Collaborative Research: Disciplinary Improvements: FAIR Pedestrian Accessibility Data for AI Research, Urban Analytics, and Civic Data Governance

This project advances research data management and open science in a high-impact domain by (1) formalizing metadata, versioning, and provenance tailored to crowdsourced geospatial accessibility - a data type that fuses subjective human judgments, artificial intelligence (AI) inference, and multi-source streetscape imagery, fitting no existing metadata template; (2) delivering interoperability via robust mapping to OpenStreetMap and General Active Transportation Infrastructure Specification and a knowledge graph that enables complex, cross‑dataset queries; and (3) introducing a Machine Learning Quality Assurance (MLQA) pipeline that quantifies and communicates uncertainty at the point of access—turning heterogeneous human-AI labels into machine-actionable, reproducible benchmarks. Currently available platforms use other types of technology, such as field data collection and LiDAR, but none have advanced to this point of developing a Findable, Accessible, Interoperable, and Reusable (FAIR) open ecosystem for their data. The benchmarks and governance protocols will lower entry barriers for AI and computer vision (CV) researchers and community scientists, enable cross-city comparison studies, and strengthen reproducibility via pre-benchmarked splits, digital object identifiers (DOIs), and provenance logs—directly operationalizing Gold Standard Science (GSS) principles of reproducibility, transparency, error communication, and collaboration. The resulting infrastructure constitutes a generalizable Research Data Management (RDM) model for citizen-science geospatial data, with contributions to data schemas, standards mappings, and uncertainty-aware quality scoring - which advance science beyond accessibility and pedestrian infrastructure. This project will transform Project Sidewalk from a successful data collection platform into a sustainable socio-technical RDM infrastructure aligned with FAIR and GSS principles. The aims are to (1) Develop a Web Content Accessibility Guidelines-compliant centralized open data portal implementing Data Catalog Vocabulary and Schema.org standards with persistent DOIs, (2) Implement interoperability Application Program Interfaces and Knowledge Graphs, (3) Build an automated MLQA pipeline that scores data reliability and packages DOI-versioned datasets enabling state of the art benchmarking, and (4) Design and deploy a CARE- (Collective benefit, Authority to control, Responsibility, and Ethics) aligned civic data stewardship governance model. The effort will serve three primary research communities — Computer Science (AI/CV), Urban Science (Transportation/Planning/Policy), and Public Health (Mobility/Safety). The project received funding support from the Directorate for Engineering and is managed by the Directorate for Computer and Information Science and Engineering (CISE) Office of Advanced Cyberinfrastructure (OAC). This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.