Equity by design as a sustainability multiplier in the citiverse: an empirical proof-of-concept

Citiverse deployments and smart-city systems (e.g., converging artificial intelligence, digital twins, extended reality, IoT, and blockchain within urban governance) are scaling rapidly, yet the equity architecture embedded in their design remains unexamined. Whereas sustainability governance scholarship has treated equity chiefly as a normative outcome, this paper theorizes it instead as a structural design property: a performance multiplier embedded in socio-technical architecture. It extends the System Design and Technology Equity Nexus frameworks to the Citiverse and asks whether equity design quality is independently associated with sustainability performance, across dimensions formally independent of equity by design. The paper draws on the ITU/UNICC/Digital Dubai Citiverse Use Case Taxonomy, the leading United Nations initiative in this domain. It develops two original instruments: the Citiverse Sustainability Impact Index (CSII) and the Equity Design Quality (EDQ) rubric. These are applied across 44 use cases, with their relationship tested through frequentist OLS using a three-specification identification strategy, together with a Monte Carlo measurement-uncertainty analysis. The results offer conditional proof-of-concept support for the central hypothesis. Equity design quality carries independent predictive power over non-equity sustainability performance. This relationship is multiplicative rather than additive once technology richness is accounted for. At this sample size, implementation maturity is the dominant structural predictor. These findings are proof-of-concept rather than causal; causal inference will require longitudinal data. The measurement infrastructure introduced here, the CSII and EDQ rubric, is designed to scale with future documentation and to serve as a replicable basis for tracking the equity design gap over time.

Cities are not neutral platforms. They are arenas of accumulated advantage and structural disadvantage, in which the design of infrastructure, services, and governance systems distributes opportunity unevenly across populations. The emergence of the Citiverse—the convergence of artificial intelligence, digital twins, extended reality, IoT, Metaverse environments, and blockchain within urban governance—represents the most consequential reconfiguration of that design logic in a generation (ITU/UNICC and Digital Dubai, 2025). The question is not whether these technologies will transform city life. They already are. The question is for whom, and whether equity is a design objective or an afterthought. This question sits at the intersection of two literature that have developed largely in parallel. Sustainability governance research treats equity as an outcome to be evaluated after deployment; scholarship on digital urban systems treats it as a design constraint negotiated during development. This paper bridges the two by theorizing equity-by-design as an antecedent structural property of urban technological systems—one that predicts, rather than merely accompanies, sustainability performance.

The global scope of this transformation is significant. By 2050, an estimated 68 percent of the world’s population will live in cities, and the technological systems that manage them are currently being developed (Kaneda et al., 2020). The International Telecommunication Union (United Nations/ITU), the United Nations International Computing Centre (UNICC), and Digital Dubai through their Global Initiative on AI and Virtual Worlds, have documented 44 active use cases across five urban governance sectors (Urban Planning, City Administration, Transport, Public Safety, and Economic Development, Education, and Tourism), each utilizing at least one of nine technology categories and each formally linked to one or more of the seventeen United Nations Sustainable Development Goals (SDGs) (Bibri, 2022).

Taken together, these use cases constitute the most systematically documented international evidence base on Citiverse deployment to date. However, a careful review of that evidence uncovers a structural gap that the SDG alignment framing hides. The vast majority of use cases target SDG 11 (Sustainable Cities and Communities), while SDG 10 (Reduced Inequalities) is addressed by only a small fraction. As it is currently being developed, the Citiverse is overwhelmingly anchored to SDG 11. SDG 10 coverage is both narrow and unevenly distributed across equity design quality, concentrated within a single domain and almost entirely absent from Urban Planning and Public Safety. This is not a moral critique of technological ambition; rather, it is an empirical observation about design choices that have measurable consequences.

Research on the causes of sustainable development impact consistently finds that equity-focused systems (those designed with specific beneficiary populations, accessible delivery methods, and clear inclusion goals) achieve broader and more long-lasting SDG results than systems optimized solely for efficiency (Gupta, 2014; Gupta et al., 2015; Hickey et al., 2015; Hulme, 2009; Kaul and Mendoza, 2003). If that relationship carries over to the Citiverse, then the current design shortfall stops being merely an ethical concern—it becomes a performance problem. Every use case built without equity at its core is leaving measurable SDG impact on the table.

The primary hypothesis of this paper is that equity design quality is independently associated with non-equity sustainability performance in Citiverse deployments, after controlling for technology richness, SDG breadth, implementation maturity, and deployment domain. Use cases with higher equity design quality are expected to score higher on the non-equity sustainability composite, suggesting that equity-by-design functions as a genuine performance multiplier rather than a duplicate measure of the same construct (Stephenson and Crash, 1992).

A secondary hypothesis proposes that this relationship is multiplicative rather than additive. Use cases that combine high equity design quality with greater technology family diversity are expected to produce sustainability gains that exceed the sum of each factor’s individual contribution. This is because equity-by-design principles, when embedded across a wider technological infrastructure, are expected to unlock sustainability pathways that neither factor would activate independently. The interaction term (EDQ × technology richness) is included in the full model to test this directly. This paper presents the first empirical test of both hypotheses.

To test them, I draw on the leading United Nations initiative in this domain—the ITU/UNICC/Digital Dubai, Global Initiative on AI and Virtual Worlds, whose dataset spans 44 use cases covering five domains, nine technology families, eleven SDGs, and four risk dimensions—and develop and implement two new measurement tools. The first is the Citiverse Sustainability Impact Index (CSII), a weighted composite score that combines SDG breadth and depth, implementation maturity, equity design quality, and risk into a single performance indicator. The second is the Equity Design Quality (EDQ) Index, a rubric with five sub-dimensions that operationalizes the concept of equity-by-design for digital urban governance contexts. I then test, through frequentist OLS with a three-layer identification strategy and Monte Carlo-based measurement uncertainty analysis, whether EDQ scores are associated with CSII performance on an outcome measure that excludes equity by formula, ensuring that any observed association reflects genuine predictive power rather than mere formula replication.

