AI dispute resolution: is a robot coming for your divorce?

'David Mazzone FCPA brings problem-solving skills and a keen sense of justice to his role as COO of the Legal Services Commission of South Australia. Photographer: Lightly Salted.

David Mazzone FCPA is pioneering the use of AI (artificial intelligence) to ease some of the pain of separation and divorce. AI takes information entered by users and compares it against family law cases to provide a starting point for dispute resolution.

By Carolyn Boyd

From the get-go, government appealed to David Mazzone FCPA. Not necessarily the politics or the cut-and-thrust of the daily barbs thrown in parliament, but rather the ability to have a real impact on people’s lives, especially when a little thinking outside the square was required.

Now the COO of the Legal Services Commission of South Australia, Mazzone heads a national project that promises to lessen the heartache for thousands of people and save them large sums of money at one of the most difficult times in their life. Using A$581,000 seed funding from the Australian Government, Mazzone is overseeing the design of an online portal that employs artificial intelligence (AI) to guide former couples through the formal separation process.

The groundbreaking project is being developed in South Australia as a prototype that will learn and improve as more people use it. It will also draw from a database of Family Court of Australia cases and consent orders. The latter will initially be inputted manually, but with more than 105,000 family law disputes in Australia each year and couples spending A$30,000 to A$100,000 to reach a solution, the tool’s potential is enormous.

Given that family law is so complex, Mazzone says if the tool – designed to be a simple-looking, user-friendly mobile phone app on rollout – is successful it could be adapted to other legal areas.

Accounting and law

Mazzone isn’t a lawyer – his degree is in commerce. However, he does have an accountant’s strong problem-solving skills and a keen sense of justice. He was a board member of the Legal Services Commission for five years and spent more than a decade working in the South Australian Attorney-General’s Department.

The commission employs 220 people, essentially making it South Australia’s largest legal firm. It briefs out 70 per cent of its work to private legal firms at a cost of A$19 million.
Mazzone stepped down as acting CEO at another, much smaller state governmentfunded body – South Australia’s former Motor Accident Commission – to take up the COO role. 

While he has had to “shelve the title” of being a CEO, Mazzone says he also weighed up the operational experience he could gain from moving to a bigger organisation to oversee an ambitious IT program.

David Mazzone FCPA brings problem-solving skills and a keen sense of justice to his role as COO of the Legal Services Commission of South Australia.“For me it was very much about ‘does this role offer something that is maybe a little bit different?’,” Mazzone says.

He has spent his career working in his home state of South Australia and resisted moving from Adelaide to bigger cities or overseas, partly because he is so close to his large Italian family and wanted to stay near his father, who was ill for many years.

When Mazzone worked with the Motor Accident Commission from 2013-2017, a campaign it undertook was instrumental in informing the development of the new online dispute resolution tool. It aimed to reduce the incidence of parents driving under the influence of drugs; a small but worrying problem with potentially devastating consequences.

“They weren’t your stereotypical drug users, they were mums and dads dropping their kids off to school getting caught with drugs in their system,” Mazzone explains. 

“We were trying to understand why people were doing that. It was a foreign concept to us because we didn’t engage in it. You can’t change behaviour without speaking to people.”

Research revealed a misconception among drivers that drugs stayed in their system for up to two weeks, rather than about 24 hours.

“If they had engaged in some of that [drug-related] activity on the weekend they weren’t prepared to not drive for two weeks, because life goes on,” Mazzone says.

We have a demographic that’s wanting to engage 24/7. Plus, we know people don’t want to come into a physical office.

However, when drivers were told they needed to wait for much shorter timeframes before drugs they may have taken dropped to levels that were not impairing their driving, the rate of drug-driving also fell.

“It was still illegal to take drugs and we didn’t shy away from that,” Mazzone says, “but when people were armed with that information they said: ‘well, if I knew that, I’d actually wait. If I’d taken drugs on Sunday, I wouldn’t drive [on] Monday’.”

The human touch

The experience demonstrated to Mazzone that to change behaviour you needed a human-centred design approach that truly understood and challenged assumptions about how people would navigate and engage with different tools.

“It also demonstrates that people want to be empowered to solve their own problems,” he says, “so we set about trying to understand how technology and AI could provide a guided pathway and empower people to solve their own disputes.”

