Capability-Informed Accessibility Metrics

Updated 18 October 2025

Capability-informed accessibility metrics is a framework that defines access by integrating individual capabilities, contextual constraints, and equity concerns.
It employs mathematical models such as isochronic analysis, gravity-based formulations, and belief functions to capture uncertainty and personalized outcomes.
The approach enables actionable insights for inclusive design in domains like urban mobility, web accessibility, and digital content evaluation.

Capability-Informed Accessibility Metrics provide a methodological and conceptual foundation for assessing access to services, environments, technologies, and opportunities in a manner that systematically incorporates user capabilities, contextual limitations, and the heterogeneity of needs. This field has evolved from aggregate, infrastructure-centric evaluations toward frameworks that are user-centered, context-sensitive, and capable of representing uncertainty, personalization, and equity concerns across domains ranging from web accessibility and urban mobility to built environments and digital content.

1. Conceptual Foundations and Frameworks

Capability-informed metrics are underpinned by the capability approach, which emphasizes what individuals are actually able to do—i.e., the real freedoms and opportunities open to them—rather than merely the presence or quantity of infrastructure or resources. This perspective reframes accessibility from a supply-side or compliance issue to one of individual or group capability.

Recent literature advocates multidimensional frameworks capturing accessibility along several axes:

Proximity: The ease and activity range for short-distance, often active access (e.g., walking to local amenities), where metrics decay with distance or time beyond thresholds (e.g., ≤15 min) (Bruno et al., 15 Sep 2025).
Opportunity: Access to broader, non-local resources (e.g., jobs, specialized services), typically encompassing longer travel times (e.g., up to 60 min) and integrating public transit and network effects (Biazzo et al., 2018, Bruno et al., 15 Sep 2025).
Value: The subjective or personal significance of accessible options, reflecting that similar numeric access to amenities or networks may be perceived differently due to reputational, quality, or relevance factors (Bruno et al., 15 Sep 2025).

These dimensions interact: high local proximity may not guarantee access to high-value opportunities, and trade-offs are context- and user-dependent.

2. Mathematical Formulations and Modalities

Capability-informed accessibility metrics rely on various mathematical formulations, frequently involving weighted averages, opportunity sums, impedance decay functions, and set-based feasibility:

Isochronic Methods: Isochrone-based analysis defines the set of reachable points within a given travel time τ from a location λ, yielding derived metrics such as the "Velocity Score"—the average expansion speed of accessible area—and "Sociality Score"—the average population accessible (Biazzo et al., 2018).
Gravity-Based and Cumulative Opportunities: A generalized accessibility metric is formulated as:

$a_i^{(k, m, t)} = \sum_{j \in R} o_j^k \cdot W_{ij}^{k,m,t}$

where $o_j^k$ is the number of opportunities of type $k$ at $j$ , $W_{ij}^{k,m,t}$ is an impedance weight (e.g., $f_{k,m}(c_{ij})$ if $c_{ij} \leq t$ , else 0), with $c_{ij}$ representing travel cost, and $f_{k,m}$ an empirically fitted decay function (Verma et al., 7 Apr 2024).

Space-Time Constraints: The SPA metric integrates real observed schedules, time budgets, and multi-anchor trip chain feasibility. An opportunity $k$ is feasible for individual $i$ if

$a^k_i = \begin{cases} 1 & \text{if } t_{\text{hw}} + t_{\text{wk}} + t_{\text{kh}} \leq tb_i \ 0 & \text{otherwise} \end{cases}$

with the aggregate SPA given by $A_i^k = \sum a_i^k$ (Liao et al., 11 Oct 2025).

Uncertainty and Belief Functions: In digital accessibility, Dempster–Shafer theory combines uncertain evidence across deficiency frames (e.g., visual, motor) to quantitatively assess accessibility, producing probability masses for accessible, non-accessible, and uncertain states. This is expressed as mass functions, for instance:

$m(Ac) = E(Ac) / [E(Ac)+E(\neg Ac)+E(\Omega)]$

with $E(\cdot)$ computed from test success/failure and confidence weights (Dubois et al., 2015).

Personalization and User Profiling: Methods now frequently incorporate user models, either as vectors of abilities/preferences or as JSON objects encoding motion primitives, reachability, and sensory sensitivities (Huang et al., 31 Jul 2025, Kumar et al., 2020).

3. Integration of Equity, Personalization, and Fairness

Equity and personalization are critical components—and often points of methodological expansion:

Equity Weighting: Metrics now formalize the assignment of weights to socioeconomically disadvantaged areas or demographic groups, affecting aggregation (e.g., Socio-Economic Disadvantage Index weights in summary metrics) (Verma et al., 7 Apr 2024).
Accessibility Fairness: In transportation systems, fairness is operationalized by minimizing population-weighted unaccessibility, defined as slack over a travel time threshold for origin–destination pairs, formalized as:

$acc = \frac{\sum_{r \in R} n_r u_{(r)}}{\sum_{r \in R} n_r}$

where $u_{(r)}$ is the unfairness per region and $n_r$ the population (Salazar et al., 30 Mar 2024).