The findings are generally consistent and methodologically robust. Use cases with higher-quality equity design outperform those with lower-quality equity on the non-equity sustainability composite. All Monte Carlo simulations show a positive coefficient under plausible single-rater measurement-error perturbations. The result is not conclusive (N = 44 is a proof-of-concept sample, not a causal estimate) but it is the first empirical evidence that equity-by-design acts as a sustainability multiplier in the Citiverse. It also shows that excluding equity from the design phase has measurable effects on SDG performance units.

The contribution of this paper is threefold. Empirically, I provide the first quantitative test of the equity-by-design hypothesis in a Citiverse context, using an original dataset and a pre-specified specification that distinguishes between mechanical and genuine predictive relationships. Methodologically, I introduce the CSII and EDQ instruments as replicable, extensible tools for future cross-jurisdictional research. Theoretically, I expand the Gender Balance System Design (GBSD) framework developed by the author and the Technology Equity Nexus Framework (TENF) into the urban technology governance context, arguing that equity-by-design is not merely an ethical imperative but a structural predictor of sustainable development impact in AI-mediated city systems.

This article is organized into five main sections. The section ‘Literature Review’ provides an overview of the relevant literature, while the next section presents the ‘Methodology’. The section ‘Descriptive Data’ provides an overview of the data, followed by the ‘Interpretation and Implications’ section.

Literature review

This paper intersects four distinct bodies of literature: the emerging research on Citiverse and virtual urban governance; the SDG-alignment studies in smart cities and digital twins; the expanding empirical work on algorithmic bias and equity in automated urban systems; and the behavioral and systems-design theories underpinning the equity-by-design concept. I examine each area individually before synthesizing them into the theoretical framework— Equity by Design as a Performance Multiplier —that guides this paper.

The citiverse: technology convergence and the governance challenge

The term metaverse, coined by Neal Stephenson in his 1992 novel Snow Crash and operationalized through successive waves of platform development from Second Life to Fortnite to Meta Horizon, has acquired substantive institutional meaning in the context of urban governance (Stephenson and Crash; Dwivedi et al., 2022; Lee et al., 2021; Mystakidis, 2022; Wang et al., 2023). The Citiverse, as conceptualized by the ITU’s Global Initiative on AI and Virtual Worlds, denotes the application of metaverse-class technologies to city systems: not entertainment or commerce, but infrastructure management, emergency planning, public participation, climate adaptation, and equitable service delivery (Preductuve Power, 2019; Allam et al., 2022; Elias Bibri and Allam, 2022; Elias Bibri et al., 2022). Cecchi-Dimeglio (2025a) argues that this distinction is fundamental to understanding the governance challenge: consumer metaverses compete for voluntary engagement; Citiverse deployments govern compulsory relationships between citizens and the systems that allocate housing, mobility, safety, and opportunity (Cecchi-Dimeglio, 2025a).

The technology families that constitute the Citiverse are not individually new. Digital twins, virtual replicas of physical systems updated with real-time sensor data, were first conceived in the 1960s by NASA and first deployed in aerospace manufacturing (Shezad Dihan et al., 2024; Tuegel et al., 2011; Grieves et al., 2017). Building Information Modelling (BIM) has been a standard in construction since the mid-1980s. What is new is their convergence within a common governance architecture, enabled by the maturation of IoT connectivity, cloud computing, and generative AI. This convergence allows urban systems to be monitored, simulated, and governed at a previously unachievable level of granularity.

A systematic review of urban digital twins identifies five enabling capabilities that define the current generation: real-time bidirectional data linkage between physical and virtual assets; integration of heterogeneous data sources across city systems; simulation and predictive modelling; multi-stakeholder collaboration support; and decision-making at machine speed (Lei et al., 2023; Batty, 2018; Elias Bibri and Krogstie, 2017).

The governance implications of these capabilities are substantial. Virtual Singapore is a digital twin of the entire country and has been used to simulate emergency evacuation scenarios for millions of residents. Rotterdam’s digital twin coordinates tidal flood defenses, real-estate development approvals, and green infrastructure planning. Dubai’s Citiverse implementation integrates data from several dozen city services into a unified decision-support environment.

What these deployments share is a design logic focused on system efficiency, resource optimization, and infrastructure resilience. However, they do not yet systematically incorporate a design approach centered on who benefits from those efficiencies, whose data is collected through monitoring systems, and which populations are prioritized when using simulation models to allocate limited city resources. Nobel Prize winner Elinor Ostrom’s work on governing shared resources provides a valuable correction: sustainable systems are not solely those optimized by centralized technical logic, but those built around collective rules, local knowledge, and active community participation (Ostrom, 2005; Ostrom, 1990). Her insight that commons governance succeeds when users directly shape the governing rules challenges the assumption that data-driven optimization is inherently neutral—an assumption that underscores the core risk of shifting toward a digital city.

In Web3 environments, which, like the Citiverse, rely on decentralized data flows, algorithmic decision-making, and immersive digital experiences, ethical considerations are not just secondary issues to address after launch. They are fundamental elements of system design that must be integrated from the very beginning. Ostrom would see this not as a technical problem waiting for a technical fix, but as a governance challenge requiring intentional institutional design from the start.

What the Taxonomy currently lacks is a systematic way to evaluate whether equity considerations are integrated into the design of individual use cases; specifically, whether, in Ostrom’s terms, the communities most impacted have a meaningful role in shaping the rules that govern them. The Citiverse Taxonomy, created by the ITU, UNICC and Digital Dubai, Global Initiative on AI and Virtual Worlds, in collaboration with more than 70 international partners, records what technologies are being used, which SDGs they align with, the risks they pose, and their maturity stage. However, documenting what exists is not the same as assessing whether those systems are equitably designed. This is the measurement gap this paper aims to address.

