A stark light has been cast on the geographic biases embedded in ChatGPT, a popular AI chatbot. The study reveals a concerning disparity in the chatbot’s ability to provide environmental justice information that caters to local needs.
Researchers have pinpointed a troubling trend: densely populated states like California and Delaware have almost universal access to tailored information, with less than 1 percent of their populations in informational deserts. Contrastingly, in sparsely populated states such as Idaho and New Hampshire, the figure skyrockets to over 90 percent.
![Environmental Justice Issues Map](jpg/environmental-justice-issues-map.jpg)
Virginia Tech’s own, Lecturer Kim, from the Department of Geography, underscores the urgency for further investigation. Kim states, “Our initial findings are a call to action — geographic biases in AI tools like ChatGPT can no longer be overlooked.”
The university’s findings are not isolated. They join a chorus of recent studies pointing out political and error-related biases in large language models, highlighting the potential for misinformation.