Technology has long provided the ability to evaluate and understand large amounts of data, but recent advances in artificial intelligence (AI) have made the process faster and more powerful, with potential applications multiplying exponentially.

Sheikh Rabiul Islam
Sheikh Rabiul Islam, assistant professor of computer science 

For Sheikh Rabiul Islam, an assistant professor of computer science at Rutgers University–Camden, one of those applications is creating safer communities. Islam recently co-authored two different research papers examining how AI could be used to address serious community threats: potential school shootings and alcohol-related road fatalities.

“This research represents a significant step toward implementing knowledge-infused artificial intelligence in real-world scenarios,” Islam said. 

Islam and his team sought to develop preventive measures against school shootings using AI. This research is the basis of the paper “Predicting Potential School Shooters from Social Media Posts.”

While these incidents are shocking when they happen, offenders often exhibit warning signs on social media that go undetected or unreported.

“We aimed to create a multimodal model made up of several variables capable of predicting sentiments simultaneously from photos and text across all of the major social networking platforms, anticipating that this would generate a unified prediction,” Islam said.

These variables were used to train the AI model in the integrated system, which allowed social media posts to be evaluated as positive or negative. The analysis revealed a reasonable level of accuracy, with the team recommending further research to make the model more robust, focusing on eventual adoption by public safety entities.

“The eventual goal for a real-world implementation and deployment of such a system would be to analyze real-time posts from multiple social media platforms,” the team wrote.

Assisting law enforcement agencies was also the focus of the second research paper Islam recently completed, "Aiming to Minimize Alcohol-Impaired Road Fatalities: Utilizing Fairness-Aware and Domain Knowledge-Infused Artificial Intelligence.”

“Biases have been observed through racial profiling, leading to some groups and geographical areas facing fewer DUI tests, resulting in many actual DUI incidents going undetected, ultimately leading to a higher number of fatalities,” the paper notes. Islam and his coauthors used AI-supported data analysis to map DUI-related fatalities more fairly and impartially. Their goal was to support improved allocation of policing resources, hopefully leading to a reduction in DUI-related deaths and a significant positive impact on road safety.

Islam and the team again developed a multivariable approach, including data on population, age, and the per-capita income of residents. Again, the AI-powered model proved effective, providing insight into how public safety officials could better monitor impaired driving in the communities they serve. Islam and the team recommended additional research to enhance the model's capabilities and sharpen the resulting predictions.

Islam sees a future where AI can help make communities safer and stronger.

“Significant research is underway in the field of computational social science, and it is not unimaginable that AI may be used in the near future to fight poverty, promote well-being, support broad educational improvements, achieve gender equity, address climate change, and reduce inequalities,” Islam said. "Many people are concerned about the potential impact of AI, but I think it has the potential to foster peace and justice around the globe."