Smart cities, digital twins and the SDG agenda

The integration of smart city technologies with the Sustainable Development Goals (SDGs) framework has been an active area of research since the 2015 adoption of the 2030 Agenda (UN-Habitat, 2021). ITU’s working definition of a sustainable smart city: “an innovative city that uses ICT and other tools to enhance quality of life, improve urban operation efficiency, and boost competitiveness” while ensuring “it meets the needs of current and future generations across economic, social, and environmental aspects”. The definition is intentionally broad, covering SDG areas from poverty reduction (SDG 1) and health (SDG 3) to climate action (SDG 13) and partnerships (SDG 17) (International Telecommunication Union, 2026). In practice, however, the literature consistently focuses on a narrower set of issues.

Urban Digital Twin (UDT) systems are mainly used in environmental sustainability areas, including power grids, energy demand, renewable energy, mobility, road infrastructure, water supply, flooding, and climate science. The focus falls most strongly on SDG 11 (Sustainable Cities and Communities), followed by SDG 3 (Good Health and Wellbeing) and SDG 9 (Industry, Innovation, and Infrastructure). Meanwhile, SDG 1 (No Poverty) and SDG 10 (Reduced Inequalities) receive less attention.

The core frameworks that shaped smart city governance reveal the same pattern at the design level. Often, smart city projects are centered on operational-efficiency goals: smart economy, smart mobility, smart environment, smart governance, smart people, and smart living. Social equity aspects of urban life, such as access to basic services, poverty reduction, and equitable distribution, tend to be seen as overall goals rather than main criteria for designing and measuring progress (Albino et al., 2015; Hollands, 2008; Meijer et al., 2016).

Urban digital twin applications have developed primarily around what is quantifiable and technically tractable, such as energy flows, traffic volumes, structural loads, atmospheric conditions, and physical infrastructure more broadly. The socio-economic and cultural dimensions of city life, who accesses services, who bears the burden of urban risks, and who participates in decision-making, are considerably harder to represent in computational models and have received comparatively little attention in current deployments (International Telecommunication Union, 2025; Kitchin, 2014; Cardullo and Kitchin, 2019). This reflects a broader pattern in the field: digital representations of cities tend to reproduce the physical environment with increasing fidelity while the underlying political, economic, and social processes that shape urban life remain underrepresented.

This gap is not simply a technical limitation that more data or better sensors will resolve. The variables selected for modelling, and those excluded, reflect choices about what counts as relevant and measurable, choices that carry distributional consequences (Sunstein, 2014; Thaler and Sunstein, 2009). When transport optimization, energy management, or land-use planning are governed by systems that model physical flows but not socio-economic access, the resulting decisions embed assumptions about whose conditions and whose constraints are treated as design inputs. The scholarship on algorithmic fairness and data-driven governance more broadly has documented how these gaps can systematically disadvantage already-marginalized populations, even in the absence of any explicit discriminatory intent (Cecchi-Dimeglio, 2023a; Barocas et al., 2016; Eubanks, 2018; O’Neil, 2016).

The Citiverse Taxonomy, with its explicit SDG tagging of each use case, provides a natural empirical setting to test whether equity design quality predicts the breadth and depth of SDG co-production.

The algorithmic equity gap: evidence from adjacent literature

The concern that data-driven systems can reproduce or intensify social inequalities is well-established in the literature. A foundational insight in this body of work is that algorithmic systems trained on historical data tend to inherit the structural biases embedded in that data, producing outcomes that disadvantage historically excluded groups even in the absence of explicit discriminatory intent (Rosen et al., 2021; Banaji et al., 2016; Umoja Noble, 2018). Discrimination, in this framing, is frequently an unintentional emergent property of algorithmic design rather than a conscious choice, a pattern that has since been documented across a wide range of system applications, from predictive policing to credit scoring to housing allocation (Bohnet, 2016).

A related line of argument holds that algorithmic transparency, while necessary, is not sufficient for equitable urban governance (Safransky, 2020; Ananny and Crawford, 2018; Pasquale, 2015). Transparency, the disclosure of what data a system uses and how it generates outputs, does not automatically produce accountability: the capacity of affected populations to contest, correct, or override algorithmically generated decisions. In contexts where a digital twin’s simulation output may directly determine which neighborhood receives flood infrastructure investment, which bus routes are maintained, or which building permits are approved, the gap between transparency and accountability carries direct distributional consequences.

Research in organizational innovation suggests that the equity gap in technology-mediated systems is not a fixed structural feature (Cecchi-Dimeglio et al., 2026; Cecchi-Dimeglio, 2025b). It responds to targeted design interventions. Experimental work in this area has found that simplifying submission processes, replacing exclusionary language with inclusive terminology, and embedding active leadership engagement in innovation workflows each produced substantial increases in participation from underrepresented groups. The consistent mechanism across these findings is the same: equity outcomes improve when systems are explicitly designed to remove barriers, rather than assuming that general-purpose platforms will automatically produce equitable participation (Cecchi-Dimeglio, 2023b; Bohnet, 2016; Ostrom, 2005). This mechanism, which I term the equity-by-design effect, is the central object of inquiry in this paper. If designing for equity in an innovation platform increases both the diversity of contributions and the overall quality of outcomes, the question is whether an analogous effect is observable in Citiverse use cases. Specifically, it asks whether systems designed with equity as a primary criterion achieve better SDG outcomes not just on the equity-specific goals but across the full SDG portfolio. Preliminary evidence from the adjacent literature on public-sector digital transformation suggests that this generalization is plausible. Research on smart citizen participation has shown that when digital city systems actively recruit diverse engagement, through multi-channel outreach, accessible interfaces, and genuine co-design, they tend to generate higher-quality decision inputs, broader public trust, and more durable implementation outcomes (Shin et al., 2021; Arnstein, 1969). Similarly, cities that have embedded equity and transparency criteria into their AI governance frameworks, including explicit procurement standards, algorithm registers, and participatory oversight, have developed more robust mechanisms for detecting and contesting algorithmic bias, and report stronger public trust in automated decision-making (OECD, 2026).

This literature provides fertile ground to support the quantitative CSII-EDQ correlation this paper aims to examine.

