-Madina Bahramova.

Digital Judges: A Virtual Reality?  

The question that has recently grabbed the public’s attention is that of digital judges. Does the advance in machine learning and AI mean we will soon be replacing our judges with a little metal box? And would it be better?

There are several different levels. Firstly, the little metal box might act simply as an advisor or facilitator. Smartsettle ONE, an AI tool, replaces the human mediator and uses an intelligent blind bid process and ‘nudge technology’, using an open offer and a secret, bottom-line one. The algorithm ‘learns’ each time about the parties’ priorities and tactics and this ‘nudges’ them towards an agreement and rewards the party who makes the greater effort to settle. This was recently used in the UK, and with impressive results: a three-month dispute over a £2,000 unpaid bill was settled in less than an hour.

Secondly, intelligent technologoy could be used to actually hand down judgments. In China, the first smart internet court was set up in 2017 in Hangzhou, followed shortly after by Beijing and Guangzhou. Reportedly, during six months in 2019, 3.14 million litigation processes were recorded on their smart court applications, with no judges or lawyers involved.

But is this to be welcomed? Do we really want to confer judicial power on an algorithm?

Digitalised dispute resolution would be quicker and cheaper. In the UK, civil litigation costs are often “disproportionate and impede access to justice”, thereby undermining the rule of law and the right to fair trial.

In addition, a computer is (in theory) neutral. It does not suffer from Monday morning/Friday afternoon syndrome. Research at Princeton showed that judges are more lenient when they arrive and immediately after lunch or a break, male defendants are often sentenced more harshly than women for the same crimes, and black and Asian defendants can be 20% more likely to be imprisoned than white offenders. The possibility of judicial corruption would also be avoided, making it fairer and more consistent, thus inspiring public trust.

However, with big data comes great power. Firstly, the metal box needs a creator, leading to possible conflicts of interest. A report shows that 16% of the equity of NAM, an early digital ADR mediator company, was owned by a large insurance company. The potential problems are clear.

The use of AI also raises the question of inherent bias. Machine learning is based on data sets. If those data sets are themselves biased then that bias will be replicated and become increasingly entrenched as the system uses and re-uses the biased data as its training set.

Bias also can be ‘taught’ to the system, directly or indirectly. When Amazon’s recruitment programme tried to find good hires, it tried to replicate successful previous hires: because most had been men, the algorithm ‘learnt’ to favour men. Similarly for indirect bias, Council of Europe literature highlights an employer who showed a marked preference for elite schools, thus training the programme (unwittingly) to show racial discrimination as the elite schools in question were practically all-white.

These flaws are not only illegal and unethical but also very difficult to detect because the bias is not in the initial source code; it is hidden in the data training sets and in the results of the machine learning process. Even the original coders will have no idea how the programme is now operating as it changes with each iteration, making it impossible for a watchdog body to oversee.

Like King Canute, the legal system cannot stand in the way of “the power of the technological tide”. And neither should it. Machine learning, big data and AI techniques offer a chance to revolutionise the system. Despite certain reservations, we have to remember that no system is ever completely consistent or efficient. One opinion poll found that 75% of Britons believe that the UK’s traditional legal system is neither fair nor transparent. The numbers drop even further for ethnic minorities. So, if our new digital judge can do better than that, it will bring value, and should be welcomed.