Beyond Proximity: Open Data and Explainable AI for Spatial Equity in X-Minute Cities

1. Introduction

The 15-minute city stands out among other concepts in contemporary urban policy debates. The idea’s attraction lies in the simple rule of thumb: everyday services and opportunities need to be within walking, cycling, or local transport commuting distance. The larger class of X-minute city paradigms builds on similar ideas but applies different thresholds (5, 10, 15, and 20 minutes) to accommodate various urban forms and mobility systems [1,2,3,4]. Current research has also identified how the concept is used in practice, signaling a trend from academic discussion to policy branding, performance measuring, and urban planning [38].

The attractiveness of the idea is also its fundamental weakness. Proximity is easy to describe but hard to manage in reality. Neighborhoods might seem 15-minute compliant, while their inhabitants face low quality, high costs, dangerous paths, poor pedestrian environment, limited capacity, short opening hours, inaccessibility for disabled people, and lack of public transportation. In this way, the real challenge is not just proximity but the ability of diverse inhabitants to use available nearby opportunities in an effective, safe, sufficient, affordable, and meaningful way [5,6,7].

Such an issue is especially relevant in the context of fast growth of the literature on proximity cities. Reviews focused on metrics show significant methodological diversity in proximity thresholds, travel modes, basket of amenities, spatial units, and accessibility indicators [4]. Large-scale studies using open data demonstrate large disparities in proximity performance between and within cities as well as global inequalities in access to basic services [39]. Meanwhile, research on urban peripheries and suburbs suggests avoiding interpretation of the 15-minute city as an urban-core paradigm that can be applied equally to all types of settlements [23,41].

Open geospatial data make it easier to measure this debate. OpenStreetMap (OSM), points of interest (POI), GTFS transit feeds, population grids, pedestrian network, sidewalk data.

AI and urban analytics add another layer to this debate. Machine learning, computer vision, natural language processing, large language models, clustering, prediction, and explainable AI are increasingly discussed in relation to urban planning, smart-city governance, walkability assessment, land-use analysis, perception, and decision support [10,11,12,13,14,15,16,17,18,47]. Recent planning scholarship stresses that AI can support different planning phases, but also that planners need social-technical evaluation, transparency, accountability, and awareness of bias before such tools can be considered legitimate [17,18,47,48].

The gap addressed in this article is therefore integrative. Existing reviews and empirical studies have examined 15-minute city concepts, X-minute metrics, accessibility equity, open spatial data, AI in planning, AI governance, AI-assisted walkability, and urban analytics. However, these streams remain insufficiently connected. This article asks how open geospatial data and explainable AI can be organized into a planning-oriented framework for diagnosing spatial equity in proximity-based urbanism.

The paper makes three contributions. First, it clarifies why proximity, accessibility, and spatial equity should not be treated as interchangeable terms. Second, it strengthens the review design through a transparent narrative-review method, corpus organization, search-string table, and evidence-role logic. Third, it develops an operational Open-Data Explainable Proximity Planning Framework that links data, metrics, AI-enabled diagnosis, explanation, governance, planning decision support, and monitoring.

2. Literature Review

2.1. Proximity-Based Urbanism and X-Minute Cities

The 15-minute city is part of a broader shift from mobility-centered planning to accessibility-centered planning. Mobility-centered planning asks how quickly people can move across urban space; accessibility-centered planning asks whether residents can reach the opportunities and services required for everyday life [24]. The X-minute city translates this accessibility logic into communicable time thresholds, but these thresholds are planning assumptions rather than universal measures [2,3,4].

Recent studies have sharpened this point. The global framework developed by Bruno et al. shows substantial heterogeneity in access to essential services and argues that proximity models must be generalized for different population densities and urban structures [39]. The limitations pointed out by Mouratidis concern issues related to sustainability, fairness, livability, and spatial thinking, warning that the ideas of proximity might be oversimplified when conceived of as an established spatial equation [6]. The empirical literature on outskirts and peripheries proves once again that the issue of short distances cannot be separated from public transport and city design [23,41,45].