Fairness by design as a performance multiplier

The theoretical foundation for equity-by-design as a performance predictor is the Gender Balance System Design (GBSD) framework, originally created by the author within the context of legal professional organizational reform (Cecchi-Dimeglio, 2025a; Cecchi-Dimeglio, 2023c; Cecchi-Dimeglio, 2023a; Cecchi-Dimeglio, 2025c; Cecchi-Dimeglio, 2019; Simon, 1969; Meadows, 2008; Senge, 1990). GBSD’s main argument is that fairness and equity are not a leftover correction applied after a system produces its results, but structural design elements that influence output quality from the beginning. Organizations that incorporate fairness and equity into their procedures, incentive structures, and evaluation criteria—rather than treating them as secondary constraints on efficiency-driven processes—consistently outperform those that do not. GBSD was developed for human organizations, where fairness is primarily procedural.

In the Citiverse, fairness, equity, and accessibility are also architectural: which populations are represented in AI training data, which connectivity standards determine access, which surveillance capabilities are embedded in smart city infrastructure, and for whose benefit.

The Technology Equity Nexus Framework (TENF) expands GBSD and the author’s decision intelligence framework to include this architectural aspect of technology. TENF’s main idea is that the impact of designing for equity is stronger in technology-heavy deployments: a use case that combines high-quality equity design with multiple advanced technology types yields a greater performance benefit than one with the same equity score but less advanced deployment. Technology richness offers more chances for positive equity design, more data streams that can be broken down, more service channels that can be made accessible, more decision points where bias can be spotted and fixed. At the same time, it creates more risks for negative equity outcomes if those choices are not intentional.

TENF asserts that the interaction between technology richness and the quality of equity design determines whether advanced deployments serve as equity amplifiers or suppressors. A technologically advanced use case without designated beneficiary populations, accessible design, or SDG 10 targeting risks prioritizing measurable outcomes like traffic throughput, energy efficiency, and property values, while neglecting less tangible ones such as accessibility for disabled residents, displacement pressures on low-income communities, and algorithmic profiling of marginalized groups. Conversely, a use case with high-quality equity design has both the capacity to recognize these risks and the structural incentives to address them.

Data

The empirical foundation of this study is the leading United Nations initiative in this domain—the ITU/UNICC/Digital Dubai Global Initiative on AI and Virtual Worlds, whose Citiverse Use Case Taxonomy (2025) provides a systematic, expert-validated inventory of 44 smart-city and Citiverse use cases spanning five urban governance domains: Urban Planning (n = 8), City Administration (n = 7), Transport (n = 8), Public Safety (n = 8), and Economic Development, Education, and Tourism (n = 13), with five cross-cutting use cases.

The unit of analysis is the individual use case, a discrete, named deployment or conceptual application of Citiverse technologies (e.g., ‘Traffic Optimization and Simulation,’ ‘Virtual Town Halls and Participatory Planning,’ ‘Multimodal Emergency Evacuation Simulation’). For each use case the Taxonomy records: (a) binary technology flags for nine technology types, Digital Twins (DT), Artificial Intelligence (AI), Augmented Reality (AR), Virtual Reality (VR), Mixed Reality (MR), Internet of Things (IoT), Metaverse environments, Blockchain, and Building Information Modelling (BIM); (b) binary coverage indicators for eleven Sustainable Development Goals (SDGs 1, 3, 4, 8, 9, 10, 11, 12, 13, 16, 17); (c) expert-rated scores (1–10) for Scale, Impact, and Feasibility; (d) an SDG Alignment score (1–10); (e) an implementation horizon classification (H1 = deployed, H2 = pilot, H3 = discovery/concept); and (f) real-world implementation evidence (cities, programs, evaluation sources). Implementation horizons play a structural role in the analysis. Implementation horizons are classified as H1 (deployed), H2 (pilot), and H3 (discovery). The horizon distribution matters because implementation maturity is both a genuine signal of design quality and a potential confound in measuring sustainability impact (Rotberg, 1993).

Citiverse Sustainability Impact Index (CSII)

The Citiverse Sustainability Impact Index (CSII) is a composite, normalized index measuring the expected sustainable development performance of each use case.

It is computed as:

The five components are defined as follows. B_i (SDG breadth) = n_SDGs/17, where n_SDGs is the raw count of SDGs covered by the use case. D_raw (SDG alignment depth) = the Taxonomy’s expert-rated SDG Alignment score divided by 10, a normalized measure of the quality, not just the quantity, of SDG linkages. M_i (implementation maturity multiplier) = 1.0 for H1, 0.70 for H2, and 0.40 for H3, reflecting the expectation that deployed use cases generate more reliable SDG contributions than pilot or conceptual ones. EDQ/10 (Equity Design Quality) is defined and measured later on. R_i (risk discount) = the mean of four ordinal risk scores, privacy/surveillance risk, social acceptance risk, data privacy risk, and financial/operational risk, where each is coded 0 (no risk), 0.5 (moderate), or 1.0 (high). The R_i term enters with a negative weight of −0.20.

The equal weighting of the four positive components (0.25 each) is selected because, at the proof-of-concept stage, no validated evidence base exists from which differential weights could be estimated, and any unequal scheme would import untested assumptions about the relative importance of breadth, depth, maturity, and equity design into the index itself. Equal weighting is the most transparent, least assumption-laden default and is standard practice for first-generation composite indicators. The risk term carries a slightly smaller weight (−0.20) so that risk attenuates, but cannot single-handedly dominate, the composite. Sensitivity to these choices is addressed directly by the three-specification design: Specification 3, the primary outcome, removes the equity component entirely, and the stability of the EDQ coefficient across specifications indicates that the substantive findings do not hinge on the baseline weights. Deriving weights empirically, through confirmatory factor analysis or structured expert elicitation, is a priority for future iterations once a larger sample is available.

