bigger ones described.

Response 1 Thank you for this observation. Changes have been made to the methodology section to clearly differentiate between data collection and the AI pipeline for analytics. This allows for improved replication of modeling.

Observation 2: The style of writing is too much AI one and there should be restructuring of the bullet points into paragraphs. Additional statement about the support of AI tools in the writing should be made more explicit as it seems that not only the analysis but also the writing of the manuscript it supported by such.

Response 2: It is important to mention that the text of the article was not generated by AI but in the original language of the authors, which after translation was reviewed to proofread potential errors in the translation. This is added to the methodology and those parts that "sound" more like an AI are changed.

Observation 3: Line 261 - What about the disparities due to different development, evolution and legacy patterns, pressure and state, not only to the regulatory content, some more context before jumping in the results or in parallel to them is needed, at least with some examples and references to data, indicators or described phenomena, otherwise the text is too abstract;

Response 3: Thank you for this comment. Although the focus of this article is on the operability of the AI tool for the analysis of urban regulatory models, it is understood that it seeks to generate a useful context for someone who comes to review this article trying to better understand the case of Santiago. We add a text to present the case in relation to its urban evolution.

Observation 4: Lines 315-316 - in order to relate to the de facto model when addressing the de jure planned urban development model it is advisable to include one or more of the following maps which would allow more context and easier navigation and comparison among the results:

- base map 
- population density map
- other thematic maps visualizing the median zonal/municipal statistics (plus min and max) of or the more detailed status inside the communities in terms of ratio between primary roads / green areas / number or gross floor area of kindergardens/primary schools and the number of residents

Response 4: A set of cartographies are added to advance in this observed problem.

Observation 5: Line 329 - How the blind spots relate to national legislation and other political requirements, is it metropolitan or municipal gap, how necessary it is to include such content in these ordinances given their role in the planning system?

 

Response 5: A deepening of this aspect has been added.

 

Observation 6: Lines 362-371 - Do the planning system in Chile addresses these issues in municipal plans or for instance accessibility is a horizontal policy which should be addressed at the investment design level and climate extremes are issues addressed in other plans considering catchment areas, the city as a whole, etc.

 

Response 6: This observation is appreciated. Details on the rationality of the urban model in Chile are incorporated, especially with regard to urban planning regulations.

 

Observation 7: Lines 377-381 - The operational issues of implementation are important part of the planning cycle but how explicit the everyday communication with the residents should be in an ordinance?

 

Response 7: It is understood that the doubt arises from not having explained more clearly how urban governance and its regulations are carried out in Chile. Details are added to improve the presentation of the case.

 

Observation 8: Lines 382-385 - This is being repeated from above and there are also other repeated observations. The final text should avoid such duplications

 

Response 8: Thanks for the observation. A review is performed to eliminate redundancies.

 

Observation 9:Lines 386-392 - This information with part of the text following it should find its place in the methodological part

 

Response 9:  The indicated information is relocated.

 

Observation 10: Line 458 - The figure is over the text at the top

 

Response 10: Thank you for the observation and the modification is made.

 

Observation 11: Line 528 - Maybe the foundations are not for a form of automated urban planning but more of automated legislative impact assessment and planning support system which can be adjusted to provide monitoring given the transparency of the approved ordinances and other relevant content.

 

Response 11: Thanks for the observation. This sentence has been restated to avoid interpretative problems.

 

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors The manuscript titled "AI-Driven Deconstruction of Urban Regulatory Frameworks: Unveiling Social Sustainability Gaps in Santiago’s Communal Zoning" presents an ambitious and innovative approach to examining urban governance through AI-enabled tools. The authors contribute to critical debates on spatial justice, institutional fragmentation, and regulatory opacity by showcasing how artificial intelligence can transcend the traditional limits of manual zoning audits. However, to further strengthen the manuscript and ensure its full academic rigor, I would like to offer the following suggestions:   1. Clarify Research Focus and Gap in the Introduction: While the introduction presents an interesting narrative, it currently lacks a clear articulation of the specific research gap this study aims to address. Given the growing body of literature on auditing urban regulatory frameworks and zoning analysis, the authors should position their work more explicitly within this scholarly context. I also recommend expanding the introduction to include a dedicated sub-section or a standalone Research Gap and Aim statement, which would clarify the contribution of the AI-driven approach in contrast to existing manual or traditional methodologies. 2. Strengthen the Methodological Framework: The manuscript would benefit greatly from a clearer and more systematic description of the data sources and analytical methods. The second section should provide a comprehensive overview of the case study area (Santiago), the nature of the zoning data used, and the process by which AI techniques were applied. If algorithmic models or formula-based interpretations were used, these should be presented clearly (e.g., through equations or flowcharts), which would help readers assess the robustness and reproducibility of the analysis. 3. Enhance the Discussion Section: The Discussion section should be expanded to directly engage with the research questions or hypotheses outlined in the introduction. This would improve the logical coherence of the paper and demonstrate the practical implications of the findings. Moreover, the discussion would benefit from deeper engagement with existing literature on urban spatial justice, digital governance, and zoning policy reforms. Doing so would not only situate the study within broader academic conversations but also highlight its unique contributions.

