Data for Good Researchers Fight Human Trafficking

With a PhD in statistical astrophysics, David Corliss works in analytics architecture at Ford Motor Company while continuing astrophysics research on the side. He is the founder of Peace-Work, a volunteer cooperative of statisticians and data scientists providing analytic support for charitable groups and applying statistical methods to issue-driven advocacy in poverty, education, and social justice.

Data for Good researchers often tackle problems reflecting the worst of society’s ills. One such area is human trafficking, which has attracted considerable interest in recent years. The Walk Free Foundation is dedicated to ending modern slavery and human trafficking and works to raise awareness; partners with governments, businesses, and NGOs; and performs research. Collecting and analyzing data plays a critical role in this advocacy.

Collecting data on modern slavery and human trafficking faces many challenges. As these activities are illegal in many areas, secrecy is required for them to thrive. Victims can be reluctant to report their plight, and perpetrators strive to prevent detection.

Walk Free worked with Gallup Inc. to develop a world poll and has conducted 48 surveys since 2014. Tens of thousands of responses have been collected. Senior researcher Davina Durgana used these to develop the Global Slavery Index. Walk Free subsequently collaborated with the International Labour Organization (ILO) and International Organization for Migration (IOM) to produce the first global estimate of modern slavery. This research found that more than 40 million people around the world were held in slavery in 2016.

Another Data for Good organization active in human trafficking research is Peace-Work. This all-volunteer cooperative of statisticians, data scientists, and other researchers applies analytics to issues in poverty, education, and social justice. Peace-Work projects are often in academic and policy research, with volunteers as likely to be found working with government economic data to write a position paper as working with a social justice organization. Projects have included education performance metrics, root cause analysis of homelessness, descriptive statistics of privilege, and the impact of racial bias.

Peace-Work’s efforts focus on human trafficking in the United States. State-level summary data published by Polaris was combined with demographic and socio-economic data and other sources. A mixed model was developed to identify economic, demographic, and other drivers of human trafficking in the United States. The number of victims reported by state was divided by population to yield per capita victim rates as the model outcome. These relative numbers are not able to estimate the number of people being trafficked; the goal of this project was to identify key drivers of human trafficking. Applying the model outcome to the statistics from smaller locations, such as metropolitan areas, can indicate where relatively high trafficking levels are expected.

Differences between individual states in legislation, administration, and the anti-trafficking efforts and organizations result in differences in reported levels unrelated to the amount of human trafficking present. The presence of these random factors in addition to variables driving the outcomes makes meta-analysis a good choice for the modeling method, identifying factors predicting a high level of human trafficking while accounting for variations between states. The fixed effects found by the state-level meta-analysis were then applied to data on large cities to yield potentially unidentified centers of human trafficking activity.

Peace-Work has begun to partner with local agencies to repeat meta-analysis of human trafficking at the level of metropolitan areas. Variations between states in the reported rate appear to be driven, in part, by differences in state laws. As a result, work has begun to advocate for legislative changes to implement best practices for finding traffickers and supporting victims across the country.

At the SAS Institute, a recent project used text analytics and machine learning to combat human trafficking. Starting with approximately 1,000 textual reports from the US State Department, SAS applied supervised and unsupervised machine learning to identify patterns in the text between source and destination countries, including types of trafficking and whether the countries involved were cooperating to help solve the problem. As described by Principal Solutions Architect Tom Sabo, the next step was to extract those patterns and present them in network data visualizations to share the insights gained with law enforcement and advocacy groups. Text clustering was also used to identify groups of words in the text that could be associated with trafficking children.

This study enables organizations fighting human trafficking to better leverage the intelligence to be gained from the State Department reports, in addition to finding key word clusters that can be used to examine other documents.

As much as the progress to date is encouraging, much more work is needed. The best efforts of statisticians and data scientists will continue to be needed in the fight against this terrible crime.