2.2. Spatial Equity, Accessibility Justice, and Capability

Equity-aware proximity planning requires a normative move from mere recognition of nearby services to the assessment of the distribution, adequacy, and usability of the access. Distributive equity refers to unequal distribution of services and opportunities between locations or social groups. Adequacy equity examines whether all individuals receive enough of an opportunity. Capabilities analysis considers how capable residents are to use opportunities, given their income, physical mobility, need to care for others, personal safety, age, gender, and time constraints [5,24,25,26,27,28,29].

It is crucial to distinguish between these types of approaches since many proximity indicators make use of thresholds which appear objective but involve normative choices nonetheless. The questions are related to the determination of what kind of service is important, what transport mode is assumed, what walking speed is used, and what time period is considered. The other factors are whether individuals with disabilities, older residents, caregivers, or poorer groups are accounted for, and whether quality and capacity of services are involved. Studies related to variation in walking access of individuals and time periods, age-related active accessibility, and X-minute city walkability show that equal spatial distance does not necessarily mean equal access [32,35,43].

2.3. Open Geospatial Data and Measurable Accessibility

Open geospatial data have made proximity-city research more reproducible and comparable. OSM networks, POI databases, GTFS transit feeds, population grids, sidewalk data, and street-view imagery allow accessibility analysis at neighborhood, metropolitan, and cross-city scales [5,9,20,34,40]. Studies using open geospatial data can identify where access is weak, how accessibility varies by mode and service category, and which neighborhoods may require priority intervention.

However, open data should not be treated as neutral evidence. OSM completeness, POI category accuracy, transit-feed reliability, population-grid resolution, and street-view coverage vary across places [34]. These limitations can bias accessibility scores and undermine the validity of equity diagnosis. Therefore, open-data proximity studies should include data audits, validation procedures, uncertainty registers, sensitivity analysis, and documentation of assumptions.

2.4. AI-Enabled Urban Analytics, Explainability, and Governance

AI-driven analytics may help with pattern recognition, perceptual modeling, development of neighborhood typologies, scenarios evaluation, and identification of gaps in access, so long as the underlying assumptions and data restrictions are made transparent. Recent studies on the use of AI in planning show that the use of AI is possible throughout all planning stages, ranging from data analytics, spatial diagnosis, policy design, citizen engagement to monitoring [17,18]. Studies on the use of AI in smart cities and algorithmic urban planning also suggest growing application in forecasting, classification, simulation, optimization and decision support [10,47].

At the same time, governance is a major issue. AI may perpetuate historical biases, hide value-laden choices, invent or exaggerate certainty, avoid accountability and politicize planning in a technical way [16,19,47,48,49]. Therefore, explanation in urban planning cannot be confined only to feature significance and model transparency. It should include public explanation, uncertainty, contestation, participation and documentation of how evidence leads to decisions [21,22,48,49]. The major review streams and the remaining integration gap addressed by this study are summarized in Table 1.

2.5. From Proximity to Equity-Sensitive Accessibility

The main drawback of proximity-based approach to planning is the consideration of proximity as a sufficient measure of access. Residents actually access services and facilities based on a range of attributes, including network distance, travel time, service quality, cost, safety, availability at a certain time, accessibility to people with disabilities, and social acceptability. Proximity of a park does not mean that it is well-maintained; proximity of a clinic does not imply affordability; proximity of a school may mean that the school is oversubscribed; presence of sidewalks on maps does not guarantee usability.

Therefore, equity-oriented proximity planning requires three assessments of accessibility opportunities. Firstly, distributional assessment – whether there are inequalities in distribution of access opportunities between different neighborhoods and population segments. Secondly, a sufficiency assessment – whether everybody has access to some acceptable minimum of services. Thirdly, a capability assessment – whether residents are capable to access their spatial opportunity given their abilities, conditions, and constraints [24,25,26,27,28,29].

