Code 8.7: How We Can Advance Collaborative Problem Solving
Computational science, artificial intelligence and machine learning have huge potential for accelerating the fight to end modern slavery and achieve Target 8.7 of the Sustainable Development Goals. Realizing that potential will require a cultivated mix of bottom-up innovation, lateral learning and strategic coordination—including with funders and donors. Without careful, strategic and coordinated action, the danger is that we will fail to exploit the real and significant opportunities to which we were all awoken at the Code 8.7 event in New York in February 2019.
Code 8.7 Day 2 at the United Nations.
From a community of interest to a community of intent
The Code 8.7 event made clear that there is now a convergence of interests between the anti-slavery community and the computational science community. That convergence reflects the overlap between the anti-slavery community’s interest in access to computational methods, tools and techniques and the computational science community’s interest in using and generating new data sets and research problems derived from real-world applications. But it is also a product of shared social motivations and an underlying commonality of intent to use these tools to solve social problems.
The challenge now is to transform this emerging community of interest into a community of intent. This requires building a framework for collaboration that creates incentives for innovation while achieving collective impact.
The good news is that we are not starting from scratch. Code 8.7 demonstrated that there is already a great diversity of basic research and operational experimentation bringing AI, machine learning and computational science into the anti-slavery field, using a variety of data sets, covering everything from remote sensing image analysis and financial transactions analysis to survivor rehabilitation case management and vulnerability mapping. There is, however, a gap between basic research and operational experimentation, on the one hand, and, on the other, the development of applied technologies at scale, delivering strengthened global understanding of modern slavery and what works to address it. The “middle of the technology pipeline is missing”, as one participant put it.
The Heilmeier Catechism, used by DARPA to figure out which research and development investment risks are worth taking, provides a powerful framework for approaching technology pipeline development. And on the more operational side, tech companies have a track record of building and scaling technological solutions. Harnessing their know-how may help accelerate progress towards this objective. As some participants noted, thought may also need to be given to coordination of funding streams to ensure the full length of the research and development pipeline is funded, covering not only government funding for basic research, but also private funding for the translation and application of basic research findings.
Still, as several participants cautioned, there were limits and downsides to the potential of AI, machine learning and computational science. We must not adopt an uncritical technological utopianism.
For starters, there is a persistent and fundamentally constraining lack of standardization in this field, even on basic technical issues such as how to describe and code forced labour, modern slavery, human trafficking and child labour. Progress in this field will require that some of the recent gains in statistical research, such as the development of a taxonomy of exploitation embedded in the most recent ILO Global Estimates of Modern Slavery, and in the recent ICLS guidelines on measuring forced labour, are translated into technical data and coding standards.
Second, there remain clear and persistent barriers to information and data-sharing. These include technological and infrastructure gaps; regulatory and legal barriers, some arising from efforts intended to ensure data protection and the right to privacy; and, even more fundamentally, a lack of trust on the part of survivors and frontline civil society actors that data will not be monetized or mishandled. Participants stressed the urgent need for solutions to these problems in the short term. At the same time, some computational science participants stressed that while such solutions might be found in the short-term, even better solutions might be available in the medium to long-term, if actors in this field were prepared to invest time and resources in research on these questions.
Third, an ongoing lack of strategic coordination among donors, funders and resource-allocators means there is likely to be significant inefficiency, duplication and funding gaps in current efforts to bring technology, AI and machine learning into this field. We say “likely” because in reality we do not know, with any accuracy, how funds are spent in this field. Almost four years after the adoption of the SDGs and several years after Alliance 8.7 was created to provide a global coordination platform, we lack a clear picture of global spending and activity to achieve Target 8.7.
The Organizing Committee for the Code 8.7 event walked away with the perception that Code 8.7 could be developed into a framework for collaboration that met some of these needs and addressed some of these obstacles. Here, in closing, we set out in broad brushstrokes what that framework might involve, seen from two different angles.