Equity design quality index

The Equity Design Quality index operationalizes the concept of ‘equity-by-design’ as a structured, multi-dimensional measurement instrument. EDQ is scored on a 0–10 scale across five sub-dimensions, each rated 0–2: (1) Named Population: whether the use case explicitly names specific equity-relevant beneficiary groups (0 = none; 1 = generic populations; 2 = specific named groups such as disabled residents, low-income households, or linguistic minorities); (2) Design Centrality: whether equity is central to the use case’s core purpose (0 = incidental; 1 = present in impact statements; 2 = primary objective); (3) Accessibility Mechanism: whether specific accessibility features or mechanisms are embedded in the technical design (0 = none; 1 = mentioned; 2 = specific, named feature); (4) Reach: the geographic and demographic scope of equity impact (0 = local/single group; 1 = city-scale; 2 = global or cross-national); (5) SDG 10 Alignment: whether reduction of inequalities is a primary or addressed goal of the use case (0 = not addressed; 1 = addressed; 2 = primary SDG target). An EDQ total of seven or above is defined as a high equity design.

The coding procedure and decision rules were as follows. Each of the 44 use cases was scored on the five sub-dimensions using only the published Taxonomy documentation—use-case descriptions, impact statements, SDG tags, and implementation evidence—applied against the pre-specified 0–2 anchors above. Three decision rules governed ambiguous cases: (1) scores were assigned on documented design features, not inferred intentions, so a use case received credit only for what its documentation states; (2) where documentation was ambiguous between two adjacent anchor levels, the lower score was assigned, a conservative rule that biases against, rather than toward, the paper’s hypothesis; and (3) the five sub-dimensions were scored independently of one another to limit halo effects across dimensions.

The five sub-dimensions are weighted equally for the same reason the CSII components are: absent prior empirical evidence that any one dimension contributes more to equity outcomes than the others, differential weights would encode untested theoretical commitments. The dimensions were specified ex ante as conceptually distinct facets of equity-by-design, and equal weighting treats them as such pending empirical validation. Because scoring was conducted by a single rater in this iteration, two safeguards address potential subjectivity. First, within-coder consistency was assessed by re-scoring the sample and is reported in the accompanying dataset. Second, the Monte Carlo analysis reported below perturbs every EDQ score by up to ±1 point—larger than plausible within-coder drift—and the directional finding survives all 1,000 draws. Multi-rater scoring with formal inter-rater reliability statistics, such as Krippendorff’s alpha, remains the appropriate standard for confirmatory work and is planned for the next Taxonomy iteration.

Descriptive Data

Table 1 presents descriptive statistics by domain. Three patterns deserve attention when reading Table 1: every domain’s mean EDQ falls well below the high-equity threshold of 7; the gap between CSII Spec. 1 and Spec. 3 within each row indicates how much the equity component contributes to composite scores; and the horizon column reveals the uneven distribution of deployment maturity that drives much of the regression analysis that follows. Figure 1 shows the mean CSII by domain and specification. Figure 2 shows EDQ distribution by domain.

Figure 1 shows mean CSII scores by domain and specification, allowing comparison of each domain’s performance across the three parallel composite formulations; the gap between Spec. 1, which includes equity, and Spec. 3, which excludes it, indicates how much of a domain’s standing depends on the equity component. Transport stands out as the consistently strongest performer across all three specifications (Spec. 3 mean = 0.475), because it has no H3 discovery-stage cases; all eight Transport use cases are at H1 or H2, so the maturity multiplier works in their favor. Transport’s lead, in other words, reflects implementation maturity rather than equity design. Urban Planning is the weakest domain under Spec. 1 (0.356), but the gap narrows under Spec. 3, suggesting that Urban Planning’s below-average score partly reflects lower equity design quality rather than weak sustainability alignment per se. The spread between Spec. 1 and Spec. 3 is most visible in Economic/Education/Tourism, where removing EDQ from the composite actually lowers the domain’s relative standing, consistent with it having the highest mean EDQ in the sample and benefiting most from the equity weight in Spec. 1.

Figure 2 shows the EDQ score distribution by urban governance domain, revealing within-domain spread, individual use case positions, and how each domain sits relative to the high-equity threshold of 7, marked by the dashed line. Economic/Education/Tourism has the highest mean EDQ (4.5) but also the widest distribution (σ = 2.96), meaning the domain contains both the most equity-forward use cases in the entire taxonomy and some with very low equity design. Urban Planning is tightly clustered at the low end (μ = 2.5, σ = 1.77); there is little variance, and no use case reaches the high-equity threshold of 7, which points to a structural gap rather than individual outliers. Transport and Public Safety sit in the middle with moderate means but meaningful variance, suggesting those domains have pockets of equity-conscious design alongside purely technical deployments. Only seven of 44 use cases (16%) cross the EDQ ≥7 threshold, and they are almost entirely concentrated in Economic/Education/Tourism; a finding with direct policy implications for where equity-by-design capacity building is most urgently needed.

Measurement validation

Prior to the primary regression analysis, two pre-specified diagnostic procedures were conducted: a Monte Carlo measurement-uncertainty analysis and a halo-effect diagnostic of the CSII component structure.

Monte Carlo measurement uncertainty analysis

To assess the sensitivity of the directional finding to plausible single-rater measurement error in EDQ scoring, all 44 EDQ scores were simultaneously perturbed by independent draws from a uniform distribution U(−1, +1), truncated to preserve the valid scale range [0, 10]. The bivariate OLS model (outcome: CSII Specification 3, the no-equity composite) was re-estimated on each of 1,000 perturbed datasets. Across all 1,000 simulations, the estimated coefficient on EDQ/10 remained strictly positive (β > 0 in 100% of draws). The median estimated coefficient was 0.1135 per EDQ/10 unit (0.0114 per EDQ point), with a 90th-percentile simulation interval of [0.0910, 0.1388]. The null value of zero falls entirely outside the empirical simulation distribution (Appendix Figure 7). The directional finding is therefore classified as sign-stable under the perturbation regime tested. Plausible single-rater coding variability is insufficient to reverse the positive association between equity design quality and non-equity sustainability performance. It should be noted that this procedure tests robustness to EDQ measurement noise specifically; sensitivity to CSII specification and model structure is addressed separately through the three-specification comparison reported in Table 2.

