I, Your Honour: what if robots dispensed judgement in small claims tribunals?

The Baltic state of Estonia has become one of the first countries to start the development of an artificial intelligence-based system that can decide small claims disputes.

Small claims are monetary disputes having a value of less than €5,000 in Malta which are decided by lawyers acting as adjudicators at the Small Claims Tribunal.

While Malta is some years behind Estonia – the country is already two decades into a project which has produced ‘the most advanced digital society in the world’ – it appears to be moving in the same direction. The island is embracing new and innovative technologies and citizens are able to access a range of services online, but how far is Malta from employing Artificial Intelligence in court?

“Such a system would be possible, provided that it is given the proper legal basis, is adequately tested before it is introduced and is properly supervised and still allows each case to be considered on its own merits,” Justice Minister Owen Bonnici told MaltaToday, who did not rule out the use of robot judges when asked about the feasibility of such a system.

He said that given AI’s potential advantages, government was “following developments”, but stressed that one could not rush since it was a “delicate subject affecting the determination of people’s rights”.

Chamber of Advocates president Louis Degabriele, asked whether a machine with the ability to learn could hypothetically oversee Malta’s justice system, emphasised the need for quality.  “It’s a matter of confidence really, of giving the public confidence that the system can work more efficiently and give the same quality of outcome as today’s system.”

He agreed that small claims cases would be an ideal testing ground for such a system.

“Unlike more complex cases, which he said could all be considered unique, debt claims are more straightforward and amenable to being decided by an automated system.

“Quite frankly [in such cases] I either owe you money or I don’t. There is a limited number of defences that can be raised and I’m quite sure that this can be done through artificial intelligence,” Degabriele said, adding that there would need to be assurances and system audits to offer those using the system, peace of mind.

Degabriele also said that implementing an AI adjudication system would require an overhaul of the present system to ensure that it is optimised for this purpose, since an AI can only be as good as the fed data. “The emphasis needs to be on quality over quantity. If most decisions at first instance end up being appealed, we’d actually be doubling the courts’ work.”

Philip Manduca, a lawyer and small claims adjudicator, agreed that AI could be useful in small claims cases, given its straightforward nature, adding that it could significantly reduce the time within which disputes are resolved. “While having the system decide might require an additional set of guarantees, systems that aided members of the judiciary arrive at their decisions already existed… [but] in order to offer peace of mind appeals would need to be heard by a human.”

Creating a system that can explain itself

Broadly speaking, AI involves a machine that can interpret, learn from, and use large amounts of data to achieve specific goals. But while one might reasonably assume that achieving that goal is a priority for developers, gaining the public’s trust will also require ‘explainability’, especially in more complex cases.

“The Al will definitely need to come up with a detailed explanation, and proper reference to all laws that it has used to come up with an interpretation,” said Angelo Dalli, an AI expert and entrepreneur who is already investing in the technology.

He said there was definitely room for the country to start off with AI-powered judicial assistance, but one needed to keep in mind that laws often needed to be interpreted within a particular context. “It’s not as simple as applying simple logic to get the right answer… This would be the ultimate impartial legal system, but unfortunately such a system does not exist anywhere, and possibly with good reason.”The ambiguous nature of some laws, he added, allowed human judges greater flexibility when required.  

Another issue Dalli pointed out was that “today’s laws often end up regulating future scenarios that might never have been envisaged by the authors”, adding that it was almost impossible for laws to keep up and create a precise model. “AI systems are not that good when it comes to dealing with ambiguous situations.”

On the ‘political’ level, he might expect significant resistance at the start, but said this would be mitigated by “always having a human in the loop” and ensuring a long enough period of testing before implementation.

Biased data, biased judgement

Asked whether he believed we might ever see criminal cases being decided by AI, Dalli pointed to problems encountered with some systems used to predict whether individuals should be granted parole in the US.

In some instances, it was discovered that the system, feeding off racial bias in past judgments, was, with all else being equal, less likely to recommend parole for black people than it was for white Americans. “The technology needs to be developed to offer a complete explanation before it can be relied upon more,” Dalli said.

Another challenge with more complex cases, Dalli explained, was the fact that lawyers tended to employ “strategies that are not truly logical in court”.

“The current legal system often relies on establishing doubt on a particular logical chain of facts, which is something AI systems can’t really do too well in their current form,” he said.

However, having said that, Dalli said AI could be easily employed to “sort through massive amounts of documentation quickly, summarise relevant facts and highlight the most pertinent information about a case. I can envisage AI being used to construct logical models of arguments and facts to find gaps in a particular argument, allowing different parties to ask more pertinent questions.”

Dalli also highlighted the fact that AI could present evidence in an impartial manner without the bias that lawyers introduce quite often intentionally. “In a perfect world such strategies wouldn’t exist, and everything would be factual, as exemplified by the traditional view of the law being blind and impartial. The increased introduction of AI in the legal process will accelerate that vision becoming a reality.”