Perception and safety are other critical issues. Traditionally, accessibility measures are represented through networks with origin points, destination points, and costs. Yet, for residents routes are experienced through fear, comfort, aesthetic value, maintenance, exposure to traffic, weather conditions, light, and social meanings. Perceptions and AI-enabled analyses of walkability and perception can help to supplement traditional network analysis provided that the limitations of image data, cultural assumptions, training sample, and proxy values are carefully analyzed [8,12].

Peripheral and suburban environments also need to be considered for proximity planning. Central areas of cities are more likely to provide short thresholds of accessibility due to clustering of services, workplaces, and multimodal connections. However, in contrast to central cities, peripheries might require particular approaches to local services, active modes of transport, public transit, polycentric growth, and metropolitan connection [23,41,45]. The goal in such contexts is not to reduce standards of access, but to develop a particular way to reach sufficiency.

2.6. Open Geospatial Data and Accessibility Measurement

2.6.1. From Concept to Measurement

Only with explicit specification of the underlying assumptions does the proximity city concept provide any analytical insights. Scholars and urban planners need to identify the important destinations, consider the modes of transport used, apply temporal thresholds, define the representation of origins and destinations, and decide on whether to measure accessibility based on counts, travel time, gravity measures, composite indexes, inequality indexes, or others [3,4,5].

This can be achieved with the use of open data through repeated measurements. However, measurement decisions themselves remain normative. While the cumulative opportunity index makes the process transparent, it is still dependent on choice of amenities and threshold values. Gravity measures reflect distance decay effect but can pose problems of communication. Composite indexes allow for the dashboard view but hide decision-making around weights. Inequality indexes help to demonstrate unequal distributions but do not, in themselves, reveal anything about causality. Enough thresholds are desirable from the point of justice in urban planning but require justifiable criteria. The principal accessibility and equity metric families used in X-minute city research are summarized in Table 2.

2.6.2. Data Sources for Proximity and Equity Analysis

Various sources of open geospatial data support different kinds of analysis. Street data allows routing on foot, by bicycle, and by public transport. Data about Points of Interest (POI) allow categorization of destinations. GTFS feeds describe planned services of public transportation. Population grids help assess population-weighted accessibility and measure inequalities. The use of sidewalk network data improves realism of pedestrian analysis. Street-level images help assess issues related to perceived safety, maintenance, presence of greenery, and route quality. While municipal data can have higher authority when it comes to local services, they can also be less comparable across cities [5,8,9,12,34,40]. The open geospatial data sources used for proximity and equity analysis, together with their planning value and limitations, are summarized in Table 3.

2.6.3. From Mapping to Diagnosis

Most maps generated by researches contain information about neighborhoods that exceed the predefined threshold. Although such maps provide better understanding, they cannot be sufficient to perform a planning diagnosis. Lack of access could be caused by absent destination, inadequate network of streets, infrequent transit, hazardous route, inaccessible infrastructure, improper service alignment, low quality, inconvenient opening hours, high cost or discrimination. Therefore, diagnostics should establish connections between the patterns and the potential mechanisms of planning.

Here the AI and explainable analytics could be applied. It is not necessary to use the AI to generate an isochrone or to count amenities. Its real value could be seen when multiple layers of information are combined to find out patterns, classify neighborhoods, determine probable access gaps, calculate perception-based barriers or analyze different scenarios of intervention. Here transparency is needed because their outputs will affect resource allocation and justification.

2.7. AI and Explainable Urban Analytics for Planning Evidence

2.7.1. AI as a Planning-Support Layer, Not a Planning Substitute

In this review, AI is not viewed as the subject matter of investigation. What matters here is proximity-based planning and the spatial justice issues it involves. The relevance of AI is limited to how AI can contribute to measurement, diagnosis, explanation, scenario comparison, or monitoring. Such a distinction is important since planning is a normative and democratic activity, not merely a technical optimization process.

