911 calls: AI flags violence against women
An innovative pilot project by the United Nations Office of Drugs and Crime (UNODC) and the Mexican National Institute of Statistics and Geography (INEGI) has tapped the power of artificial intelligence (AI) to detect unreported violence against women.
By UNODC Mexico City
“911 – what’s your emergency?”
The three-digit emergency response hotline is increasingly familiar worldwide.
Skilled operators dispatch police, fire crews, and ambulances to resolve dangerous situations and save lives. Rapid decisions and action are key to help victims of violence.
But often, the immediate response overlooks ongoing abuse, leaving victims at risk and their abusers untouched.
A new project from the UNODC-INEGI Center of Excellence in Statistical Information on Government, Crime, Victimization and Justice in Mexico City uses artificial intelligence (AI) to sift through unstructured data and capture subtle warning signs.
The AI transformer model – a form of neural network that learns context from sequences of words – analyses transcripts of “911” emergency calls to identify incidents of gender-based violence that might otherwise go unnoticed.
“AI is rapidly transforming various sectors, and its application in public safety and preventing violence against women offers significant opportunities to create a more inclusive and peaceful world,” says the UNODC Representative in Mexico, Kristian Holge.
Catching the warning signs
In most countries, emergency operators classify incoming calls based on the care needed first – whether it’s for burns, broken bones or other injuries. But the traditional approach can mean overlooking situations where the patient’s intimate partner causes the emergency.
The authorities, consequently, miss opportunities for timely intervention. Victims of gender-based violence, most often women harmed by men, receive less support than they could. Affected children and families are similarly neglected.
The UNODC-INEGI Center of Excellence aims to change this situation.
The “Saving Life 911” platform helps identify – and properly categorize – intimate partner violence events that were originally labelled as medical incidents, fires or accidents.
More broadly, the experimental model demonstrates how evidence-based AI applications can improve public services such as police response and investigation, social service and shelter provision, education, and more.
Protecting individual privacy
The analysis of sensitive data complies with strict legal standards to protect people’s privacy and ensure data is handled responsibly.
Clear legal frameworks and data protection protocols will be key for wider replication of the model in Mexico and other countries.
AI use can help create a more comprehensive and integrated approach to public safety, bringing together first responders, service providers, statisticians and data analysts to address violence against women.
AI-driven analysis can also enhance standardized protocols for data sharing and analysis, providing solutions that are not just innovative but scalable and sustainable.
Going beyond Mexico
The project’s implications extend far beyond Mexico. Emergency response resources are stretched thin in many countries, and response centres often get overwhelmed.
Misclassifications, whether due to a lack of training or the sheer volume of calls, leave vulnerable populations unprotected.
The AI-driven approach, by improving classifications and maintaining a longer view, can improve national capacities to detect and respond to gender-based violence.
The Mexico City project’s initial results highlight the power of AI to drive positive social change, save lives, and promote justice and equality.
“As the technology comes into wider use, it could transform public safety and gender equality efforts globally, making the world a safer and more inclusive place for everyone,” says Holge.
The UNODC-INEGI Center of Excellence presents “Saving Life 911” at SDG Digital on 20-21 September.
Use SDG Digital resources and promote on social media.
Header image credit: Adobe Stock