Report on using ethical AI to support modern slavery survivors and improve policy outcomes in the UK. First published on the Modern Slavery Policy & Evidence Centre website
Based on Project RESTART (The Reporting Experiences of Survivors to Analyse in Real-Time), a research project conducted by Aberystwyth University (Dept. of Law and Criminology), FiftyEight, Trilateral Research and Causeway. The project was funded through an open call for proposals by the Modern Slavery and Human Rights Policy and Evidence Centre (Modern Slavery PEC), which in turn is funded and supported by the UK Arts and Humanities Research Council (AHRC).
Background
Since the introduction of the National Referral Mechanism (NRM), the UK’s dedicated framework for identifying and supporting survivors of modern slavery, conversations have focused on its effectiveness in meeting survivor support needs. Arising out of the recognition that the UK still falls short of adequately protecting survivors of modern slavery, project RESTART sought to provide a proof of concept for a new, and more effective, method for understanding survivors’ needs.
The project used Natural Language Processing (NLP), an AI technology, to analyse large and complex data sets held by Causeway, a charity that supports survivors of modern slavery, that would otherwise remain underutilised due to resource constraints. Survivors were consulted on the design and use of NLP through a Lived Experience Advisory Panel.
The project also sought to integrate survivor voices in research by enabling survivors to actively document their individual experiences and assess their own needs and goals via a smartphone app (called MeL) over a four-month period. Subsequently, the data generated by survivors underwent analysis using NLP techniques.
Key findings
On the use of AI technologies to analyse modern slavery related data sets:
- Whilst substantial resources are initially required for training and validating NLP models, once established, data can be analysed rapidly. In the long term, investing in NLP is both efficient and effective for stakeholders seeking deeper insights into modern slavery and human trafficking, and its impact.
- Natural Language Processing is an effective tool for uncovering insights from textual data, such as survivors’ stories, and can identify the types of support needed by survivors of modern slavery. However, it may be less effective when analysing culturally specific language. Human involvement is crucial in supporting survivors by overseeing the insights derived by AI technology, determining the ‘so what’, and providing direct support.
- Survivor needs identified using NLP are consistent with the needs identified by conventional methods. However, NLP enables this identification to be conducted at scale and in real-time.
On the use of mobile app technology in delivery of support for modern slavery survivors:
- Access to an appropriate app yields benefits to survivors of modern slavery by providing a space where they can record their recovery journey, and thoughts and feelings throughout it, independently. The MeL app’s distinction between immediate needs and long-term goals further encouraged app-users to reflect on their recovery in a more holistic way.
- The potential therapeutic benefits of journaling within a suitable app, particularly in contexts where access to sustained, formal mental health and well-being services is limited, hold significant promise for survivors of modern slavery and human trafficking. The MeL app provided avenues for emotional processing through journal entries, enabling survivors to freely express and record their thoughts and feelings at any time.
- Maximising survivor engagement with app technology requires suitable assistance. This was evident in participant’s initial hesitancy toward using the app, which was subsequently overcome with the assistance of a dedicated Participation Facilitator. As one participant expressed, “If you told me to use an app I wouldn’t know how and wouldn’t have the confidence to. But the information they gave, it gave me confidence and I did it.”
Key recommendations
To the UK Government:
- Incorporate, with associated funding, the use of AI technologies such as NLP across UK Home Office and First Responder agencies as a means of identifying fluctuations in modern slavery trends and survivor support needs in real-time, with a view to sustained enhancement of support measures and mechanisms. The Independent Anti-Slavery Commissioner should consider facilitating and overseeing the amalgamation of these diverse data sets as part of her role in supporting research.
- Training should be introduced at the national level to ensure all statutory and state-funded support services collect and record data in line with strict data privacy and protection measures and in a uniform, consistent manner. This would better enable efficient analysis and anonymisation by AI technologies.
- Subject-matter experts (including lived experience experts) should be involved in the development and vetting of any use of AI tech and other technologies to ensure they are customised to respond to the complexities related to modern slavery.
- Access to mobile devices and internet data packages should be made available through state-funded survivor support services, and the use of apps to help survivors to manage their needs and goals should be encouraged.
For UK Practitioners:
- Education and upskilling programmes should be integrated into survivor support services and should enhance tech literacy. Such programmes should include support for survivors wishing to participate in consultation activities and mentoring opportunities.