Tech-Driven Tools To Uncover Labor Exploitation

Pauline Göthberg coordinates sustainable procurement for the Swedish government. Like many people passionate about ensuring that the items we buy are produced fairly, she’s been frustrated by the failure of traditional factory audits to uncover exploitation and even slavery.

This was very evident during the Covid-19 pandemic, with reports of continued abuses at the Malaysian factories producing medical gloves for the world. And even though in-person audits have long been flawed, it’s been especially hard to carry them out during the pandemic.

Göthberg believes that the current voluntary “tsunami” of standards and codes of conduct “has not really taken us that far in 20 years.” This voluntary infrastructure often depend on social audits, one of the main responses to the wave of public pressure over sweatshops since the 1990s. Göthberg is one of a number of people calling for a rethink of conventional practices when seeking to shut exploitative labor out of the supply chain.

The problems with social audits

In social audits, auditors visit factories to check on working and living conditions. Yet all too often they protect companies rather than workers, and companies are not transparent about audit results.

In any case, workers may be constrained in speaking frankly, or may understandably mistrust the intentions of the auditors. In one study of over 40,000 factory audits, spanning 12 countries, 45% were found to be rooted in falsified or unreliable information.

The report Hidden Harm: Audit Deception in Apparel Supply Chains and the Urgent Case for Reform, by the nonprofit organization Transparentem, documents many cases of deception during social audits in India, Malaysia, and Myanmar. Workers are coached to lie, for instance about having access to their passports when in fact these have been confiscated; or given falsified documents, for instance indicating that recruitment fees were within the legal limit rather than far higher. And safety conditions in work sites may be massaged to appear better during auditor visits than they are the majority of the time. For instance, one worker in Myanmar reported that her factory had soap and medicine only during audits.

As well, underage employees are routinely hidden away in order to evade policies restricting child labor. Some minors have been locked into rooms or made to hide under trucks during factory audits.

Given how rife practices like these are, it’s unsurprising that social audits have had limited effects in actually stopping worker exploitation. Import bans have been much more effective at driving change, for instance when US Customs and Border Protection has prevented the entry of gloves from Malaysian manufacturers associated with labor violations.

Still, not every dodgy shipment is going to be stopped at the border, and social audits remain necessary even despite their weaknesses. Göthberg reflects, “I’ve been very skeptical to social auditing, but at the moment, it’s the best method that we have. So what we can do is actually to improve the social audits.” This shouldn’t be just about doubling up on such audits. “As a public buyer, I’m not sure whether we should just put another layer of social audits. I mean, this is a million-dollar business…for governments and public procurement to go exactly the same way as the corporate sector does not make sense.”

In general, governments have a big role to play in limiting the exploitative labor that feeds their supply chains. Göthberg comments that in OECD countries, “within governments, procurement accounts for 12% of GDP. It’s a massive amount of money and governments are the largest buyers in many markets. So governments should really assure policy coherence and that they do protect human rights, labor rights, and the environment through their purchasing practices.”

Undoubtedly, there are ways to improve the accuracy of social audits that governments as well as other organizations depend on, such as involving local organizations rather than relying on auditors parachuted in from distant communities. And Hidden Harm lists some ways to more accurately assess worker ages, such as by asking young workers about their education history and family members’ ages, and reviewing multiple identity documents.

More broadly, one way forward may be to complement in-person audits with remote technologies.

New techniques

Göthberg believes that “there’s a lot of new techniques governments should start exploring, rather than social auditing.”

For instance, the Global Fund to End Modern Slavery has developed an open-source platform called the Forced Labor Automated Risk Estimator, which the organization claims is 80% accurate according to a pilot in the Indian garment sector.

And the Swedish nonprofit Globalworks uses machine learning to mine millions of social media posts, job postings, and other webpages for clues about working conditions. For instance, a Globalworks report on Apple’s supply chain in China scraped over 1 million social media posts and 1,000 job posts related to 32 manufacturers. This revealed many problems including violence and discrimination at Apple suppliers – problems that still haven’t been addressed over a decade after first being identified.

“We’re just small kitchen-table activists,” says Stefan Brehm, the cofounder and research director of Globalworks. Each member of the Globalworks team works part-time, and often pro bono. The funding they do receive is primarily from paid research projects.

The system they’ve devised, known as Social@risk, provides services to nonprofit and public organizations, rather than corporations. Brehm explains that this focus and methodology are by design: “We want to have unsolicited information.” Rather than research projects urging workers to fill out surveys or participate in interviews, Social@risk aims to capture unfiltered, unprompted information, which people are already sharing online. This is particularly important in spaces where they’re prevented from organizing.