Code 8.7 as a strategic framework for applied research
The community of interest that coalesced around the Code 8.7 event could, with some encouragement, develop into a community of intent. Keeping survivors front and centre in that discussion will be important; and it may be necessary to bring in additional stakeholders, such as more representatives of the private tech sector. But the central challenge is to define and record our shared intent, and turn it into a strategic framework.
Beyond “accelerating progress towards Target 8.7”, what is the intent we share? Answering this question, through a structured process intended to scope the participating stakeholders, the flow of inputs—funds, tech capabilities, research capabilities—and expected outputs will allow this community to develop a shared purpose. Within the Organizing Committee, CCC and Turing may be particularly well placed to convene sessions or processes intended to crystallize this shared intent. And there may be scope for such a process to connect in to other ongoing processes, such as the policy research agenda formulation exercise that ILO will lead within a US Department of Labor-funded project exploring future research inputs to Delta 8.7.
Centrally, this process should also allow the Code 8.7 community to clarify not only the kinds of research questions it is seeking to address, but also the conceptual and operational hierarchy among them. Which questions or operational issues need to be addressed first, to enable forward progress on the others? Answering this question will help Code 8.7 translate a shared strategic objective into a shared strategic framework, with sequencing and prioritization of interim milestones towards the ultimate goal.
Listening to discussions at Code 8.7 in New York, the Organizing Committee tentatively identified the following three dimensions of a strategic framework. It intends to explore these three dimensions and consider at an upcoming meeting how they might be translated into a public, timebound strategic research framework.
Dimension 1 – Data standards and collection frameworks
In an initial phase, Code 8.7 activities would focus on rapidly building out shared data standards and collection frameworks to allow participating members to rapidly scale-up data-related activities. This may need to cover both technical norms (such as coding typologies) and data-sharing infrastructure, potentially through the establishment of a “data trust”. There may be lessons to be learned here from related fields, notably public health, and there will be a need for engagement with survivors, and legal and ethics experts. Where possible, efforts along this dimension would build on and scale up existing, trusted initiatives, such as CTDC’s work on data standards, or other emerging data-sharing initiatives.
Dimension 2 – Technology pipelines
Simultaneously, or potentially in sequence, Code 8.7 would provide a framework underpinning activities by members developing new technology pipelines. This could include framing and inception workshops; collaborative research design and implementation; and research evaluation. We heard four possibilities coalesce during the two days of Code 8.7:
- Global Slavery Observatory – existing initiatives, such as the Slavery from Space project, others involved in imagery analysis, Delta 8.7, the Brazilian Slavery Observatory and CTDC, could work together to develop a Global Slavery Observatory, offering close-to-real time mapping of incidence and dynamics;
- Case tracking and management – there were strong signs that AI, machine learning and computational science could radically transform case tracking and management in the anti-slavery field. There appears to be particular scope for AI-assisted case management software to nudge decision-makers towards practices and outcomes that are more effective (whether seen in terms of outcomes for victims and survivors, likelihood of conviction of perpetrators, asset recovery or some other systemic factor).
- Risk modelling and mapping – Target 8.7 risk modelling is undergoing something of a revolution, drawing on new statistical techniques, and fuelled by the private sector’s growing need for supply-chain risk data, not least as a result of supply-chain due diligence legislation in California, the UK, Australia, France and beyond. A technology pipeline focused on this area could help ensure experimentation in this area converges in ways that deliver systemic value, such as interoperability with prevalence measurement analysis or case management.
- Financial data analysis – another boom area, we see a rapid rise in the use of computational science in banking, remittance and investment transaction data to understand risk and returns, manage down risk and identify value investment opportunities. Again, a technology pipeline focused in this area could have systemic benefits.
