Artificial intelligence reducing the justice gap
Justice Connect, an Australian not-for-profit organisation that connects disadvantaged people to legal assistance and promotes social justice, has embraced natural language processing (NLP) – the branch of artificial intelligence (AI) concerned with giving computers the ability to comprehend language – to boost its efficiency and enable more people to gain the essential legal advice they need.
Working with a network of more than 10,000 pro bono lawyers, Justice Connect helps thousands of people each year who would otherwise not receive legal assistance in cases ranging from housing and employment issues to preventing elder abuse and escaping family violence. It also provides practical legal information via its online resources and tools and campaigns for change to address the root causes of harsh and unjust laws.
But providing such a service is not without challenges – every case is different and needs to be categorised and assessed so that Justice Connect can offer the most practical and effective assistance, decide which cases are most urgent, and assess which lawyers are best placed to help. Up until one year ago, sorting their case load was a complex and time-consuming manual process.
This is where Australian native Professor Timothy Baldwin, a world leading expert in large language models (LLMs), a Melbourne Laureate Professor at The University of Melbourne and the recently appointed Provost at Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) in Abu Dhabi, United Arab Emirates, stepped in to help overcome the challenge, immediately seeing the potential to devise and implement an NLP-based solution.
"Justice Connect offers a perfect example of how NLP – and AI more generally – can help to bring enormous benefits to society, empowering people to take control of their lives and protect what's most precious to them," Professor Baldwin said. "Having worked on NLP for many years, I was keen to develop practical applications with the potential to do good in society."
Tom O'Doherty, Head of Innovation at Justice Connect, became acutely aware, along with his team, of inefficiencies in the organization's digital intake tool, which had been deployed a few years earlier. After undertaking a thorough assessment of the tool, they realized that issues with the system meant that the charity was wasting resources and taking longer to help people than necessary.
"Our mission is to help some of Australia's most vulnerable people access legal advice – advice that is essential for their wellbeing, family and livelihood," Mr O'Doherty explained. "To achieve this, we always aim to ensure our systems are as efficient and effective as possible."
Professor Baldwin, who is also a Professor of NLP at MBZUAI – the world's first graduate-level, research-based AI university - has spearheaded several high-impact LLM products, including the world's most advanced Arabic LLM called Jais and LLMs for underrepresented languages.
"We went on to develop the models, integrate them and get really positive results," Professor Baldwin said. "Justice Connect's work is vital to the wellbeing of thousands of people in Australia, and it was quickly apparent that an NLP-based solution could be a game changer," he added. Aside from working with the not-for-profit sector, Professor Baldwin also champions eliminating bias and ensuring fairness in AI models and leads research aimed at providing language technologies for under-resourced languages to develop the NLP-based solution, Justice Connect gathered anonymized data from its intake tool and devised a 'tagging game' by which its legal partners – consisting of around 50 legal firms and 289 lawyers – would look at legal requests received by the charity, tag them, and select the most applicable category options from the intake tool. This refined data was then used to train the model how to triage incoming case work, and, with some supervision, it started to achieve remarkable results.
The NLP-based tool is capable of understanding help seekers' queries – covering the full gamut of potential legal requests – and directing them to the most suitable solutions.
The impact of the integrated model has been significant during the past 12 months, with incomplete entries down 50 percent and a 10 percent increase in service delivery, helping the charity increase its efficiency and accuracy, and enabling people to find the legal assistance they need faster.
"We have also gained valuable feedback from our paralegals who say that the cases referred to them via the AI tool are easier to deal with, thanks to the tool aligning the users' needs with the most suitable solution," Mr O'Doherty said.
The NLP-based system has worked so well that Justice Connect is now expanding it and enhancing the functionality to improve the segmentation of users, improve accuracy, and help the charity offer more effective 'self-help' legal advice via the website itself.
Justice Connect also sees potential to develop the technology for the wider judicial ecosystem by building an application programming interface (API) that is open to all not-for-profit organisations in Australia to build their own types of tools within their own platforms.
"If we can help other similar organisations to adopt this type of technology to increase social justice, that will be another great outcome," Mr O'Doherty said.
