This interdisciplinary doctoral thesis investigates the convergence between Computer Science and Architecture, fields that often appear to have diverging foundations. Computer science is basically formal, quantitative, and data-driven, whereas architecture and urban planning are rooted in contextual specificity, human-centered design, and complex stakeholder interactions. Based on this observation, the thesis addresses a core research question: Where and how could Computer Science and Architecture converge to advance the design and management of urban systems? To answer this, the thesis introduces INTERPRET: an INtegrated and adapTive framEwoRk to support Policy-makers in the urban EnvironmenT. The framework combines Human-Centered Conceptual Design (HCCD), spatial clustering methods (SiMBA: Systematic clustering-based Methodology to support Built environment Analysis), tools based on Geographic Information Systems (GIS Tools), and a novel recommendation framework designed for fairness and diversity in group decisions: ADAPT (fAirness and Diversity for sequentiAl grouP recommendaTions). HCCD is introduced as a novel method to structure the dialogue between domain experts and computer scientists, enabling the co-design of data models that reflect real-world constraints and objectives: it lays the foundations for transparency and interpretability in subsequent data-driven analyses, and constitutes a significant methodological innovation in interdisciplinary design processes. SiMBA is employed as a clustering strategy to identify emergent patterns and typologies across complex urban data. Clustering is particularly suited for exploratory urban analysis, as it enables the detection of meaningful structures in heterogeneous, noisy, and spatially distributed datasets. Rather than prescribing predefined categories, clustering allows urban planners and stakeholders to recognize urban-area typologies in a bottom-up fashion, thus supporting scenario definition, comparative analysis, and context-sensitive design. The GIS plugins CA (Cluster Analysis) and CTA (City Transport Analyzer) integrate spatial clustering and transportation diagnostics directly into the Q-GIS (Quantum-Geographic Information Systems) platform. These tools enable planners and decision-makers to perform advanced spatial analyses without requiring programming knowledge, facilitating usability and dissemination. Finally, ADAPT is a group recommendation framework that models sequential decision-making among diverse stakeholders. The choice of a recommendation-based approach stems from the increasing need to assist group decisions in urban planning processes, where multiple actors must negotiate conflicting priorities. In this context, Fairness and Diversity are critical not only to ensure social acceptance and equity, but also to avoid systemic bias and improve long-term satisfaction with the implemented policies. ADAPT formalizes these concerns and embeds them into its optimization logic. Each module is grounded in real-world case studies - including Milan and Rio de Janeiro - and evaluated across computational performance and practical applicability. Notably, the thesis demonstrates that urban case studies do not merely benefit from data-driven methods; they also pose novel, non-trivial challenges that stimulate advancements in computer science. The challenges addressed in this work include: (i) the absence of ground truth in urban diagnostics, (ii) the scarcity of labeled data, (iii) the small datasets typical of real-world urban planning contexts, and (iv) the need to model fairness and diversity in sequential group recommendations. These aspects are systematically tackled through the methodological innovations introduced in HCCD, the validation-by-interpretability approach in GIS diagnostics, and the fairness-aware logic embedded in ADAPT. The findings suggest that urban system analysis can greatly benefit from the reciprocal exchange between the formalization strengths of computer science and the contextual richness of architecture. This work provides both a practical toolset and a conceptual roadmap for future research at this disciplinary intersection.
L'incontro tra informatica e architettura, due ambiti spesso percepiti come distanti per approccio e finalità, rappresenta oggi una sfida strategica per la progettazione e la gestione dei sistemi urbani. L’informatica è caratterizzata da approcci quantitativi, formali e data-driven, mentre l’architettura e la pianificazione urbana sono profondamente dipendenti dal contesto e caratterizzati da metodi spesso difficilmente generalizzabili, fortemente dipendenti dalla complessità dell’interazione tra gli attori coinvolti e che richiedono un forte coinvolgimento umano nel design delle soluzioni. A partire da questa osservazione, la tesi affronta una domanda di ricerca centrale: Dove e come dovrebbero convergere informatica e architettura per supportare la progettazione e la gestione dei sistemi urbani? Per rispondere, viene proposto il framework INTERPRET: INtegrated and adapTive framEwoRk to support Policy-makers in the urban EnvironmenT. Il framework integra quattro componenti principali: lo Human-Centred Conceptual Design (HCCD), metodi di clustering spaziale (SiMBA: Systematic clustering-based Methodology to support Built environment Analysis), strumenti diagnostici basati su Geographic Information Systems (GIS), e un sistema di raccomandazione sequenziali di gruppo (ADAPT: fAirness and Diversity for sequentiAl grouP recommendaTions) focalizzato su fairness e diversità delle raccomandazioni. HCCD è una nuova metodologia introdotta in questa tesi per strutturare il dialogo tra esperti di dominio e professionisti tecnici, permettendo la co-progettazione di modelli di dati che riflettano vincoli e obiettivi reali. Rappresenta un’innovazione metodologica che pone le basi per analisi successive trasparenti e interpretabili. SiMBA viene utilizzato per identificare pattern emergenti all’interno di dati urbani complessi. Il clustering è scelto per la sua capacità di supportare l’esplorazione di dataset rumorosi ed eterogenei, senza imporre categorie predefinite: ciò consente la costruzione bottom-up di scenari e tipologie urbane significative. I plugin GIS, Cluster Analysis (CA) e City Transport Analyzer (CTA), integrano