Better Targeted Anti-Trafficking Inspection Efforts in Brazil

The Human Trafficking Data Lab has developed and deployed a new decision support system that uses human-centered artificial intelligence to enhance the work of Brazilian anti-trafficking investigators. The tool is designed to help investigators accurately identify and target multi-agency task force inspections more effectively, identify clusters of cases and opportunities to bundle site inspections together, and direct limited resources more efficiently, and reach victims faster.  

The tool embeds powerful AI assistance and analytics into existing case management platforms, enhancing workflows while keeping humans in the loop, and limiting barriers to uptake. Functionalities include: 

1. AI-driven case classification – uses a Large Language Model (LLM) to interpret incoming tips in real time, classifying them according to risk factors critical for investigators. These include whether child labor or forced labor is suspected, and whether the tip alleges sexual exploitation, debt bondage, violence, or sweat shops.
2. Machine learning risk modeling – integrates a machine learning model to identify and flag the highest priority cases based on the key characteristics of each case: whether children are involved, whether sexual exploitation or violence is reported, how many victims might be affected, and others.
3. Geospatial analytics – includes interactive maps visualizing the geographic distribution of cases, allowing investigators to quickly identify the largest most emergent clusters of cases, or cases that share characteristics like economic sector and might be bundled into one investigation.
4. Automated data pipelines – automates many of the investigative tasks typically performed by prosecutors, such as searching administrative databases for records of past labor law violations, worker safety complaints, and environmental regulation violations. The decision support tool organizes these documents and produces concise summaries to assist prosecutors in building more comprehensive cases.

For decades, anti-trafficking law enforcement has faced intractable challenges sifting through cases and acting in time to rescue victims before traffickers move their teams and evade detection. Modern AI technologies bring efficiency and analytics needed to assist authorities in discerning the cases most likely linked to trafficking and respond quickly. Since its launch in October 2023, we have observed widespread adoption and prosecutor workflows changing to adapt to the new capabilities.  

The decision support tool demonstrates The Lab’s commitment to building evidence-based and technology-driven solutions to the fight against human trafficking. We are also committed to building our scientific understanding of “what works” and are formally quantifying the impact of our tool using a gold standard, nationally implemented randomized controlled trial (RCT). The RCT is focused on measuring the tool’s ultimate impact on the number of victims rescued from trafficking, and studies three possible mechanisms: (1) Do anti-trafficking task forces get better at successfully identify trafficking? (2) Do task forces target more cases per task force? (3) Do task forces focus on the largest cases with the greatest number of victims?  

Location: Brazil (nationwide)

Funder

Partners