How data analytics can help tackle the scourge of human trafficking

There are an estimated 50 million people living in modern-day slavery worldwide. Using big data analytics tools, data scientists, law enforcement and victim advocacy groups can focus their efforts, working to combat human trafficking by identifying correlating factors and criminal hotspots, and finding and sharing examples of anti-trafficking best practice.
Human trafficking has many faces:
- Men are enticed from Mexico and Central America with offers of attractive hourly rates for agricultural work in the US – but once they arrive, they’re treated like slaves
- Young children cook, clean and wash for families nonstop, with no rest even on weekends, for just a few dollars a month
- Women and girls are forced into sex work. Some have reported servicing up to 30 men each day
All of these men, women and children are victims of human trafficking. All are largely invisible. And they’re all around us.
International Labour Organization (ILO) reports from 2021 estimate that 50 million people are living in modern-day slavery across the globe. Of those, 28 million are in forced labour and 22 million are trapped in forced marriage. A study of forced labour from the ILO highlights a US$150bn industry that continues to worsen as the international crisis of human trafficking intensifies.
Few industries are immune. Agriculture, construction, garment and textile work, caterers and restaurants, domestic work, food processing, health care, cleaning, entertainment or the sex industry – all harbour trafficked humans.
Over the past 15 to 20 years, the US and other nations have taken steps to fight the problem. The US fully complies with international minimum standards for the elimination of human trafficking and has instituted numerous programmes to combat the issue.
For example, the State Department’s 2022 Trafficking in Persons Report for fiscal year 2021 shows that the Department of Homeland Security (DHS) conducted 542 trafficking-focused interviews using a trauma-informed approach. And DHS victim assistance helped 728 victims of human trafficking. The Department of Justice (DOJ) forensic interview specialists conducted 202 human trafficking forensic interviews of victims, and DOJ’s 172 victim specialists provided services to human trafficking victims in 708 cases.
Over the years, dozens of states have passed hundreds of laws to address sex trafficking and the exploitation of children. Yet law enforcement continues to struggle with identifying victims and implementing laws. “Most governments are in the nascent stages of recalibrating their approach – if the issue is on their radar at all,” states Governing magazine in the article “Fighting Sex Trafficking Is Harder Than It Seems.”
Now, organisations like Peace-Work and SAS are successfully applying sophisticated analytics to help combat human trafficking. Part of the growing Data for Good movement, Peace-Work is a volunteer cooperative of statisticians, data scientists and other researchers who apply analytics to issue-driven advocacy.
Peace-Work identifies human trafficking conditions
Who would have thought that the degree of slavery in a location in 1860 would correlate to high levels of human trafficking today? That was a surprising association Peace-Work researchers identified that’s helping law enforcement pinpoint where victims are most likely to be located.
Starting with summary data from National Human Trafficking Resource Center victim hotlines in individual states, Peace-Work volunteers used meta-analysis to perform a single, unified analysis on data sets across multiple states. They uncovered several additional factors that correlate to human trafficking:
- African-American and mixed-race population
- Per capita income
- Percentage of households below the poverty line
- Home foreclosure rates
- Areas with large income disparities
Next, the group looked at these factors across different metro areas and identified 18 where rates of human trafficking were likely to be high. While law enforcement already knew about 10 of them, the other eight were previously unknown.
With no full-time staff, Peace-Work completes projects like this by breaking down larger statistical projects into small pieces to accommodate volunteers’ busy schedules. Each volunteer is given access to SAS® Analytics.
With its first project complete, Peace-Work is planning several additional projects. “Our next step will be to look at states and metros that are better at finding and reporting trafficking to identify best practices that other states can use,” says David J. Corliss, PhD, founder and director of Peace-Work.
“One of the most effective techniques is training emergency room doctors to recognise signs of human trafficking. We also expect to look at the international situation and identify best practices for victim support.”
Putting text analytics on the job
SAS recently sponsored a project making information currently hidden in text reports more accessible to those seeking to fight human trafficking. The State Department publishes roughly 200 reports each year about human trafficking in various countries. SAS used supervised and unsupervised machine learning to assess patterns of human trafficking currently buried in the text of these reports.
“Previously, organisations interested in taking advantage of these reports would have to read through them to get a hunch about the issues and trends they contained,” explains Tom Sabo, an advisory solutions architect at SAS.
“Our goal was to surface useful data that could help advocacy groups or overseeing agencies better determine where to focus anti-trafficking efforts,” Sabo continues. “We used text analytics to comb through all the Trafficking in Persons reports since 2013 and identify patterns that were not apparent previously. We then created visualisations to make the information highly accessible.”
For example, the team used text analytics to simultaneously identify the source and destination countries for trafficking worldwide and depict these connections. SAS Visual Analytics displayed this information geospatially, with lines drawn between source and destination countries. The colour-coded lines indicated whether the two countries involved were cooperating to address the issue. Text analytics also determined the cooperation indicator.
The team initially used text clustering to analyse the introduction and prosecution sections of reports. For example, a cluster of “forced”, “child”, “beg” and “street” may indicate child exploitation. Based on these clusters, text analytics can extract the most common sentences that back up the clusters as evidence. Researchers can also drill down to view the supporting text within the actual documents to fully understand the context of the statistical findings.
Over the past several years, the team extended its work by creating an analytics framework to counter international human trafficking. Researchers approached the problem from several different angles. For example, they assessed patterns in the Trafficking in Persons report and drilled into other data sources, like migration data from the United Nations and armed conflict data from ACLED.
Now the team can further evaluate trafficking by accessing advertisements and hidden recruitment posts from sources like the former backpage.com. They can also evaluate how anti-money laundering programmes could assist in their efforts.
The counter-human trafficking framework is designed to incorporate new data sources as soon as information emerges. Using AI techniques, it assesses patterns in this data, such as those that identify specifically how victims are being exploited.
Human trafficking is a heinous crime that thrives in the shadows. By taking advantage of big data analytics, data scientists and researchers can shine a light on the problem. In turn, law enforcement and victim advocacy groups can find better ways to address it.