strumenti avanzati di clustering e diagnostica dei servizi di trasporto direttamente nell’ambiente Quantum-Geographic Information Systems} (QGIS), offrendo funzionalità sofisticate a professionisti non programmatori. ADAPT è un sistema di raccomandazione pensato per supportare decisioni sequenziali all’interno di gruppi eterogenei. La scelta di un approccio basato su raccomandazioni nasce dalla necessità, crescente nella pianificazione urbana, di assistere processi decisionali collettivi caratterizzati da preferenze conflittuali. In tale contesto, fairness e diversità non sono solo requisiti normativi, ma elementi fondamentali per garantire l’accettazione sociale e prevenire distorsioni sistemiche. Ciascun modulo è stato testato attraverso casi studio reali, tra cui Milano e Rio de Janeiro, e valutato sia per le sue prestazioni computazionali sia per la rilevanza pratica. In particolare, la tesi dimostra come i contesti urbani non solo traggano beneficio dai metodi data-driven, ma pongano anche sfide metodologiche nuove e non banali all’informatica. Tra le criticità affrontate nella tesi le più significative sono: l’assenza di ground truth nei dati urbani, la scarsità di esempi etichettati, dataset ridotti e la necessità di modellare fairness e diversità nei processi di raccomandazion sequenziali di gruppo. Questi aspetti vengono affrontati attraverso soluzioni dedicate in HCCD, nella diagnostica GIS basata sull’interpretabilità, e nel framework ADAPT. Nel complesso, i risultati evidenziano come l’analisi dei sistemi urbani possa trarre notevoli benefici da uno scambio reciproco tra il rigore formale dell’informatica e la ricchezza contestuale dell’architettura. Questa tesi propone così un set di strumenti concreti e una visione concettuale per promuovere future sinergie tra design e dati.
Urban analysis frameworks: from data modeling to sequential group recommendations
LENZI, EMILIA
2024/2025
Abstract
This interdisciplinary doctoral thesis investigates the convergence between Computer Science and Architecture, fields that often appear to have diverging foundations. Computer science is basically formal, quantitative, and data-driven, whereas architecture and urban planning are rooted in contextual specificity, human-centered design, and complex stakeholder interactions. Based on this observation, the thesis addresses a core research question: Where and how could Computer Science and Architecture converge to advance the design and management of urban systems? To answer this, the thesis introduces INTERPRET: an INtegrated and adapTive framEwoRk to support Policy-makers in the urban EnvironmenT. The framework combines Human-Centered Conceptual Design (HCCD), spatial clustering methods (SiMBA: Systematic clustering-based Methodology to support Built environment Analysis), tools based on Geographic Information Systems (GIS Tools), and a novel recommendation framework designed for fairness and diversity in group decisions: ADAPT (fAirness and Diversity for sequentiAl grouP recommendaTions). HCCD is introduced as a novel method to structure the dialogue between domain experts and computer scientists, enabling the co-design of data models that reflect real-world constraints and objectives: it lays the foundations for transparency and interpretability in subsequent data-driven analyses, and constitutes a significant methodological innovation in interdisciplinary design processes. SiMBA is employed as a clustering strategy to identify emergent patterns and typologies across complex urban data. Clustering is particularly suited for exploratory urban analysis, as it enables the detection of meaningful structures in heterogeneous, noisy, and spatially distributed datasets. Rather than prescribing predefined categories, clustering allows urban planners and stakeholders to recognize urban-area typologies in a bottom-up fashion, thus supporting scenario definition, comparative analysis, and context-sensitive design. The GIS plugins CA (Cluster Analysis) and CTA (City Transport Analyzer) integrate spatial clustering and transportation diagnostics directly into the Q-GIS (Quantum-Geographic Information Systems) platform. These tools enable planners and decision-makers to perform advanced spatial analyses without requiring programming knowledge, facilitating usability and dissemination. Finally, ADAPT is a group recommendation framework that models sequential decision-making among diverse stakeholders. The choice of a recommendation-based approach stems from the increasing need to assist group decisions in urban planning processes, where multiple actors must negotiate conflicting priorities. In this context, Fairness and Diversity are critical not only to ensure social acceptance and equity, but also to avoid systemic bias and improve long-term satisfaction with the implemented policies. ADAPT formalizes these concerns and embeds them into its optimization logic. Each module is grounded in real-world case studies - including Milan and Rio de Janeiro - and evaluated across computational performance and practical applicability. Notably, the thesis demonstrates that urban case studies do not merely benefit from data-driven methods; they also pose novel, non-trivial challenges that stimulate advancements in computer science. The challenges addressed in this work include: (i) the absence of ground truth in urban diagnostics, (ii) the scarcity of labeled data, (iii) the small datasets typical of real-world urban planning contexts, and (iv) the need to model fairness and diversity in sequential group recommendations. These aspects are systematically tackled through the methodological innovations introduced in HCCD, the validation-by-interpretability approach in GIS diagnostics, and the fairness-aware logic embedded in ADAPT. The findings suggest that urban system analysis can greatly benefit from the reciprocal exchange between the formalization strengths of computer science and the contextual richness of architecture. This work provides both a practical toolset and a conceptual roadmap for future research at this disciplinary intersection.| File | Dimensione | Formato | |
|---|---|---|---|
|
PhD_Thesis_July25.pdf
accessibile in internet per tutti
Dimensione
7.38 MB
Formato
Adobe PDF
|
7.38 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/241477