Author Response

Reviewer 2

 

Observation 1: Clarify Research Focus and Gap in the Introduction: While the introduction presents an interesting narrative, it currently lacks a clear articulation of the specific research gap this study aims to address. Given the growing body of literature on auditing urban regulatory frameworks and zoning analysis, the authors should position their work more explicitly within this scholarly context. I also recommend expanding the introduction to include a dedicated sub-section or a standalone Research Gap and Aim statement, which would clarify the contribution of the AI-driven approach in contrast to existing manual or traditional methodologies. 

 

Response 1: Thanks for the feedback. An addendum has been made in the introduction to make explicit the research gap and the aim of the study that allows us to advance in understanding the usefulness of this type of research.

 

Observation 2: Strengthen the Methodological Framework: The manuscript would benefit greatly from a clearer and more systematic description of the data sources and analytical methods. The second section should provide a comprehensive overview of the case study area (Santiago), the nature of the zoning data used, and the process by which AI techniques were applied. If algorithmic models or formula-based interpretations were used, these should be presented clearly (e.g., through equations or flowcharts), which would help readers assess the robustness and reproducibility of the analysis. 

 

Response 2: Observation is understood. A workflow diagram is made to help clarify the rationality of the process and the list of each enacted ordinance used is shared in Appendix A.

 

Observation 3: Enhance the Discussion Section: The Discussion section should be expanded to directly engage with the research questions or hypotheses outlined in the introduction. This would improve the logical coherence of the paper and demonstrate the practical implications of the findings. Moreover, the discussion would benefit from deeper engagement with existing literature on urban spatial justice, digital governance, and zoning policy reforms. Doing so would not only situate the study within broader academic conversations but also highlight its unique contributions.

 

Response 3: Addenda have been made to the discussion to generate a more direct link to the literature reviewed.

 

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This manuscript introduces an innovative methodology for auditing urban regulatory frameworks using artificial intelligence (AI) and natural language processing (NLP), with the metropolitan area of Santiago, Chile, as a case study. The study conducts a semantic analysis of 31 Communal Zoning Ordinances (PRC) to reveal how legal structures actively exacerbate socio-spatial inequalities despite appearing normatively neutral. By employing the DeepSeek-R1 model, fine-tuned for Chilean legal-urban discourse, the methodology achieves 88% classification accuracy, enabling the detection of normative asymmetries, omissions, and structural fragmentation. Key findings indicate that affluent communes (e.g., Vitacura and Las Condes) exhibit detailed and incentive-rich regulations. At the same time, peripheral municipalities lack provisions for social housing, participatory mechanisms, or climate resilience, thereby reinforcing exclusionary patterns. The paper also proposes a scalable rubric-based evaluation system and GIS visualizations to synthesize regulatory disparities across the metropolitan area. Methodologically, the study demonstrates how domain-adapted AI can extend regulatory scrutiny beyond manual limitations, while substantively contributing to discussions on spatial justice, institutional fragmentation, and regulatory opacity in urban planning. The results call for binding mechanisms to align local zoning with metropolitan equity goals and highlight the potential of automated audits to inform reform agendas in the Global South.

The following issues and suggestions are provided for improvement:

Introduction Section: The introduction does not identify gaps in the existing literature. Please elaborate on the limitations of current research identified by the authors and explain how this manuscript addresses these gaps.

Materials and Methods Section: It is recommended to include a research framework or technical roadmap diagram in Section 2 to provide readers with a more intuitive understanding of the study.

Figure Formatting: Figure captions should be placed below the figures. Additionally, the title of Figure 2 is not visible; please optimize the formatting.

Map Elements: It is suggested to include a north arrow in each spatial analysis map to enhance map completeness.

AI and NLP Terminology: The manuscript repeatedly presents AI and NLP as parallel concepts; however, AI broadly encompasses NLP. The authors are advised to refine the description of AI methods and consider adding statistical analysis techniques.

In summary, while the study makes a valuable contribution, improvements are needed in manuscript formatting, clarity of innovation, and methodological rigor. Specifically, incorporating a research framework figure and detailing statistical methods would strengthen the presentation of the study’s novelty and professionalism.

Author Response

Reviewer 3

Observation 1: Introduction Section: The introduction does not identify gaps in the existing literature. Please elaborate on the limitations of current research identified by the authors and explain how this manuscript addresses these gaps.

Response 1: Thanks for the observation. Improvements have been made to the introduction and in the methodology section, which has been considerably expanded, the limitations of the study are also added, to review how these limitations were faced for the research.

Observation 2: Materials and Methods Section: It is recommended to include a research framework or technical roadmap diagram in Section 2 to provide readers with a more intuitive understanding of the study.