Brehm refers to this as “online solidarity” – the quiet resistance of factory workers uploading photos of their cramped living conditions, 30-minute lines to get lunch, or documents showing illegal amounts of overtime. Social@risk scours not only social media sites, but also websites like online complaint forums and message boards for work sites of interest, such as forums for Foxconn factories.

Even in China, with its internet restrictions, workers find ways to organize online and share relatively candidly. “Workers are frank within the boundaries of what is permitted,” is how Brehm puts it. Some have actually posted photos of bodies of Foxconn factory workers who died by suicide, amid horrendous working conditions.

Social@risk can scrape posts in multiple languages, reflecting the different nationalities represented at each site. The Globalworks team can then track which topics in particular locations are mentioned most often, changes in these patterns, and relationships between words. For example, “intermediary” might lead to “fee,” then to “cheating” – suggesting how prevalent it is for middlemen at some work sites to charge inflated recruitment fees, often under false pretences.

These kinds of social media posts are freely available, and unlike private audits the data can’t be locked away in corporate vaults. But that doesn’t mean that they’re easily Google-able by researchers. “You have to scrape a large amount of information, and you have to make this somehow accessible to people who have the subject matter knowledge,” Brehm says. Labor and human rights experts aren’t necessarily working with big data, so Globalworks is trying to make the big data available to the subject matter experts, who can then provide the necessary context.

Though the Globalworks researchers believe in open data, there are limits. They don’t allow everyone into the full database because they want to protect worker identities, as well as adhere to their principle of only using information in workers’ interest. When they cite information, for instance, they refer to groups rather than individuals. Focusing on collectives also allows them to triangulate information.

At her procurement desk, Göthberg then uses this kind of service to assess the risk of forced labour in the factories that Swedish public agencies are considering using as suppliers. “This was a really easy technique to use, and quite cheap to be honest,” she says.

The limits of all audit tools

Of course, it’s hard to put a price tag on fair working conditions. But the reality is that the massive industry that’s developed around traditional factory audits hasn’t actually rooted out the problems.

Tech-driven tools are no silver bullet either. Brehm readily acknowledges this: “The problem that we saw is no matter how advanced and refined your technology is, it always will just perpetuate the same social order.” Restrictions on worker organizing will remain a core problem, which can’t be fixed by an algorithm.

Brehm stresses that the technology exists to challenge power structures, not as an end in itself. “It’s not really about technology in the end. It’s about making technology more accessible to social forces.”

And he notes of Social@risk, “It’s not the only solution.” For instance, there are certain topics that people don’t write a lot about on social media, such as child labor and sexual harassment. Brehm believes that “it’s not a universal tool, but it’s one that is actually very good for human rights due diligence because it helps you find priorities,” based on what workers themselves are choosing to write about.

Any one method of assessing working conditions isn’t enough on its own. As the project team of the Modern Slavery and Human Rights Policy and Evidence Centre comments, “No audit protocol alone is sufficient. There needs also to be an effective and transparent grievance mechanism, demonstrable track record of the effective resolution of grievances, and evidence of worker agency.”

Sadly, there are limits to any attempt to monitor working conditions. But with care and attention, procurers can reduce the likelihood of contributing to abuses. As Göthberg says, “No supplier can in their right mind sign a contract saying that there will never be any violations against human rights and labor rights. But what they can sign is having implemented proper due diligence processes.”

What’s harder to do, yet important for long-term improvements, is building relationships with suppliers and capacity to understand labor exploitation at all levels of the supply chain. Göthberg muses, “From my perspective, it’s very easy for us as public buyers to point fingers at our suppliers and tell them what to do. It’s equally important for us to identify and mitigate risks and build competence within our purchasing organizations. We need to increase resources and competence because we need to know what specific risks to look for. Otherwise, we tend to rely on these broad audit schemes and tick-the-box kinds of exercises.”

Follow-up matters too, and has been too often neglected, Göthberg says. “You have to include remedy in the contract requirements. Before we were just requiring corrective actions if any violations were found, but not ensuring effective remedies for persons whose human rights had been violated.” For instance, workers who have been found to have paid excessive (and illegal) recruitment fees can be reimbursed for these as a start.

Overall, Göthberg emphasizes, sustained change will come only with mandatory regulations. “We should not just rely on these voluntary rules and measures for follow-up and counting on free market principles to solve the defining crises of our time. I don’t think it works. It hasn’t worked.”

So digital tools like Social@risk can’t make up for the hard in-person grind of ensuring that factory labor is just, and enforcing legislation that protects workers. But this next generation of tools does provide a useful supplement, allowing a broader peek into what’s going on in the factories that make so many of our belongings.