ChatGPT has geographic biases on environmental justice: Report

A recent study conducted by researchers from Virginia Tech has revealed potential geographic biases in the artificial intelligence (AI) chatbot ChatGPT, particularly concerning its ability to provide locally tailored information on environmental justice issues. The report suggests variations in outputs across different counties, underlining the disparities in access to area-specific data.

According to the findings, ChatGPT exhibits limitations in delivering information relevant to environmental justice issues based on geographical location. The researchers identified a notable trend indicating that such information was more readily available for larger, densely populated states.

"In states with larger urban populations such as Delaware or California, fewer than 1 per cent of the population lived in counties that cannot receive specific information," the report noted.

Conversely, regions with smaller populations, such as rural states like Idaho and New Hampshire, experienced a significant lack of access. The report highlighted that over 90 per cent of the population in these areas lived in counties that could not receive local-specific information on environmental justice issues.

The research paper included a map illustrating the extent of the US population without access to location-specific information on environmental justice issues. The map showcased areas in blue where residents can view such information and areas in red where residents cannot.

Image credit: Virginia Tech

Earlier this year, scholars from the United Kingdom and Brazil published a study asserting that large language models like ChatGPT produce text containing errors and biases that may mislead readers.