The commission hired a market research firm to conduct in-depth interviews with separating couples. Mazzone says it completely changed the approach to design.

“We had originally segmented the market, as a lot of people do, and accountants are very trigger-happy in looking at income bands,” Mazzone says.

“However, income had no bearing on whether people wanted to use the tool or not. It actually had really strong take-up – we had 78 per cent interested in using the tool across all income bands. It came down to two factors: how amicable they were with their partner and how complicated the situation was, or they perceived their situation to be. That set us on a course to [segment] the market to say it had different needs and people’s journey through the system was quite different.”

Mazzone says the AI tool could be adapted to other legal areas if successful.Former partners on speaking terms wanted to sort out arrangements quickly and painlessly. 

“They wanted limited education around that – they just wanted to know what they needed to know,” Mazzone says.

When the project team originally spoke to legal experts, it was told domestic violence and abuse could not be dealt with through online dispute resolution.

“We built exceptions out of this process and started to look at a system and scope it up.” Mazzone says. 

“However, when we spoke to the users we received some interesting feedback, particularly from people who had a history of domestic violence in their situation.

“What they said was, ‘we want to use a system like this, but we want it to be one-sided because we want to be empowered. When we know more, we’re empowered and we’re stronger’.”

Others stuck in more complicated situations saw the tool as a potential way to engage their former partner.

Teaching AI new tricks

The tool is not intended to replace existing services but it could alleviate pressure on an overstretched legal system. However, it’s not the first time an online resolution tool has been built. The Netherlands tried and failed.

“We’ve learned from their failings and, in particular, it seemed like there was a one-size-fits-all approach,” Mazzone says. 

“Our market research came back really strongly that the one-size-fits-all approach wouldn’t work. They also charged people upfront to use the system. We envisage a system that is free up until the point that you get a product, then we might look at how we charge.”

The system will be able to produce agreements, consent orders and parenting plans. “I guess the real point of difference for us is the artificial intelligence,” Mazzone notes. 

The AI built into the system takes the information entered by users about their dispute and compares it against Australian family law cases and consent orders to generate a suggested division of assets for the parties. This provides a starting point for a resolution that is fair and referenced against outcomes accepted by the courts.

“We’re starting to populate more and more data, and the more we can train the model, the more accurate it will be,” Mazzone says. 

“The idea is to get that accuracy up to a level that we’re satisfied with before we go out to a public trial and start to let the public use it.”

For me it was very much about ‘does this role offer something that is maybe a little bit different?’.

He is confident the system can be rolled out efficiently and at low cost because, being bespoke, it does not have expensive licences. While the trend is to steer away from custom-built systems, the approach has allowed the commission to create an agile product. User testing has driven further refinement of the AI, including building in an element to keep conflicting parties behaving amicably online.

“Disputes resolve far easier when parties are amicable and play fair,” Mazzone says. 

“We’ve built in some natural language and sentiment analysis, so that if you start to use aggressive language through the tool, it will come up and say, ‘hey, that’s not fair, this process works better if everyone stays calm’. It actually won’t let them proceed.”

An individual trying to claim an exorbitant level of assets won’t be able to move forward, either. Using AI, the system can identify if either person proposes to change the division of assets to claim a percentage outside what is deemed an acceptable range.

Meeting new expectations

Some people probably won’t enjoy a computer scolding them, but Mazzone says the reaction has been positive. 

“They liked it because they could see the other party couldn’t take it to a level that would start to get the negotiation off the rails.”

He believes take-up of the system will be high. 

“We’re lucky in Australia that we don’t face the barriers that maybe some other cultures do – people are tech-savvy, people willingly give technology a go particularly if they think it makes their life easier or it’s cheaper,” he says.

“We have a demographic that’s wanting to engage 24/7. Plus, we know people don’t want to come into a physical office.

“People want to deal with it online. They want to do it when they finish work, after they put the kids to bed. It’s really around this expectation from the community that government services and services more generally will be available 24/7.”

Read next: Using robots to detect forest fires

Like what you're reading? Enter your email to receive the INTHEBLACK e-newsletter.
October 2021
October 2021

Read the October 2021 issue of INTHEBLACK magazine.

Each month we select the must-reads from the current issue of INTHEBLACK. Read more now.