Halo effect diagnostic

A potential source of construct contamination in the CSII is collinearity between its SDG breadth component (B_i, the raw count of SDGs addressed, normalized to [0, 1]) and its SDG alignment depth component (D_raw, an independent expert-rated alignment score). If these two components were strongly correlated, their joint inclusion in the composite would effectively double-weight a single underlying dimension, undermining the formula’s additive structure. Pearson correlation between B_i and D_raw across the full sample (N = 44) yields r = 0.300 (p = 0.048), indicating a moderate, statistically significant, but not severe association. A use case’s raw SDG coverage count is not a reliable proxy for expert-rated alignment depth, confirming that the two components capture partially independent aspects of sustainability integration. Both are therefore retained in the CSII as non-redundant contributors.

Table 2 presents bivariate and multivariate OLS regression results for CSII Spec. 3, showing EDQ’s independent predictive contribution after controlling for implementation horizon, risk, domain, and technology count. Figure 3 plots EDQ against CSII Spec. 3 for all 44 use cases with the OLS fit and 95% confidence band. The attenuation of the EDQ coefficient across the two columns of Table 2 (from 0.121 to 0.066), alongside the jump in R 2 (from 0.075 to 0.891), shows how much of the raw association runs through implementation maturity and risk—and how much remains as EDQ’s independent contribution.

Table 2 presents bivariate and multivariate OLS estimates for CSII Specification 3 as the dependent variable. In the bivariate specification, the EDQ coefficient is positive and directionally consistent with the hypothesized relationship (β = 0.121, SE = 0.065, p = 0.071), though it falls short of the conventional α = 0.05 significance threshold. This result should be interpreted as suggestive rather than confirmatory, and is reported as the transparent baseline prior to the introduction of structural controls.

The multivariate specification clarifies the picture considerably. After controlling for implementation horizon, deployment domain, technology count, and risk burden, the EDQ/10 coefficient attains statistical significance (β = 0.066, SE = 0.026, p = 0.014), indicating that equity design quality retains independent predictive power over non-equity sustainability performance once the dominant structural covariates are held constant. This finding supports Hypothesis 1: higher EDQ is associated with higher CSII Spec. 3 scores net of confounds, and that association is not an artifact of the equity component’s presence in the CSII formula, from which it was excluded by construction.

The horizon dummies account for the majority of the explained variance. Pilot-phase cases (H2) score 0.112 points below deployed cases (H1) on CSII Spec. 3 (p < 0.001), and discovery-phase cases (H3) score 0.233 points below the H1 reference (p < 0.001). The risk composite is also significantly negative (β = −0.243, p = 0.013), consistent with the theoretical expectation that higher implementation risk attenuates realized sustainability impact. Domain dummies and technology count are jointly non-significant in the full specification.

The increase in R 2 from 0.075 in the bivariate model to 0.891 in the multivariate model (adjusted R 2 = 0.866) is the key diagnostic. It establishes that the implementation lifecycle position is the primary determinant of a use case’s CSII Spec. 3 score, with EDQ contributing a statistically significant but substantively modest increment above that structural baseline. The partial contribution of equity design quality, positive, significant, and robust to the inclusion of nine covariates, constitutes the paper’s central empirical finding, understood explicitly as a directional signal in a proof-of-concept sample rather than a precise causal estimate.

Figure 3 shows the bivariate OLS regression of EDQ against CSII Spec. 3 across all 44 use cases, with the fitted line and 95% confidence band illustrating the direction and uncertainty of the relationship. The positive slope is visible, but the cloud is wide; individual use cases with the same EDQ score span a CSII range of roughly 0.15–0.20 points, reflecting the horizon and risk variance that the bivariate model cannot see. The 95% confidence band widens noticeably at both ends of the EDQ scale, a direct consequence of the small number of use cases at the extremes: only 7 cases score EDQ ≥7, so the right-hand side of the plot rests on a thin empirical base. The Transport cluster consistently sits above the regression line at moderate EDQ values, while several City Administration cases fall below it; both are consistent with the matched-pairs finding that domain-level implementation maturity can override equity design quality as a predictor of CSII performance at N = 44.

Robustness check

Three robustness tests were conducted to probe the stability of the main finding. Within-domain fixed effects yield β_EDQ = 0.137 (SE = 0.067, p = 0.045) on domain-demeaned values, confirming that the positive relationship between equity design quality and CSII Spec. 3 holds within domains and is not simply an artefact of between-domain variation. Restricting the sample to the n = 15 deployed (H1) use cases, where the maturity confound is absent by design, yields β_EDQ = 0.081 (SE = 0.044, p = 0.091); the direction remains positive, but precision is low given the reduced sample size, and this result is treated as directional evidence only. Finally, three within-domain matched pairs were examined on the primary outcome (CSII Spec. 3) (Figure 4).

The secondary (multiplicative) hypothesis H2 was tested by adding the EDQ × TECH_COUNT interaction term to the full model. The interaction coefficient is β = 0.048 (SE = 0.023, t = 2.06, p = 0.048), significant at α = 0.05. At the sample mean of 3.93 technology families, the marginal effect of a one-unit increase in EDQ/10 is −0.1331 + 0.0475 × 3.93 = 0.054 CSII points, substantially larger than the additive estimate. H2 is therefore supported conditionally: equity design quality and technology richness combine multiplicatively. Given N = 44 and 10 parameters (4.4 obs/param), this result is treated as directional; replication with a larger sample (N ≥ 90) will be required before stronger inferences can be drawn.

