AIMS built the first exploration of the application of machine learning to automate the analysis of statements produced by businesses under the UK Modern Slavery Act to boost compliance and help combat and eradicate modern slavery. Project AIMS is now open source on GitHub!
Project AIMS (Artificial Intelligence against Modern Slavery)
March 17, 2021
Since 2015, the UK Modern Slavery Act (MSA) has required companies making over 36 million £ (approx. USD 45 million) per year to report on the steps they are taking to eradicate modern slavery from their direct operations and supply chains. Similarly, under the Australian Modern Slavery Act, since December 2020, Australian companies making over AU $100 million (approx. USD 69 million) per year have also been required to produce an MSA statement. According to the BHRRC Modern Slavery Registry, there are now over 16,000 published MSA statements in the UK.
The Future Society (TFS) launched Project AIMS (Artificial Intelligence against Modern Slavery) in partnership with Walk Free. Project AIMS built upon the work from BHRRC Modern Slavery Registry, WikiRate, and Walk Free, to develop methods and tools to ‘read’ the statements produced by companies under the Modern Slavery Act. Some of those methods used metrics designed by Walk Free in line with the UK Government guidance. This work can support that of WikiRate and others interested in analyzing those statements.
Project AIMS uses the latest data science and machine learning techniques, including NLP (Natural Language Processing) and computational linguistics, allowing for deep dives into each report, assessing and improving the overall understanding of the quality of the reports produced under the UK and later Australian Modern Slavery Acts.
This project builds on research conducted by Adriana Bora’s “Using augmented intelligence in accelerating the eradication of modern slavery – Applied machine learning in analyzing and benchmarking the modern slavery businesses’ reports” (2019; unpublished manuscript). Bora focused on creating the first global analysis of all the businesses’ reports collected under the Modern Slavery Registry.
Project AIMS is now open source on GitHub
The first phase of Project AIMS has now ended and this work is now available open-source on our GitHub. The GitHub repository contains the notebooks developed under Project AIMS from April 2020 to June 2021, clearly documented, for anyone who would like to use, reproduce or develop further the project. Those notebooks are organized in the following format:
- 🗄️ Data and text extraction
- 📔 Initial Metrics Exploration
- 📔 Model for multi-class and multi-label classification for core metrics
- 💡 Sub-projects for multi-class and multi-label classification for core metrics designed for the community to pick up and develop further. Those include a clear scope, questions, expected outcomes, a proposed methodology, and resources.
- 🔎 Other explorations on the modern slavery corpus
We would like this work to continue, so we invite you to join us in taking this project to the next phase. We hope you stay in touch, join our community, and share this opportunity with your network, finding those interested in using their skills to fight modern slavery.
Once developed, the tool will help increase efficiency and transparency for consumers and businesses, while being aligned with the best practices in terms of ethics by design.
Project AIMS in the media and more:
- Walk Free Foundation Press Release.
- UNESCO IRCAI AI Award 2021 – Adriana Bora for Project AIMS
- Publication: Digital Insights into Modern Slavery Reporting: Challenges and opportunities of machine readability
- DATA FOR FUTURE podcast episode #17: AI AGAINST MODERN SLAVERY – with Adriana Bora talking about AI governance and the conceptualization of Project AIMS
- Adriana Bora presented at the United Nations World Data Forum 2021: Measuring and combatting modern slavery with data-driven solutions in times of a pandemic.