Cultural Infrastructure Accessibility Score (CIAS)

• CIAS is a multi-dimensional metric defining access to cultural venues through proximity, opportunity, and quality measures.
• The methodology integrates spatial tessellation, travel-time computation, and normalized venue ratings to generate an aggregate accessibility score.
• Empirical applications reveal spatial gradients and social disparities, guiding urban policy and resource allocation to address cultural deserts.

The Cultural Infrastructure Accessibility Score (CIAS) is a rigorously defined, multi-dimensional metric devised to capture the spatial, modal, and qualitative dimensions of access to cultural venues—including libraries, museums, theaters, galleries, and monuments. CIAS formalizes accessibility in relation to proximity, opportunity, and value, and is designed to support comparative, policy-relevant, and equity-focused cultural infrastructure planning at urban and metropolitan scales (Bruno et al., 15 Sep 2025, Verma et al., 7 Apr 2024, Pranto et al., 8 Nov 2025). CIAS implementations reflect both the structural complexity of cultural access and the social disparities inherent in urban layouts.

1. Theoretical Underpinnings and Dimensions

CIAS emerges from the critique of mono-dimensional accessibility metrics, extending prior frameworks to address the interplay between three principal dimensions:

1. Proximity Access ($A^P$): Quantifies active, short-range, typically pedestrian access to nearby cultural venues. Proximity is operationalized via an exponential distance-decay function:

$f^P(d_{ij}) = \exp(-d_{ij}/d_0)$

where $d_{ij}$ is pedestrian distance and $d_0$ a decay constant (e.g., $d_0=1$ km). The summed, possibly weighted, proximity score assigns maximal value to venues within a few minutes' walk, with sharply diminishing influence at greater distances.

2. Opportunity Access ($A^O$): Reflects network-based access to relevant non-local assets via longer-range, multimodal travel—commonly transit or cycling. The form proposed is:

$$f^O(t_{ij}) = N\,\exp(-\alpha T/t_{ij})\exp(-\beta t_{ij}/T)$$

with $t_{ij}$ being the mode-specific travel time, $T$ a typical travel-time budget, and $N$ a normalizing constant.

3. Value Access ($A^V$): Encapsulates the quality or perceived desirability of reachable venues, incorporating non-spatial attributes such as visitor ratings or expert scores. The value-access score takes the form:

$A^V_i = \sum_j v_j \exp(-t_{ij}/\tau_V)$

where $v_j$ is a normalized quality score and $\tau_V$ represents the willingness-to-travel for quality.

These axes capture distinct, policy-relevant notions: convenience (proximity), coverage (opportunity), and experiential worth (value), avoiding reductionist aggregation (Bruno et al., 15 Sep 2025).

2. Methodologies for CIAS Construction

CIAS can be implemented at varying geographic and analytic resolutions—individual grid cells, census tracts, or origin zones—depending on application requirements (Pranto et al., 8 Nov 2025, Verma et al., 7 Apr 2024). The typical workflow is as follows:

• Spatial Tessellation: Define a fine-grained regular grid (e.g., 500×500 points citywide) or use census-defined polygons.
• Venue Data: Collect location and attribute data for all relevant cultural facilities, using sources such as OpenStreetMap, municipal open data, or GTFS feeds.
• Distance/Travel-Time Computation: Compute $d_{ij}$ (pedestrian) and $t_{ij}$ (multimodal) matrices using routing engines (OSRM, pgRouting, Connection Scan Algorithm, or r5r).
• Attribute Assignment: Assign meaningful weights $w_j$ (by size, attendance, or binary inclusion) and quality scores $v_j$ where applicable.
• Normalization: For each component, normalize $A^P_i, A^O_i, A^V_i$ to [0,1] across all locations.
• Aggregation: Combine axes via a weighted sum:

$$CIAS_i = w_P \tilde{A}^P_i + w_O \tilde{A}^O_i + w_V \tilde{A}^V_i$$

with $w_P + w_O + w_V = 1$. Weights can be set equally, via stakeholder input, or derived by principal components analysis.

• Trade-off Management: In cases of dimension imbalance (e.g., high proximity/low quality), apply penalty terms:

$$CIAS_i = w_P\tilde{A}^P_i + w_O\tilde{A}^O_i + w_V\tilde{A}^V_i - \gamma|\tilde{A}^P_i - \tilde{A}^V_i|$$

Alternatively, in the "generalized accessibility" approach (Verma et al., 7 Apr 2024), practitioners combine cumulative opportunity (step-function thresholding) and gravity-based (continuous decay) variants, potentially integrating equity weights:

$$CIAS_i = s_i \sum_{j\in J} W_j\,f_\theta(c_{ij})\,1\{c_{ij} \leq T\}$$

where $s_i$ upweights origins with high socioeconomic disadvantage.

