The tech and the good: Behind a tech for good start-up

In light of the UN Sustainable Development Goal to end child labour by 2025, many businesses and investors are now looking to address instances of child labour within their supply chains. However, to be able to do this effectively, they need access to a range of information and data, and for that data to be reliable, usable and up to date. According to the International Labour Organization (ILO), over 160 million children aged 5-17 years old are estimated to be involved in child labour globally. This figure is rising, so it’s never been more important to take action on this global issue.

Although this data does exist in various guises and from a multitude of sources, it’s often presented in a way that makes taking actions and decisions from it very difficult. General statistics don’t offer enough detail around why children are driven into labour, or information about what work they’re doing – which makes them almost impossible for organisations to take any positive action from.

HACE founder Eleanor Harry realised that there was a real need for investors and organisations to be able to easily access all of these data points from a single source, but that this need was, at the time, not being met. So she set out to create exactly that.

The idea was strong, but the process was difficult. Team members at HACE spent a long time manually collating and analysing data sources to ascertain the risk of child labour appearing in specific supply chains, but it quickly became evident that this task was bigger than humans could undertake alone. They needed to automate this process in order to scale it to the extent that it could be comparable, usable and reliable.

So HACE enlisted the help of Equal Experts, who quickly stood up a team to help HACE achieve their goals.

“I really love the project. I love the cause of eradicating child labour, it’s really interesting. And the project itself, from a technology point of view, was really challenging, and because of that, really interesting. I had to work with some technologies I’ve never used before…I wanted to use something that was really good for HACE.” – Pedro Sousa, Software Engineer, Equal Experts, Portugal

HACE had already identified three main areas which would create a complete picture of the company’s performance on the issue of child labour. They also had a clear view of how these pieces of information could be aggregated and analysed to create an overall score.

1. An organisation’s Company Disclosure

The first area of interest is the organisation’s Company Disclosure – the publicly available data a listed company displays on their website. This is a statement that demonstrates how well a company identifies, manages and mitigates the risk of child labour in their supply chain. This presented some challenges, as organisations aren’t required to use standardised formatting or language to present this information; it could be a web page, a PDF, or embedded in an image.

The Equal Experts team worked to create an algorithm that can locate and scrape this information from whatever location and format it’s displayed in. It then gets searched for keywords, analysed and allocated a score based on the content of the disclosure, and the quality of it with regards to child labour.

2. An organisation’s Public Perception

The second area is Public Perception. This score is based on how a listed company is perceived to be associated with child labour by the general public and therefore, a company’s potential consumers. To create this score, we needed to be able to locate, gather and analyse news sources and articles to ascertain the sentiment of these publications (are they negative or positive stories, which businesses or supply chains these organisations are linked with, etc.) We used a mixture of existing software and bespoke functionality that we created to be able to create this score.

The technology used to scrape websites is nothing new, but we needed something bespoke to measure the sentiment within an article. We harnessed AI and LLMs (large language models), and wrote specific prompt engineering to suit HACE’s needs. This technology can ingest historical and new data regularly, analyse the sentiment of the language used, and then assign that content a score. It then aggregates those scores along with any other existing scores from other articles to produce a final output. It can analyse multiple languages and can be set to scrape for new content at any level of frequency.

3. An organisation’s Supply Chain

The third area is Supply Chain – this score is based on how susceptible a listed company’s supply chain is to child labour, based on the sector and geography they operate in. It incorporates data from externally assigned sectors via a stock ticker, and combines it with data on commodities that have been proven to be produced with child labour.

These 3 scores are then aggregated, which gives each organisation a final score to be published in the index.

“The best thing I could do for HACE was to make sure there was a good EE team on it. Everybody who’s been involved in it loves it. It’s a great thing to be involved in, technically it’s really interesting, and obviously it’s for a great cause. I think the team has done an incredible job. It was a lot of functionality to get in in a short period of time.” – Simon Case, Head of Data at Equal Experts

We then had to figure out the best way to contain and display this information to users. The chosen technology was challenging, due not only to the complexity of the software, but also the time constraints we found ourselves under. Using Test Driven Development (TDD) the team were able to build a solid and functioning platform, that also acts as an independent API, within 3 months. We built the platform on AWS (Amazon Web Services cloud) using services including Amplify, OpenSearch, S3, ECS and DynamoDB – with AWS providing some funding towards HACE’s initial Proof of Concept.

The testing and validation of this data has been rigorous and thorough. There are automated tests developed specifically for HACE. Then there’s manual validation, where specific scenarios are put through the algorithms and the results are compared with an expected outcome. There’s also a working group of leading academics in fields such as human rights, modern slavery supply chains and ethical AI who are consulting with HACE on the social sciences behind this data.

“It was a joy to work on something bigger than myself, commit to making a difference, and develop an impactful solution freely. These are the ingredients of one of the most rewarding experiences of my life.” – Gerhard van Deventer, Senior Data Engineer, Equal Experts, South Africa

The credibility and reliability and transparency of these scores are of the utmost importance – for HACE, for the end users of the index, and crucially, for the children involved in child labour. As a tech for good organisation, HACE has a responsibility to the investors they serve and the children and communities they represent to provide valid and reliable data, and it’s a responsibility we all take incredibly seriously.