In 2021, IRCAI launched its first AI Award campaign. Aiming to highlight the work of individuals addressing the UN SDGs by means of artificial intelligence (AI), IRCAI strongly encouraged nominations from all world regions and all backgrounds. Upon a careful selection procedure, this year’s award has been granted to Adriana-Eufrosina Bora (PhD Candidate at the Centre for Data Science School at QUT) for her work in Project AIMS. Striving to eradicate modern slavery, the project uses AI to read and benchmark companies’ statements.
Each year, about 16,000 to 17,000 modern slavery statements are published by UK businesses alone. This can be quite a tiring process: It takes approximately 1 hour for volunteers to manually analyze one statement and for the researchers to validate their results. To respond to the need of accelerating this analysis and creating accountability at scale, Adriana has been gathering resources required to build her AI-based over a period of six years. In the meantime, Adriana and The Future Society were taken onboard by the Walk Free Initiative to boost business compliance and eradicate modern slavery, by jointly automating the analysis of modern slavery statements.
The analysis of the statements is basing itself on the UK Modern Slavery Act. Adapted from the requirement set formulated in the act, the project considers a list of 16 metrics. Each of these metrics refers to a question about the quality of the companies’ modern slavery statements. Questions include: Is the statement signed? Is it approved by the Board of Directors? Does the company have a mechanism to facilitate whistleblowing or the reporting of suspected incidents of slavery?
The team behind the project notes that the analysis proceeds in two steps: In the first step, the corpus of the text is extracted from the modern analysis (usually documents stored in a PDF or HTML format). In the second step, the methodology consists in bringing along the “understanding needed to prepare the training of the machine learning (ML) solutions”. In other words, the second step is where the analysis of the 16 metrics and the identification of the supporting statement segments takes place.
Whilst trying to come up with an effective computational method, the team behind the AIMS project invited a consortium of modern slavery research experts to a semantic workshop. Striving to imbed the experts’ knowledge in the methodology, the experts formulated a set of keywords, which were then used to facilitate the identification and exploration of the metrics in the corpus.
To present her work to a wider audience, Adriana will be giving a keynote speech at the IRCAI AI Award presentation on March 17th at 2pm GMT+0. Her keynote will be followed by a panel discussion featuring Adriana’s colleagues and partners from The Future Society, Walk Free, Mila and WikiRate. The event will be broadcast live on Zoom or YouTube Live. The YouTube live recording will be available later on IRCAI’s YouTube channel as well.
For more details on the statement analysis methods employed by Project AIMS, feel free to have a look at the paper “AI against Modern Slavery: Digital Insights into Modern Slavery Reporting – Challenges and Opportunities”. Furthermore, a technical project evaluation has been conducted by Mila, which is available on request.
