Can ChatGPT get the EV transition back on track?

If Massachusetts is to have any hope of reducing carbon emissions to reach its climate goals, hundreds of thousands of drivers will need to switch to electric vehicles. But the growth in sales of EVs slowed this year, particularly for the market leader, Tesla.

Getting the EV transition back on track is critical in a state where transportation accounts for 39 percent of all emissions, the highest of any sector in the economy. Is the problem high-priced vehicles, fears about limited driving range, or a lack of charging stations? Or have EV-interested consumers grown wary of Tesla chief executive Elon Musk?

Omar Asensio, director of the Data Science and Policy Lab at Georgia Institute of Technology, is spending some time at Harvard delving into the issue with a new tool: artificial intelligence. As a climate fellow at Harvard’s Institute for Business in Global Society, Asensio and a team of researchers used a program with the same generative AI technology underlying ChatGPT to find out what consumers really think about EVs — and what might be holding them back.

The program churned through more than one million reviews of EV driving by consumers and charging experiences posted on sites such as Plugshare.com, a specialty site for EV owners in search of trends.

“What we found in the US is there is not major concern about range anxiety,” Asensio said. “The number one issue is charger reliability,” including reliable access.

In the data from the reviews, more than one in five attempts to charge failed. Sometimes a charger was out of order, sometimes a driver’s payment could not be processed, sometimes the wait for an available charger was too long. And sometimes a gas-powered car was parked in front of the charger, blocking the space.

“The AI started picking up this interesting lingo people were using that ‘I’m getting ICED,’” Asensio said. “We were thinking what the heck is that? We had to confirm our machine learning algorithms were working.”

“ICED” is a slang term used by EV drivers when they are blocked from charging by a gasoline-powered car parked in the spot. It is derived from the acronym for “internal combustion engine.”

The AI almost always (more than 90 percent of the time) correctly picked up the negative connotations from the mentions of drivers being ICED. The researchers compared that to people looking at the same reviews. Humans only grasped the annoyance factor correctly half of the time.

Now Asensio is pondering where else he can deploy AI analysis tools to improve climate research projects. The team has also studied corporate sustainability reports and looked into consumers’ conservation and recycling behavior.

“Very quickly, a new community of scholars and business networks are evolving around the use of AI-powered tools applied to solving more climate change themes,” he said. “It’s a very exciting time to be doing thjis kind of work.”

Aaron Pressman can be reached at aaron.pressman@globe.com. Follow him @ampressman.

Originally Appeared Here