Modern slavery in global supply chains is hard to measure. It’s not always visible and it covers a range of different forms of exploitation that can develop and change quite quickly. We need to understand it better, and to have better evidence about what investors and businesses can do to help address it. Machine learning tools have the potential to help find the right data, combine it in useful ways, and produce analysis that can help investors and businesses make better decisions.
The Modern Slavery PEC, in collaboration with CCLA Investment Management and the Finance Against Slavery and Trafficking (FAST) initiative, recently co-hosted a roundtable discussion on this topic. It included experts from several different sectors, highlighting challenges but also bringing out ambitious and hopeful thoughts for how we might move forward.
Here are some of our takeaways from the event.
- Among the 48 participants at the roundtable were investors (both asset owners and asset managers), businesses, civil society organisations, academic researchers, policymakers and AI specialists. Every one of those groups has a piece of the puzzle and only in collaboration can we make progress on using AI to analyse modern slavery risk in supply chains and business action to address that risk. Figuring out how to gather and analyse modern slavery data, and how to use it to inform real-world decisions, requires an unusual combination of practical and technical expertise that goes well beyond any of these individual groups.
- But it’s not just the importance of a diverse range of experts - we also need new evidence, and new research to provide that evidence. We need evidence on what businesses are doing, of course, but also on the risks of modern slavery across particular regions or sectors from where businesses buy their goods and services. We need evidence on the actions of industry groups and NGOs to address those risks, and also on what government regulation exists and how effective it is.
- AI can help both in gathering relevant data and also in combining and analysing it in innovative ways. But there are challenges: incomplete or hidden data, difficult to access or in formats that are not machine-readable, for example. And can AI be taught to deal with what’s not there, drawing conclusions from gaps in the data and things not said? We must also be aware of potential bias: AI tools are only as good as the way in which they have been designed and trained. Expert oversight of such tools is vital.
- Research to begin answering these questions needs to bring together expertise on modern slavery and on machine learning techniques. It must also be done in partnership with the investors, businesses, NGOs and others who understand the practical challenges of gathering and using data to address modern slavery. For researchers, collaboration offers access to practical experience and knowledge, but also the opportunity to have a real influence.
What this roundtable made clear is that the expertise and the commitment are there, across many sectors, to push forwards and make an impact. This conversation was only a first step, and we’ll continue to work with others to develop the next steps on this important issue.