We recently shared a story about the development of an electric car battery that can be charged in just 5 minutes, but the issue of charging station accessibility still makes some drivers nervous about transitioning to an all-electric vehicle. 

In an effort to improve charging station infrastructure, researchers from the School of Public Policy at the Georgia Institute of Technology have trained an AI system to analyze vast numbers of charging station user reviews to assess where charging infrastructure is missing or inadequate. 

The researchers trained the AI using reviews from 12,720 US charging stations to categorize stations based on eight critical categories: functionality, availability, cost, location, dealership, user interaction, service time, and range anxiety. The valuable system not only greatly cuts down on research costs for companies and governments interested in expanding EV infrastructure, but also offers valuable insights into consumer preferences and needs in specific areas. 

Training an AI system to read user reviews was no easy task. The reviews range in length from a few words to whole paragraphs and often contain misspellings, slang, or regional references that aren’t easily interpreted. 

So far, the system has achieved 91 percent accuracy in uncovering user meaning in reviews and more importantly, has identified communities, mostly in the West and the Midwest United States, that are seeking more EV resources. These areas, in states like California, Oregon, Utah, and Nebraska, are some of the easiest places to implement the transition to electric vehicles because the demand is already there. They just need infrastructure to support it. 

With states like California and Massachusetts aiming for all-electric vehicle sales by 2035, massive infrastructure updates are in order including improved charging station availability and efficiency. Once expanded, this AI system will offer valuable insights to governments and companies as to how best to provide reliable EV infrastructure for all citizens, regardless of geographic location.