Quantcast

Could AI be the answer to solving the congestion crisis?

CAMBRIDGE, Mass., March 28, 2024 – The research paper, AI-generated visuals of car-free US cities help improve support for sustainable policies, published in Nature Sustainability combines insights from behavioral science with recent advances in AI. The study was led by Rachit Dubey, a postdoctoral research fellow at MIT’s Sloan School of Management in collaboration with Rahul Bhui, assistant professor of marketing at MIT Sloan. Additional co-authors are Thomas L. Griffiths, professor of psychology and computer science at Princeton University, and PhD candidate Mathew D. Hardy at Princeton University.

“America’s infrastructure is extremely car-centric, so using more sustainable modes of transportation is currently inconvenient or unavailable to many people,” said Dubey. “And, unfortunately, public transportation has increasingly become a polarizing topic for both the American public and elected officials who are generally reluctant to support policies that try to increase these investments.”

Bhui said the research results may be significant for policymakers who hope to promote sustainability without directly referencing climate change.

The research involved two experiments with approximately 3,200 participants across the political spectrum who were given a hypothetical transport bill to increase America’s investment in public transportation, improve and expand sidewalks for better walkability, and build dedicated bus and bike lanes on major roads with a price tag of $500 billion.

The study subjects were shown “before” images including Google maps visuals of heavily used roadways such as the I-35 expressway through Austin, Texas, and “after” images that depicted how the streets would look if the proposal went through.

In the study, participants were randomly divided into three groups. In each case, participants were presented with the same “before” images of car-centric downtown city streets. However, the three groups were presented with different “after images”: either a sentence about how the proposal would transform the location, a cartoon-like illustration, or the AI-generated scenes.

Participants were then asked to indicate on a scale of 1 to 10 how they felt about the hypothetical bill, where 1 indicated strongest opposition and 10 indicated strongest support.

When analyzing participant responses, it was observed that participants who viewed the AI images provided an average support rating of 6.4 out of 10 for the bill. This level of support was significantly higher compared to participants who saw no image or those who saw the cartoon-like images.

Notably, study participants who identified as Republicans made the biggest shift. While the majority opposed the bill when they saw no images (average support rating of 3.5) or when they saw cartoon images (average support rating of 3.7), they were more supportive of the bill after viewing the AI images (average support rating of 4.8).

“Nudging someone to change their entire belief system about the world is a very hard thing to do. But we can still influence the passing of sustainable policies at a local or national level if we help people imagine specific outcomes of a policy,” Bhui said.

In the future, Dubey said he’d like to test this approach’s effectiveness for real-world proposals. With the climate crisis intensifying, he believes it is more important than ever to understand human decision-making. “There’s rich psychological literature on imagination that shows when people are able to imagine something better, they’re more motivated to do something about it,” he said. “We need to show people how much better their life will be if a green or a sustainable policy were passed.”

Source: “AI-generated visuals of re-imagined sustainable cities enhanced Americans’ willingness to support green transportation policies,” Mar. 28, 2024 MIT Sloan School of Management press release.

Corresponding, connected home-page-featured image: Michigan Department of Transportation




The material in this press release comes from the originating research organization. Content may be edited for style and length. Want more? Sign up for our daily email.