Figure 4 shows matched-pairs comparisons within three domains, tracking how a lower- and higher-EDQ use case differ on EDQ, CSII Spec. 1, and CSII Spec. 3 side by side. Because each panel holds domain constant, the deltas isolate design differences from sector effects. The comparison yields mixed results for the primary outcome. The Transport pair (TR-04 vs. TR-03, Δ = +0.06) supports the equity-by-design hypothesis, with the higher-EDQ use case outperforming on CSII Spec. 3. The Urban Planning pair (UP-01 vs. UP-02, Δ = −0.05) and the City Administration pair (CA-04 vs. CA-01, Δ = −0.17) run in the opposite direction, with the lower-EDQ case scoring higher. The City Administration reversal is the most pronounced and is largely explained by CA-01’s H1 deployment status; its full implementation maturity boosts CSII Spec. 3, independently of its lower equity design quality. Taken together, the 1-for-3 result is consistent with the regression finding that implementation horizon is a stronger predictor of CSII performance than EDQ at this sample size, and it underscores why the within-domain fixed effects and multivariate controls are necessary to isolate EDQ’s contribution.

Figure 5 shows CSII and EDQ by implementation horizon, tracing a maturity gradient from deployment (H1) to discovery (H3). Deployed use cases (H1, n = 15) score markedly higher on both CSII Spec. 1 (0.497) and Spec. 3 (0.550) than pilot cases (H2: 0.375/0.408) and discovery cases (H3: 0.250/0.261), with the gap between H1 and H3 exceeding 0.28 points on the primary outcome; more than three times the size of the EDQ coefficient in the multivariate regression. EDQ/10 follows the same gradient but with a shallower slope: deployed use cases average EDQ/10 = 0.400 versus 0.317 for discovery cases, suggesting that equity design quality tends to be more fully elaborated in mature deployments, though the relationship is not deterministic. The overlap in standard deviations across horizons confirms that implementation stage and equity design quality are correlated, but distinct dimensions; both matter, but at N = 44, maturity dominates.

Figure 6 shows SDG coverage rates by EDQ tier (high vs. low equity design quality) across all eleven SDGs addressed in the taxonomy; large gaps between the paired bars mark equity-driven divergence. SDG 11 (Sustainable Cities) dominates the coverage profile across both tiers, but with a notable inversion: low-EDQ use cases cover it at 81.1% versus 57.1% for high-EDQ cases, reflecting that SDG 11 is simply the baseline ambition of virtually every Citiverse deployment, regardless of equity intent. The sharpest equity-driven divergences appear in SDG 4 (Education: 71.4% high vs. 8.1% low), SDG 10 (Equality: 42.9% vs. 2.7%), and SDG 1 (Poverty: 28.6% vs. 0%); SDGs that require deliberate targeting of marginalized populations and are almost absent from use cases with weak equity design. Conversely, SDG 9 (Infrastructure) is covered exclusively by low-EDQ use cases (24.3% vs. 0%), consistent with technically driven deployments that prioritize connectivity and efficiency over distributional outcomes. SDG 13 (Climate) and SDG 16 (Justice) show moderately high-EDQ premiums (42.9% vs. 27.0% and 42.9% vs. 24.3%, respectively), suggesting partial but not systematic alignment between equity-conscious and sustainability-oriented design. The overall picture is one of SDG concentration risk: the taxonomy is heavily anchored to SDG 11, while SDGs most directly linked to inequality reduction remain structurally underrepresented.

Interpretation and implications

The regression results offer conditional empirical support for the central hypothesis. In the bivariate specification, the EDQ coefficient is positive and directionally consistent with the hypothesized relationship (β = 0.121, SE = 0.065, p = 0.071), falling marginally short of the conventional α = 0.05 significance threshold, consistent with the limited statistical power at N = 44. In the nine-regressor multivariate model, the EDQ/10 coefficient strengthens and attains statistical significance (β = 0.066, SE = 0.026, p = 0.014, R 2 = 0.891, adjusted R 2 = 0.866), indicating that equity design quality carries independent predictive power over non-equity sustainability performance once implementation horizon, deployment domain, technology richness, and risk burden are held constant. Estimated on an outcome measure from which the equity component was excluded by construction, this association reflects a genuine cross-construct relationship rather than formula overlap. The Monte Carlo perturbation analysis corroborates the directional finding: across 1,000 simulations introducing plausible single-rater measurement error, the EDQ coefficient remained strictly positive in all draws (median β = 0.1135 per EDQ/10 unit, 90th-percentile interval [0.0910, 0.1388]), consistent with the theoretically expected attenuation under classical measurement error given the estimated reliability ratio of λ = 0.943.

The dominant structural finding, however, is the role of implementation horizon (Figure 5). Discovery-phase cases (H3) score 0.233 points below deployed cases on CSII Specification three (p < 0.001), and pilot-phase cases (H2) score 0.112 points below the H1 reference (p < 0.001). Together with the significant negative risk coefficient (β = −0.243, p = 0.013), these results establish that lifecycle position and risk burden jointly account for the overwhelming share of explained variance, with R 2 increasing from 0.075 in the bivariate model to 0.891 in the full specification (ΔR 2 = 0.816). Domain dummies and technology count are jointly non-significant after these controls are introduced. EDQ’s contribution is real and statistically reliable, but it operates within a dataset currently dominated by maturity-driven variation. This ordering of effects is itself a substantive finding: as the proportion of deployed use cases grows across successive Taxonomy iterations and maturity-driven variance narrows, equity design quality is expected to account for a relatively larger share of the remaining performance differences between use cases. The current estimate is therefore more plausibly a lower bound on EDQ’s long-run predictive relevance than a ceiling.

Three implications follow directly from these results. First, the significance of the EDQ coefficient in the full model, net of domain fixed effects, indicates that the equity design gap documented in the descriptive analysis (Figures 2, 6) is not fully attributable to domain-level structural characteristics. Of the seven use cases that cross the high-equity threshold (EDQ ≥7, representing 15.9% of the full sample), five belong to the Economic, Education, and Tourism domain, one to City Administration, and one to Transport; Urban Planning and Public Safety contribute none. This concentration reflects design choices that are, in principle, independent of sector and therefore amenable to targeted intervention. Second, the large and highly significant horizon coefficients suggest that the window for embedding equity-by-design principles is concentrated at the pre-deployment stage. Use cases currently in pilot or discovery phases—which together constitute 65.9% of the sample, have not yet fixed their design parameters irreversibly, making the present moment a high-leverage point for equity-oriented policy intervention. Third, the significant negative risk coefficient indicates that high-risk domains face structural constraints on sustainability performance that equity design quality alone cannot offset, implying that risk mitigation and equity-by-design should be treated as complementary policy levers rather than competing investment priorities.

