Jiatong Li

Jiatong Li

Auto-generated profile, based on publicly available information

Jiatong Li


Research Focus & Metrics

Academic Career Length

1Year (2026-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
29 July 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
JWJiali Wang
Jiali Wang
JLJiatong Li
Jiatong Li
YLYuan Lai
Yuan Lai
Urban planning increasingly confronts complex and wicked problems that require planners to draw on experiences accumulated through previous practice and adapt interventions to local contexts. This case-based knowledge is largely embedded in unstructured narratives such as planning reports, policy documents, and project descriptions. Despite advances in AI-driven urban analytics, such experiences remain difficult to systematically compare, retrieve, and reuse due to their contextual complexity and fragmented representation. This study presents CaseFlow, a human-AI collaborative framework for structuring and retrieving empirical experiences to support cross-case analysis and traceable reasoning in urban planning. The framework integrates large language models (LLMs), knowledge graphs, and GraphRAG into a unified workflow that transforms unstructured case narratives into structured, computable knowledge networks. A dynamic schema mechanism enables domain experts to define entity-relation structures for specific research questions. To improve controllability and transparency, the framework employs a multi-agent, human-in-the-loop pipeline in which researchers supervise entity extraction and relation inference at each stage. A hybrid retrieval and reasoning engine—combining semantic search, full-text search, and graph expansion—supports cross-case comparison and traceable evidence retrieval. Evaluation on 50 urban regeneration cases demonstrate the framework's effectiveness in structuring planning experiences into knowledge graphs and supports both pattern extraction and context-sensitive case retrieval. Overall, this urban case experience extraction framework enables experience to be structured, compared, and interpreted alongside formal rules and institutional conditions, supporting more transparent and context-sensitive planning.

11 August 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
JLJiatong Li
Jiatong Li
AHAnni Hu
Anni Hu
SYSijie Yang
Sijie Yang
KZKun Zhou
Kun Zhou
ZFZicheng Fan
Zicheng Fan
YLYuan Lai
Yuan Lai
Cities are complex systems where nonlinear interactions among micro-scale urban elements produce emergent macroscopic structures. Understanding this complexity requires identifying the meso-scales where local diversity balances collective order. This study reframes the scaling problem in planning by centering on urban locality—the distinctive spatial identity of an area. As a collective understanding shaped by human interactions, locality provides a human perspective on complex urban systems. However, its geographical scale that is critical for translating place-based diversity into coherent planning units, remains poorly defined. This study proposes a multi-scale analytical framework to identify the emergent scale of urban locality through urban form and function. Using self-supervised Graph Neural Networks and multi-scale spatial clustering, we analyze street view imagery and point-of-interest data from six global megacities. By examining how system complexity changes with scale, we identify an emergent locality scale for each city, and compare it with existing administrative hierarchies. Results reveal substantial variation in locality scales across cities, yet they generally correspond to neighborhood-level administrative units, underscoring relevance for neighborhood design and community planning. At the emergent scale, locality reveals distinct patterns where city centers develop locality through functional mixing and homogeneous urban form, whereas suburban areas exhibit locality rooted in functional homogeneity. Overall, this study provides a scalable methodology for identifying urban locality as a human-centered and empirically measurable meso-scale, offering insights into complex urban systems while supporting neighborhood governance, context-sensitive interventions and cross-city analysis.