The use of AI analytics may help with: pattern recognition over large geospatial datasets; extraction of street level features from imagery data; clustering of neighborhoods according to accessibility, vulnerability, and urban morphology; simulation of accessibility impacts; analysis of planning documents or public comments; and scenario comparison of interventions. However, such uses of AI must always be constrained by considerations of planning. Outputs from models need to be used as pieces of evidence for public deliberation, not as the final decision-makers themselves [10,17,18,47].

2.7.2. Explainability as Public Justification

Explanation in the context of urban planning is characterized by four different types. First of all, there is model explanation that includes such types of explanation as feature importance, local explanation, counterfactual reasoning, and surrogate model. The second type of explanation is related to causal caution, which helps not to present statistical relationship as causal one. The third type of explanation is stakeholder explanation, which is the interpretation of technical results into planning narrative. The fourth type is accountability documentation.

Tools such as LIME and SHAP can support local or feature-level explanations, but they do not by themselves produce planning legitimacy [36,37]. A feature may be statistically important without representing a causal planning lever. A model may be accurate but rely on biased data. A dashboard may be visually convincing while hiding normative choices. Therefore, explainable urban analytics should connect model outputs to planning mechanisms, uncertainty statements, and participatory validation.

2.7.3. Governance Safeguards for Urban AI

Governance safeguards are not a final add-on after the technical pipeline has been built. They should be present from problem definition through data collection, modeling, explanation, intervention selection, and monitoring. In proximity planning, governance begins with apparently simple choices: which services count, whose needs are represented, which thresholds are legitimate, and which groups are prioritized.

The ethical concerns of AI in planning include bias, privacy, misinformation, accountability gaps, lack of transparency, and overreliance on automated judgment [48]. Critical accounts of generative AI and anticipatory urban governance also warn that AI systems can become quasi-oracular sources of future-oriented policy authority if uncertainty and contestability are not built into the process [49]. For proximity planning, this means that every analytical output should be accompanied by a record of assumptions, limitations, and routes for public challenge. The AI and urban analytics methods relevant to proximity planning, including their potential uses, planning value, and principal risks, are outlined in Table 4.

3. Research Methodology

3.1. Review Type and Purpose

This article is a narrative and critical review with a conceptual framework contribution. It is not intended to be a systematic review, a scoping review, a bibliometric review, a meta-analysis, or a PRISMA review. This review does not purport to identify or screen out all studies that fit the criteria through an exhaustive methodology. Instead, it summarizes selected peer-reviewed and seminal literature in order to connect various separate concepts such as proximity planning, spatial equity, open geospatial data, and AI-enabled urban analytics.

The review relies on evidence in five different functions. First, proximity city conceptual studies explain the planning principles and assumptions behind 15-minute and X-minute urbanism [1,2,3,4,38,39]. Second, studies of accessibility and open data provide examples of methods for quantifying proximity and spatial equity [5,9,20,23,32,34,40,45]. Third, transport justice, accessibility, and capabilities frameworks form the normative approach separating proximity from equitable access [24,25,26,27,28,29,30,31]. Fourth, studies of AI and urban analytics give an example of how computational techniques can be used for prediction, perceptual analysis, planning support, and urban governance [10,11,12,13,14,15,16,17,18,46,47]. Finally, critical governance literature highlights the issues associated with bias, opacity, loss of public trust, accountability, and substituting political decision-making for technology [16,19,21,22,47,48,49].

3.2. Literature Identification and Evidence Organization

To improve transparency while preserving the narrative-review character of the article, the corpus was organized through a structured but non-PRISMA identification strategy. Searches were conducted across major scholarly databases and publisher platforms, with emphasis on peer-reviewed journal articles, books, and landmark theoretical works available up to June 2026. The recent update prioritized 2024-2026 publications on 15-minute city implementation, open geospatial measurement, AI in urban planning, urban digital twins, explainability, and AI governance.

Backward and forward citation checking was used selectively to identify foundational sources that anchor the conceptual argument. The review corpus was then coded thematically according to evidence role: conceptual framing, accessibility measurement, spatial-equity theory, open-data infrastructure, AI/urban analytics, explainability, governance, planning decision support, and monitoring. Because the corpus is selective, conclusions should be interpreted as a conceptual synthesis and research agenda rather than as a statistically representative map of the field. The corpus organization and evidence-role protocol used for this narrative review are summarized in Table 5.

