A recent study published in PNAS Environmental Sciences warns that using artificial intelligence (AI) and Google Street View (GSV) images for urban planning may lead to misleading conclusions. These errors could negatively impact public health efforts aimed at reducing obesity and diabetes.
How is AI Used in Urban Planning?
AI has become increasingly integrated into important sectors like public health and urban planning. For instance, GSV images can be analyzed with deep learning to assess health outcomes linked to neighborhood characteristics defined by census tracts. GSV data reveals environmental details such as vegetation types and urban structures, helping to devise local interventions for mental health, cardiometabolic diseases, and COVID-19.
However, AI-driven predictive models face challenges. They can misidentify biased data and create false correlations, which may distort health predictions. This problem worsens when other factors influence the relationship between environment and health.
What Did the Study Show?
The study explored how GSV features relate to obesity and diabetes rates in New York City’s census tracts. It also examined how these health issues connect to physical inactivity, a major contributor to these conditions. The findings indicated that more crosswalks corresponded to lower disease rates. The effect of physical activity on obesity was greater than on diabetes, aligning with prior GSV-based research. However, no link was found between sidewalk density and health outcomes, differing from earlier studies.
Physical Inactivity Intervention vs. GSV Features
The prevalence of crosswalks and sidewalks influenced health outcomes primarily due to physical inactivity levels in the area. Therefore, changes in health were more closely related to physical activity than to the built environment. A decrease in physical inactivity was associated with significant declines in obesity and diabetes rates.
Built Environment vs. GSV Features
The study found discrepancies between the built environment and GSV data. For example, sidewalks might appear near bridges or highways even when they are missing, leading to inaccurate assumptions about accessibility. This suggests that AI might provide flawed estimates of health interventions because it relies heavily on GSV data without considering critical mediating factors.
Conclusions
This study stands out by comparing GSV features to actual ground conditions. Using a causal framework, the researchers found that improving physical activity in less active neighborhoods could lead to significant reductions in obesity and diabetes prevalence. Nonetheless, limitations in data and changes in the built environment and individual behaviors need careful consideration when using this data for public health initiatives.
Related topics:
- Michael Jordan’s Friend Shares Health Update On The Chicago Bulls Legend After Photos Show Him With Yellow Eyes
- Hayden Panettiere Responds to Concerns Over Health After Emotional Interview
- Nearly 28 Million Americans Affected by Asthma: New Report Highlights Challenges