Personalized Evaluation and Adaptive Models: Web and built environment accessibility metrics employ user-specific simulations (e.g., movement time, reachability), dynamic user modeling (via LLMs or explicit feedback loops), and the use of common user profile formats for cross-system adaptability (Huang et al., 31 Jul 2025, Kumar et al., 2020).
Path- vs. Flow-Based Disparity: Flow-level average fairness in network optimization does not guarantee individual path-level equivalence, prompting calls for refined allocation algorithms (e.g., per-path slack minimization) (Salazar et al., 30 Mar 2024).

4. Domain-Specific Implementations

The application of capability-informed accessibility metrics spans domains:

Web Accessibility: Evaluation tools (e.g., WAccess) check WCAG guideline violations and combine automatic scoring over observable criteria with fix suggestions, yielding quantitative accessibility snapshots (e.g., per conformity level, code-specific analysis) (Boyalakuntla et al., 2021). Simulation-based models and user profiles extend these with user-centered, interaction-based assessments (Kumar et al., 2020).
Digital Content Accessibility: AI pipelines generate, evaluate, and optimize alt-text using visual-semantic models and context integration, with performance measured via Cosine Similarity and BLEU, and validated by user studies (Shen et al., 30 Dec 2024).
Urban and Transport Accessibility: Routing-enabled frameworks compute space- and time-aware metrics, support scenario simulation, and provide tools for equity-driven planning and infrastructure assessment (Biazzo et al., 2018, Lang et al., 2020, Verma et al., 7 Apr 2024).
Built Environments: Human-centric models (e.g., SHAPE) generate weighted accessibility graphs (vertices: spatial positions; weights: multidimensional human factors such as energy expenditure, slopes, step types) for room/building/topography level optimization, combined with standard graph algorithms for evaluation (Schwartz, 2021).
Robotic Systems: Accessibility in robotics is treated as transition ease between poses, informing initial state distributions that maximize coverage and efficiency of state space exploration (e.g., K-Access clustering) (Zhang et al., 2021).

5. Evaluation Strategies and Empirical Findings

Robust evaluation protocols demonstrate practical impacts:

Empirical Benchmarks: Quantitative metrics (e.g., Cosine Similarity, BLEU, error reduction rates) are routinely reported for digital content accessibility, while urban metrics rely on city-level and demographic stratification (e.g., Gini indices for spatial/racial disparities) (Shen et al., 30 Dec 2024, Ju et al., 17 May 2025, Verma et al., 7 Apr 2024).
User Studies and Human-in-the-Loop Assessment: Qualitative feedback from end-users (especially in disability contexts), Likert scale usability ratings, and scenario-based evaluations validate metric relevance, personalization success, and the detection of out-of-scope or hallucinated concerns (Huang et al., 31 Jul 2025, Shen et al., 30 Dec 2024).
Simulation and Case Study: Large-scale, cross-domain studies (e.g., 100 French news websites, 3 major US metropolitan areas, Bay Area EV infrastructure) identify persistent gaps, document the evolution of accessibility over time, and facilitate scenario planning (Dubois et al., 2015, Verma et al., 7 Apr 2024, Ju et al., 17 May 2025).

6. Methodological Evolution and Future Directions

Development trends emphasize:

Information Fusion: Combining outputs of multiple assessors (web, transport, perceptual) using theoretical frameworks (e.g., belief function theory) to handle uncertainty and imprecision (Dubois et al., 2015).
Contextualization and Dynamic Adaptation: Metrics increasingly prioritize integration of external context (e.g., surrounding text for alt-text, article or environmental context for image description), as underscored by deficits in referenceless evaluation (Kreiss et al., 2022).
Scalability and Open Source Tooling: Many methodologies scale to thousands of instances via efficient spatial indexing, modular codebases, and batch analysis—providing both accessibility metrics and actionable diagnostics (Lang et al., 2020, Boyalakuntla et al., 2021).
Toward Comprehensive Capability Models: Enhanced metrics are expected to integrate personality, preference, group, and adaptation layers, transitioning from static checklists to continuously updated systems accommodating life-stage, context, and evolving infrastructure (Huang et al., 31 Jul 2025, Liao et al., 11 Oct 2025).

7. Practical and Policy Implications

Capability-informed metrics have direct impact on design, governance, and policy:

Targeted Interventions: Equity-weighted and opportunity-aware metrics support resource allocation to reduce disparities (e.g., in public charging infrastructure or essential services) (Verma et al., 7 Apr 2024, Ju et al., 17 May 2025).
Design Diagnostics: By decomposing metrics into quality factors and subattributes, such as perceptiveness, operability, and localization, practitioners can identify specific accessibility shortcomings and formulate targeted recommendations (Kuz et al., 27 Sep 2025, Schwartz, 2021).
Standardization and Adaptability: The systematic formalization via standardized frameworks (ISO/IEC, WCAG), detailed submetrics, and scaling mechanisms enhances the repeatability, comparability, and rigour of accessibility evaluation (Kuz et al., 27 Sep 2025, Dubois et al., 2015).
Personalization at Scale: AI-enabled, LLM-driven systems indicate the feasibility of delivering tailored accessibility assessments across environments, highlighting both opportunity and sensitivity to variation in user needs (Huang et al., 31 Jul 2025).

In summary, capability-informed accessibility metrics are distinguished by their integration of user capabilities, context, equity, and uncertainty, realized via formal mathematical models, personalized user models, simulation, and empirical validation. The result is a set of tools and frameworks that move beyond generic or infrastructure-centered accessibility measurement, providing nuanced diagnostics and actionable insights for designing inclusive digital, physical, and urban environments.