3.3. Guiding Questions

The synthesis is organized around five guiding questions: (1) How have 15-minute and X-minute city studies conceptualized proximity, accessibility, and neighborhood completeness? (2) How have recent studies critiqued proximity-only approaches from the perspective of equity, perception, peripheries, and governance? (3) What open geospatial data sources and accessibility metrics are used to operationalize proximity-based planning? (4) How are AI and urban analytics being used to support walkability, accessibility, perception, and planning decision-making? (5) What framework can connect open data, spatial equity, explainable AI, and governance into proximity-city planning?

4. Research Results and Discussions

4.1. Open-Data Explainable Proximity Planning Framework

4.1.1. Rationale

The preceding sections show that proximity-based planning requires more than a spatial threshold. The 15-minute and X-minute city literature provides a communicable planning concept. Accessibility studies provide metrics and open-data methods. Spatial-equity literature provides justice criteria. AI and urban analytics offer pattern detection, perception analysis, and scenario comparison. Governance scholarship provides safeguards against opacity and technocratic decision-making. The proposed framework connects these streams into a planning process.

The Open-Data Explainable Proximity Planning Framework is not an empirical model and does not claim that explainable AI automatically improves planning outcomes. It is a conceptual and operational framework for organizing evidence. Its purpose is to help planners move from proximity goals to transparent data, explicit measurement, equity diagnosis, interpretable analytics, governed decision support, and monitoring.

4.1.2. Framework Figure

Figure recreation note. The figure can be recreated in Word, Graphviz, or any vector-drawing tool as nine rounded rectangles connected by downward arrows. The governance safeguard is shown as a dashed node connected to the data, equity, AI, explainability, and decision-support layers. A dashed feedback arrow returns policy monitoring to the initial proximity goal, representing iterative learning.

4.1.3. Operational Layers

The framework has nine connected layers. Each layer should be operationalized through input data or assumptions, a specific analytical task, a planning output, an explicit governance risk, and at least one example indicator. This prevents the framework from remaining a purely descriptive diagram and turns it into a planning-audit structure. The inputs, analytical tasks, planning outputs, governance risks, and example indicators associated with the nine framework layers are detailed in Table 6.

Figure 1. Open-Data Explainable Proximity Planning Framework. The DOT-style diagram links proximity goals, open geospatial data, accessibility measurement, spatial-equity diagnosis, AI/urban analytics, explainability, governance safeguards, decision support, and policy intervention/monitoring through an adaptive feedback loop.

Figure 1. Open-Data Explainable Proximity Planning Framework. The DOT-style diagram links proximity goals, open geospatial data, accessibility measurement, spatial-equity diagnosis, AI/urban analytics, explainability, governance safeguards, decision support, and policy intervention/monitoring through an adaptive feedback loop.

4.1.4. From Indicators to Planning Action

The framework is designed to prevent pass/fail proximity mapping from being mistaken for planning action. If low access is caused by lack of destinations, relevant interventions may include facility siting, land-use change, service incentives, or public-service provision. If the problem is poor connectivity, interventions may involve crossings, pedestrian links, sidewalk upgrades, traffic calming, and cycling infrastructure. If weak transit is the barrier, planners may consider route restructuring, frequency improvements, or multimodal integration. If safety or perception is the barrier, maintenance, lighting, design, social programming, and community safety measures may be more relevant. If affordability is the problem, spatial proximity alone is insufficient.

Planning decision support should therefore compare intervention scenarios rather than produce a single access score. Adding new services may improve access but create gentrification pressures. Sidewalk improvements may increase walkability while leaving service affordability unresolved. Transit improvements may strengthen metropolitan accessibility while not solving local service shortages. The framework requires planners to make these trade-offs explicit.