This paper set out to test whether equity design quality is independently associated with non-equity sustainability performance in Citiverse deployments, and to develop the measurement infrastructure necessary to conduct that test on a systematic, international evidence base. The results are affirmative on the first question and generative on the second. The positive, significant, and measurement-robust association between EDQ and CSII Specification three in the full model provides proof-of-concept evidence that equity-by-design functions as a genuine performance signal in this context rather than a definitional artifact. The measurement instruments introduced, both the CSII and the EDQ rubric, are transparent, internally documented, and designed to scale with future data collection.

These findings extend two theoretical frameworks. Within the Gender Balance System Design tradition and the Technology Equity Nexus Framework, the results provide the first quantitative evidence that equity-by-design principles, when operationalized at the use-case level of urban technology governance, are associated with measurably broader sustainability outcomes across dimensions that are formally independent of equity by construction. This operationalizes a core theoretical claim, that equity is a structural performance signal rather than merely an ethical constraint on system design, in a governance context to which neither framework had previously been applied empirically. The modest but statistically significant effect size is consistent with a multiplier mechanism at early proof-of-concept scale, where the compounding benefits of equity-embedded design across the full implementation lifecycle have not yet materialized.

The results also contribute to the smart city and digital twin literature, where equity considerations have been increasingly theorized but rarely measured quantitatively at the use-case level. The finding that 77.3% of Citiverse use cases address SDG 11 while only 9.1% address SDG 10, and that this distributional asymmetry is associated with measurable differences in non-equity sustainability performance, provides an empirical foundation for arguments that have remained largely normative in the existing literature.

For policymakers and urban governance stakeholders, the EDQ rubric is directly operational, and its leverage is greatest in the pilot- and discovery-stage (H2/H3) systems that constitute roughly two-thirds of the current Taxonomy.

Three concrete pathways follow. First, procurement and funding: cities and national programs can embed minimum EDQ thresholds, or require scoring against the five sub-dimensions, in smart-city tender specifications and grant calls, making named beneficiary populations, accessibility mechanisms, and SDG 10 alignment entry conditions rather than post hoc aspirations. Second, stage-gate review: because design parameters are not yet fixed in H2/H3 systems, program owners can score use cases at each lifecycle gate and require documented improvement on the weakest sub-dimension before advancement; the rubric’s 0–2 anchors make this a lightweight, auditable exercise. Third, portfolio monitoring: standards bodies such as the ITU can incorporate EDQ scoring into successive Taxonomy iterations, turning the equity design gap into a tracked indicator and directing capacity-building toward the domains identified here as structurally weakest, Urban Planning and Public Safety. None of these applications requires new data infrastructure; each relies on documentation that mature procurement processes already generate.

Several limitations constrain the strength of the conclusions and define the priorities for subsequent research. The sample of 44 use cases, while constituting the most systematically documented international evidence base on Citiverse deployment currently available, yields 4.9 observations per regression parameter in the full specification, which is below the conventionally required threshold of 10 for stable multivariate inference.

All results should therefore be interpreted as directional estimates with wide uncertainty bounds rather than precise causal quantities. The CSII component weights are assumed equal in the baseline specification and sensitivity-tested across three formulations, but are not empirically derived; independent weight estimation via confirmatory factor analysis or expert elicitation awaits a larger dataset. The EDQ rubric was applied by a single researcher in this iteration, and the inter-rater reliability reported in the accompanying dataset reflects within-coder consistency rather than externally validated agreement between independent coders. This distinction bears directly on the inferential weight that can be placed on the EDQ coefficient and should be acknowledged wherever the results are cited.

Finally, the cross-sectional design precludes causal inference. The positive association between EDQ and CSII Specification three is consistent with equity-by-design functioning as a sustainability performance signal, but it is equally consistent with reverse causation or omitted variable confounding at the jurisdiction or program level. Addressing this identification problem rigorously requires either longitudinal data tracking use cases from design through deployment, or natural experiments arising from policy changes that exogenously shift equity design requirements.

The Citiverse is now being designed, and the equity architecture embedded in its foundational use cases will shape the distribution of its sustainability benefits for decades. This paper provides the first quantitative evidence that equity design quality is associated with non-equity sustainability performance in this context, using an original dataset of 44 internationally documented use cases, two new measurement instruments, and an analytical design that separates genuine predictive signal from formula overlap by construction. The finding is modest in magnitude, EDQ accounts for a small share of variance in a dataset currently dominated by deployment maturity, but it is positive, statistically significant after full controls, directionally stable under plausible measurement error, and estimated on a performance outcome from which equity was excluded. It is also precise about what it does not establish: causality requires longitudinal data, and precision requires a larger sample.

The distributional finding may carry the most immediate policy relevance. Only seven of 44 documented Citiverse use cases cross the high-equity design threshold, and they are concentrated in a single domain. Urban Planning and Public Safety, which together encompass 16 of the 44 use cases, contribute none. The evidence presented here suggests that this concentration reflects a measurable design gap, not merely a values gap, with quantifiable consequences for sustainability performance. Equity-by-design is a design decision made, or forgone, at the point of conception; it cannot be reliably retrofitted into systems already operating at scale.

The methodological contribution of this paper (the CSII, the EDQ rubric, the three-specification sensitivity design) is intended to provide a durable infrastructure for the research program that follows. As the ITU Global Initiative on Virtual Worlds and AI expands its documentation of Citiverse deployments and as national and municipal programs generate larger and more diverse portfolios of use cases, the framework developed here provides a replicable basis for tracking whether the equity design gap is narrowing, stable, or widening over time. Establishing that trajectory is the next empirical task the field needs to undertake.