3. Incorporation of Equity and Multimodality

Equity and modal integration are central to advanced CIAS applications (Verma et al., 7 Apr 2024, Pranto et al., 8 Nov 2025). Facility weights $W_j$ allow venues to be differentiated by type, size, or significance. Origin weights $s_i = 1 + \gamma S_i$ permit targeted upweighting of socioeconomically disadvantaged neighborhoods, where $S_i$ is a normalized disadvantage index and $\gamma$ a policy parameter. This ensures CIAS can function as a tool for spatial justice as well as aggregate efficiency.

For multimodal scenarios, travel-time matrices $c_{ij}^{(m)}$ are computed for each mode $m$ (e.g., walking, biking, transit, driving), with overall accessibility scored via weighted sums:

$$CIAS_i = s_i\sum_{m=1}^M \omega_m \sum_{j\in J} W_j f_\theta^{(m)}(c_{ij}^{(m)}) 1\{c_{ij}^{(m)}\leq T_m\}$$

where modal weights $\omega_m$ are subject to application or policy.

4. Computational Implementation and Data Sources

Efficient computation of CIAS at city scale leverages spatial data structures and open data sources (Pranto et al., 8 Nov 2025). For high resolution, KD-Tree nearest-neighbor queries provide rapid evaluation of distance-decay sums for each grid point. Grid-level CIAS can then be aggregated to population-weighted areal units—tracts, blocks, or user-defined zones—enabling comparisons and mapping.

Key data requirements include:

• Point-layer geocoded venues with attributes (type, size, ratings)
• Population, socioeconomic, and demographic variables at suitable scale (e.g., census tracts)
• Multimodal network graphs (from OSRM, OpenTripPlanner, GTFS)
• Auxiliary data for calibration (e.g., travel surveys, mobility data)

Normalization and post-processing steps often involve z-score standardization of continuous features or imputation of missing data. Scenario testing (e.g., adding venues or new transit lines) is supported by re-running the workflow with updated inputs.

5. Applications and Empirical Patterns

Empirical studies reveal pronounced spatial gradients and social disparities in CIAS distributions (Pranto et al., 8 Nov 2025). In New York City:

• Museums, theaters, and galleries are highly concentrated in the urban core, yielding steep core-periphery drops in CIAS for non-library venues.
• Public libraries display broader coverage, closely trailing population density and serving many lower-income and diverse neighborhoods.
• Statistical correlations: non-library CIAS is positively associated with median household income (e.g., galleries $r \approx +0.29$), while library CIAS is weakly negatively associated with income and strongly positively with density.
• Racial/ethnic analysis: White residents are over-served in cores; Black residents are under-served in peripheries, with libraries helping to ameliorate the deficit.

Explaining these patterns, interpretable AI models (Random Forest, SHAP) show that CIAS features for galleries and museums contribute modestly to income predictions; population density and tract-level racial/ethnic composition are stronger factors.

Applications for policy include identifying “cultural deserts,” prioritizing investment, equity benchmarking, and simulating the impacts of planned infrastructure. Validation methods include mapping CIAS, correlating with observed attendance, and sensitivity analysis over decay and threshold parameters (Verma et al., 7 Apr 2024).

6. Adjustments for User Profiles and Policy Goals

CIAS supports user- and context-specific calibration (Bruno et al., 15 Sep 2025):

• Mobility-impaired users: Decrease $d_0$, use steeper decay in $f^O$, and increase $w_V$ to emphasize quality.
• Tourists: Increase travel budgets $T$ for flatter opportunity decay and raise $w_O$.
• Residents: Raise $w_P$ to favor proximity.
• Equity interventions: Impose axis minima/thresholds and target reweighting to maximize gains in the lowest-performing neighborhoods.

Scenario-based analysis is enabled by updating input data and recomputing all axes and the aggregate CIAS.

7. Limitations and Ongoing Developments

Current known limitations include reliance on Euclidean rather than network-based distances in some implementations (Pranto et al., 8 Nov 2025), uniform weighting of venues, lack of intra-tract demographic differentiation, and potential omission of non-public or informal infrastructure. The cross-sectional nature of available data constrains causal inference.

Active research directions target incorporation of network-travel times, richer and more heterogeneous venue attributes (e.g., capacity, admission cost, opening hours), direct linkage to usage metrics (e.g., attendance, social-media check-ins), and extension to global metropolitan comparisons. These extensions seek to improve both the explanatory power and policy relevance of CIAS in both descriptive and normative urban analytics.

A plausible implication is that future CIAS frameworks will play an increasingly central role in evidence-based cultural facility planning and monitoring spatial equity in the cultural sector.