4.2. Discussion: Research Agenda and Planning Implications

As such, the literature suggests that future generations of research on proximity-cities need to move from simply measuring proximity to models that are understandable, equitable, and governance-focused in the context of proximity planning. This research agenda does not seek to replace urban planners with artificial intelligence systems, but to build up the evidence base that supports urban planning decisions, by: (a) setting clear thresholds, (b) auditing open data, (c) diagnosing access deficiencies, (d) using AI systems only in situations where they add value through their explanatory power, and (e) reporting how evidence influences decision-making.

Methodological examples from other domains can inform the evaluation of urban analytics. Synthetic-data augmentation for imbalanced breast-cancer classification [50] and staged random-forest intrusion detection in IoT networks [51] illustrate why data balance and class-specific performance deserve attention alongside aggregate accuracy. For proximity planning, the corresponding research question is whether model errors differ across neighborhoods and population groups; synthetic samples cannot replace missing observations about local access.

Graph–language integration links relational representations, retrieved evidence, and reasoning [52], while autonomous zero-trust network-slicing research examines trust management and policy enforcement across connected domains [53]. These perspectives suggest research directions for evidence traceability and controlled access in urban decision-support platforms. Their usefulness for spatial-equity diagnosis requires urban-specific and participatory evaluation. These four cross-domain sources are contextual additions outside the June 2026 review corpus and do not provide empirical validation of the proposed planning framework.

In Urban Science and other urban-planning journals, this framework also draws attention to a specific interdisciplinary opportunity. AI and open data are only meaningful for urban planning if and when their technical output is translated into spatial diagnosis and interventions as well as public accountability. The future research directions, recommended approaches, planning value, and risks identified by the review are summarized in Table 7.

4.3. Review Limitations and Boundary Conditions

The present paper employs a narrative and critical review approach rather than that of systematic reviews. As such, it does not provide PRISMA screening figures, quality assessments scores, full coverage of all databases searched or meta-analysis. The merit of the paper is in its ability to integrate knowledge between disciplines; the weakness of the paper is in its selectively curated and reviewed corpus.

The review is likely to underestimate the literature about technical GeoAI studies, studies by government about algorithmic accountability, studies about informal mobilities, non-English planning literature and locally conducted accessibility studies. In this regard, the proposed conceptual framework serves more like an agenda for further research and a planning audit tool but not a fully operationalized model, which requires empirical validation in different urban environments especially those that are peripheral, small, Global South and data-scarce.

Another limitation is open data and governance of AI. While open geospatial data can improve transparency and reproducibility of results, they might be incomplete, outdated, spatially unbalanced and biased. AI analytics could help diagnose problems and analyze scenarios, however, such analysis depends on data, assumptions, model design and governance.

5. Conclusions

The 15-minute city and its larger family of X-minute cities emphasize the issues related to accessibility, local living, active mobility, and neighborhood completeness. The strength of these concepts in terms of conceptualization lies in their communicability, and the weakness lies in the assumption of proximity being a sufficient measure of spatial justice. This review argues that proximity is a necessary but insufficient concept: the challenge of planning is to enable different populations to access necessary services effectively, safely, affordably, and sufficiently.

The current literature review includes all the necessary elements of a better approach to planning – proximity city concepts, accessibility indicators, open geospatial data, theory of spatial equity, urban analysis enhanced by AI technologies, explainability techniques, and mechanisms of governance. The contribution of this paper consists in bringing these components together into the framework of Open Data Explainable Proximity City Planning.

This framework consists of the following elements: proximity city objectives, data infrastructure, accessibility measurement, spatial equity diagnosis, urban analytics enabled by artificial intelligence, explanation, governance, decision-making, and monitoring.

This paper provides a conceptual review, not an empirical validation of any particular technique. The usefulness of the suggested framework can be estimated only with the help of its application to open data case studies, participatory planning practices, and longitudinal evaluation of the implemented interventions. The main message of this review is that proximity planning should be led by planning and assisted by AI. Artificial intelligence will help to diagnose, interpret, simulate, and communicate evidence. But it will not replace public consultation, professional judgment, or democratic responsibility.