NYU Researchers Develop Machine Learning Tool to Detect Underreporting of Heat and Hot Water Issues in NYC

A team of researchers from New York University has developed an automated machine learning tool aimed at identifying underreporting of heat and hot water shortages in New York City’s 311 complaint system. The 311 system allows residents to report various quality-of-life issues, but reporting rates can vary significantly across neighborhoods due to factors like language barriers, socioeconomic status, or awareness of the system. This underreporting can result in some heating and hot water problems going unaddressed, as city inspections are often triggered by these complaints.

The NYU study, published in the Annals of Applied Statistics, introduces a model that analyzes building and neighborhood characteristics to estimate where underreporting is likely occurring. The researchers used two main approaches: first, identifying buildings with no reported problems but similar characteristics to those with frequent complaints; second, spotting buildings with fewer complaints than expected based on size and estimated problem duration. Factors such as building age, rental status, number of units, and neighborhood demographics (including the proportion of limited-English speakers, elderly residents, and families with children) were incorporated into the analysis.

If adopted by city agencies, this tool could help direct inspections and resources to buildings and communities where residents may be less likely to report issues, ultimately promoting more equitable service delivery. The authors hope their methods will assist city agencies and advocacy groups in improving access to 311 and ensuring that heat and hot water problems are addressed more effectively.