Dimension 3 – Research mapping and monitoring
Finally, a central element of a shared research framework would need to be both a mapping of research opportunities and a system for monitoring progress over time. This could, for example, begin with a research-focused white paper, identifying different research questions and problems and different strategies for addressing them over time. Numerous problem-sets emerged already from this two-day conference, including around data fusion to deal with limited ground-truths; network analysis; natural language processing, especially of unstructured data, to identify causal pathways related to modern slavery. This Dimension might also consider the development of research and research impact metrics, and a system for periodic review of research developments.
From left to right: Laura Hackney (AnnieCannons), Jessica Hubley (AnnieCannons), Kevin Bales (Rights Lab), James Cockayne (UN University Centre for Policy Research), Anjali Mazumder (The Alan Turing Institute), Dan Lopresti (Computing Community Consortium), Zoe Trodd (Rights Lab), Hannah Darnton (Tech Against Trafficking).
Code 8.7 as a human learning community
Whether or not the Code 8.7 community is able to translate its shared interest and intent into a formal, public strategic framework for applied research, we saw strong interest at the New York event in Code 8.7 continuing to support mutual learning. Code 8.7 might have been about artificial intelligence and machine learning, but we left the event with a strong sense that it is natural intelligence and human learning that is central to this enterprise.
There are numerous ways that Code 8.7 could serve as a vehicle for the development, publication and reporting on research, operational and analytical work on these issues, whether or not that is reduced to a formal strategic framework.
First, with a small secretariat, Code 8.7 could organize membership events. These would:
- encourage information sharing and lateral learning;
- allow the deliberate development of a structured and progressively-developing discourse on these issues. Code 8.7 could, for example, publish analytic reports, proposals or joint research reports on specific issues, to be debated at membership events;
- provide a venue for teaching;
- help maintain and strengthen the social capital in this network.
Second, Code 8.7 could play a match-making role, connecting demand for computational science, AI and machine learning expertise within the Target 8.7 community to suppliers of relevant expertise. A centralized match-making or clearing-house function would reduce the costs of search for those in the market, foster interest from researchers and students, and, importantly, allow the secretariat to develop a sort of “market trends” analysis, which will allow regular insight for the community about how collaborative research and problem-solving are developing.
Third, Code 8.7 could support joint research and fund-raising by its members.
Fourth, Code 8.7 could foster the development of good practice insights and standards development. Initially, this might be linked to the development of technical data standards, as explored in Dimension 1 above. In time, this might develop into a broader norm-development or norm-dissemination role, for example allowing the promotion of standards developed elsewhere, such as the Worker Engagement Supported by Technology (WEST) Principles (focusing on technology-driven efforts to address human rights abuses in supply chains) or the Ethical OS Toolkit (intended to help technologists think critically about the different ways their proposed technology influences human behaviour and interactions). A soft norm-development role for Code 8.7 would contribute to strengthened interoperability, comparability and automation in Target 8.7 data initiatives; help strengthen transparency and accountability in those efforts; and foster the protection of survivor rights.
As survivor leader Sharlena Powell reminded us all as the Code 8.7 event in New York closed, whether and how computational science, AI and machine learning make a real difference to efforts to achieve Target 8.7 of the Sustainable Development Goals depends on us, not the technology. Code 8.7 may have helped to awaken some in the anti-slavery community to the potential gains, and some of the risks involved. And it may have awoken some in the tech community and the computational science research community to the urgency of the issues addressed by Target 8.7. But if we do not “stay woke”, as Sharlena Powell exhorted us all, not only to the opportunities but to the risks, this positive potential will pass us by.
The Organizing Committee intends to work with interested parties in the months ahead to figure out a framework for moving forward. As Sharlena’s exhortation should also remind us, survivors need to be central to this exercise. We will need active participation from researchers, the tech community, anti-slavery actors and governments. And funding will be a real question, as it always is. The challenges are real. But the potential from closer collaboration between the anti-slavery and computational science communities is far too significant to allow that to deter us.
This article has been prepared by the authors as a contribution to Delta 8.7. As provided for in the Terms and Conditions of Use of Delta 8.7, the opinions expressed in this article are those of the authors and do not necessarily reflect those of UNU or its partners.