Future Tense

A.I. Could Help Combat Modern Slavery, if Humans Don’t Mess It Up

Nepalese labourers work in a brick kiln in Lalitpur.

Prakash Mathema/AFP/Getty Images

Though it’s been over it’s been more than a century since the end of the trans-Atlantic slave trade, the practice of using forced human labor has proven to be a stubbornly modern problem. According to 2016 estimates by the Global Slavery Index 45.8 million people are currently enslaved worldwide.

Despite clear laws banning such exploitation, these numbers have remained high as practices continue to evolve. One of the biggest obstacles those who want to combat forced labor face is the difficulty identifying and accessing the places where it’s happening—which are often in remote or unstable areas.

But now some human rights activists may have a new tool to track some of the most notorious sites of slavery in the world: artificial intelligence. The technology holds promise to vastly expand and accelerate their important work. However, as experts point out, its implementation doesn’t come without potential pitfalls.

One leader in the A.I. fight against slavery is the University of Nottingham group Slavery From Space. SFS recently launched a program to help train A.I. technology to identify satellite images of brick kilns, which are common sites for slavery in South Asia. Many of the laborers at these kilns are debtors and their family members, who pledge to work in hazardous conditions for undefined periods of time as part of an exploitative repayment scheme. SFS estimates that 68 percent of the 4.4 million to 5.2 million brick kiln workers at these sites are not paid.

Such technologically aided tracking could be a big breakthrough for anti-slavery organizations that work in the region. These groups have already identified broad swaths of Pakistan, Nepal, and India—nicknamed the Brick Belt—where these kilns operate. The kilns are easy to spot in aerial photos, so the team has been using satellite images to pinpoint their locations. But they’ve been slowed by the vetting speeds of human volunteers. By manually mining the enormous set of images from the region, they’ve only been able to catalogue a fraction of the estimated 20,000 to 50,000 sites. Now, however, SFS aims to have a computer do that painstaking work. And to do it faster. They plan to use volunteer-classified images to train a machine-learning algorithm to flag similar iterations of the kilns’ distinct shape and color.

With an A.I. to automatically collect and share kiln locations, SFS hopes it may be able to pass this information to organizations that can use the information to step in and investigate the sites. According to its website, SFS also plans to share the data publicly to help academics, policymakers, and humanitarians measure and respond to the prevalence of this modern form of slavery.

Certainly, it’s a laudable goal. But Faine Greenwood, a Future Tense contributor and assistant researcher at the Harvard Humanitarian Initiative who investigates how drone technology and satellite imagery can be used in humanitarian contexts, believes that the public nature of the data could harm, rather than help, the enslaved.

In an email earlier this week, Greenwood pointed out that if the images are published, there would be nothing stopping kiln owners from using the website to access the imagery themselves and determine if their site has been identified. In doing so, they may be alerted to the fact that they’ve been identified, buying them time to transfer their laborers or otherwise further obscure their activities. Whether volunteers and others who want to access the information are vetted before sharing it, she said, should be considered.

“It’s dangerous to assume that only well-meaning people will view this imagery,” she wrote.

Greenwood also said that she has concerns about overreliance on AI. In some ways, she said, human volunteers can be more useful because they’re less likely to miss subtleties and outliers that a machine may not catch. Technology may help with the process, but we also need to take account of its risks and limitations.

Though we’re just beginning to leverage the full potential of artificial intelligence, it’s promising to see the ways in which human rights advocates are already thinking about how to use the technology as a tool for good. SFS’ efforts seem to be on the right track. But let’s hope they—and others—avoid the pitfalls that would make them a cautionary tale.