Response 2: A workflow diagram is added.

Observation 3: Figure Formatting: Figure captions should be placed below the figures. Additionally, the title of Figure 2 is not visible; please optimize the formatting.

Response 3: The necessary format changes are made.

Observation 4: Map Elements: It is suggested to include a north arrow in each spatial analysis map to enhance map completeness.

Response 4: North is added to context maps.

Observation 5: AI and NLP Terminology: The manuscript repeatedly presents AI and NLP as parallel concepts; however, AI broadly encompasses NLP. The authors are advised to refine the description of AI methods and consider adding statistical analysis techniques.

Response 5: The observation is appreciated. To address this problem, IA and NLP sites have been replaced with AI (specifically NLP), along with a useful taxonomic review to illustrate this methodological decision.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

Please find attached.

Comments for author File: Comments.pdf

Author Response

Reviewer 4

 

 

Observation 1: The manual annotation process (1,800 clauses) is central to model training but is not fully described in terms of coder agreement, training duration, or annotation software. Add inter-annotator reliability (e.g., Cohen’s kappa) and more detail on rubric validation.

 

Response 1: The observation is appreciated. The methodology has been changed.

 

Observation 2: While the authors include 31 PRCs, two (Estación Central and Pedro Aguirre Cerda) are omitted. The rationale for their exclusion should be more explicitly stated. Were their plans unavailable or too incomplete for analysis? This omission may bias the spatial conclusions.

Response 2: Thanks for pointing out this. Estación Central PRC (2003) is under judicial review and unavailable; Pedro Aguirre Cerda has no approved PRC. This explanation is now included.

Observation 3: Although DeepSeek-R1 is adapted to the Chilean legal context, details about the fine-tuning process, training epochs, loss function, and language preprocessing steps are insufficiently elaborated for replication. A supplementary appendix or link to a GitHub repo would strengthen methodological transparency.

 

Response 3: The methodology section now incorporates a more detailed explanation.

 

Observation 4: Figures 1–3 show important spatial patterns but lack sufficient resolution and explanation of symbology. Ensure final publication includes high-resolution images and clear legends. Also, consider including confidence intervals or uncertainty bands, especially for norm-based rubrics.

 

Response 4: Observation is understood. The original maps have been attached separately to ensure good resolution.

 

Observation 5: Line 15: The term “88% classification accuracy” needs clarification—across all classes? Was this macro-averaged or per-class?

 

Response 5: This detail is very technical and is better explained in the methodology and has been removed from the abstract.

 

Observation 6: Section 3.3: Consider briefly comparing with international examples to contextualize Santiago's zoning incentive disparities.

 

Response 6: We have incorporated international examples that allow us to give comparative value to the findings.

 

Observation 7: Line 577: The idea of a transferable AI model is promising—expanding on specific transfer learning potentials (e.g., to Brazil or Colombia) would be helpful.

 

Response 7: An idea of transfer has been made that can be replicated, which has been detailed in the conclusions.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

I find the answers to the review relatively short but the editions in the text are significant and addressing well the raised questions in the first version of the manuscript.

The work is more finite and more easily replicable with good showcase of innovative and ethical use of AI in urban planning affairs in the Global South. The last further increasing the value of the article.

 

Small remarks can be made on the following:

• line 180 - informal urban? ... development?
• pages 6 & 7 can be squeezed to half a page explaining the mid (2nd) level steps and the rest of the details which are still valuable but slightly contributing to an unbalanced volume of the methodological section which can be considered as third+ level steps can be provided in additional second appendix (the one with the ordinances is also valuable and provides more clarity). The figure (2) can be preceding the description as it shows 1st level steps and can serve as intro to the mid level description of the undertaken steps.
• line 385 - revise the sentence

Best wishes to the author

Author Response

Observation 1: line 180 - informal urban? ... development?.

Response 1 : Thanks for the precise information. It was amended.

Observation 2: pages 6 & 7 can be squeezed to half a page explaining the mid (2nd) level steps and the rest of the details which are still valuable but slightly contributing to an unbalanced volume of the methodological section which can be considered as third+ level steps can be provided in additional second appendix (the one with the ordinances is also valuable and provides more clarity). The figure (2) can be preceding the description as it shows 1st level steps and can serve as intro to the mid level description of the undertaken steps.

Response 2: Thanks for this suggestion. Appendix B was created and the new version of the manuscripts adopts these recommendations for reshape the methodological description.

Observation 3: line 385 - revise the sentence.

Response 3: Thanks for the precise information. It was amended for clarity.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The revised manuscript has improved a lot and meets the publication standard.

Author Response

Thank you very much. Best wishes. 

Reviewer 3 Report

Comments and Suggestions for Authors

All my concerns raised in the previous round have been thoroughly addressed. I recommend accepting this manuscript. 

Author Response

Thank you very much. Best wishes.