Johns Hopkins University researchers have developed an artificial intelligence (AI)-based tool capable of identifying the risk factors that contribute to car crashes and accurately making predictions about future incidents.

The tool, dubbed SafeTraffic Copilot, was developed to provide experts with both crash analyses and crash predictions in a bid to reduce automotive accidents.

Using Large Language Models (LLMs), the SafeTraffic Copilot was trained using a combination of text (such as descriptions of road conditions), numerical values (such as blood alcohol levels), satellite images and on-site photography. The model can also evaluate both individual and combined risk factors, thereby offering a more detailed understanding of how the interaction of these elements influence crashes.

The team explained that SafeTraffic Copilot features a continuous learning loop so that prediction performance will improve as more crash-related data is entered into the model, which will make the tool even more accurate over time.

"By reframing crash prediction as a reasoning task and using LLMs to integrate written and visual data, the stakeholders can move from coarse, aggregate statistics, to a fine-tuned understanding of what causes specific crashes," the researchers explained.

According to the team, the model promises to give policymakers and transportation designers a reliable tool to identify combinations of factors that increase crash risks.

An article detailing the tool, “SafeTraffic Copilot: adapting large language models for trustworthy traffic safety assessments and decision interventions,” appears in the journal Nature Communications.