top of page

Can Artificial Intelligence Address Equity Gaps in Urban Health Surveillance? Evidence and Implications from Boston's Infrastructure Monitoring Systems

  • 3 days ago
  • 2 min read
A new working paper from Khahlil Louisy examines whether AI-based urban infrastructure monitoring can address equity gaps in public health surveillance, using evidence from Boston’s 311 complaint system and AI-detected pavement data. The paper shows that complaint-based systems reflect socioeconomic patterns of civic participation, and argues that AI can complement but not replace participatory surveillance without rigorous validation and equity safeguards.

A new working paper from Khahlil Louisy examines whether AI-based urban infrastructure monitoring can address equity gaps in public health surveillance, using evidence from Boston’s 311 complaint system and AI-detected pavement data. The paper shows that complaint-based systems reflect socioeconomic patterns of civic participation, and argues that AI can complement but not replace participatory surveillance without rigorous validation and equity safeguards.


Abstract

Background: Infrastructure defects, particularly pavement quality, contribute substantially to outdoor pedestrian falls, which now exceed motor vehicular injuries as a source of serious pedestrian harm.1,4 Public health agencies are increasingly relying on resident complaint systems like 311 platforms for infrastructure surveillance, yet these systems may systematically underrepresent conditions in disadvantaged communities. AI-based monitoring is often proposed as an equity solution and an expanding list of cities are adopting the technology, but empirical evidence is limited.


Methods: I analyzed socioeconomic patterns in 2.97 million Boston 311 reports (phone calls, messages to operators, and use of app) from 2011 to 2024, across 216 Census tracts using multivariable regression, and compared complaint-based versus AI-detected pavement conditions in the Boston neighborhood of Jamaica Plain (5 tracts).


Results: Median household income strongly predicted complaint volume (r=0.42, p<.001) independent of digital access. A $10,000 income increase associated with 14 additional reports per 1,000 residents annually. In Jamaica Plain, the ratio of AI-detected distresses to resident complaints varied 2.4-fold across tracts, indicating systematic divergence between surveillance modalities.


Conclusions: Complaint systems reflect civic capacity rather than underlying conditions alone. AI-based monitoring can complement but not replace participatory surveillance. Responsible integration requires multi-modal approaches combining algorithmic detection with community engagement and independent validation. This preliminary work (5 tracts with AI data) illuminates critical tensions requiring expanded multi-city research with health outcomes linkage.


Access the full paper here:


bottom of page