Rule of Robots vs Rule of Law
03 January 2019
Artificial intelligence (AI). Technology. Two (or three, if we are being picky) words which seem to appear on my news feed every morning, as I am sure they have been for many of us out there. It feels like not a day goes past without another article being published about the ever-increasing spread of AI and technology – from transforming clinical trial recruitment1 to concocting new perfume combinations2. We are now even hearing about how researchers have taught an AI program to identify galaxies in deep space from the facial recognition technology used on Facebook3. It really is incredible to see the astonishing pace at which technology is advancing and how we are seemingly coming to rely on it in nearly every facet of our lives. So, what does it all mean for us in the legal sphere?
Many of those working in private practice, as well as those in government and in-house counsel, are already aware of how AI technology can have a beneficial effect in the work place as an aid to their daily administrative responsibilities. In fact, survey results from the LexisNexis 2018 Roadshow Report showed that 44 percent of legal practitioners have seen technology remove the grunt work from legal practice and 42 percent acknowledge that the nature of legal work has been changed4. Through technology, data-intensive tasks such as practice management (i.e. timekeeping, billing and file management), document review, legal research, document and legislation analysis and comparison can now be completed in a much quicker time and potentially with greater accuracy.
What we are seeing more of now, however, is the use of artificial intelligence in the form of algorithm-based technology outside of the office space and increasingly in the public arena of justice (i.e the courthouse). In some American criminal courts5, judges determining bail have experimented with using algorithm-based ‘scores’ which rate a defendant on several factors, such as the likelihood of skipping trial and reoffending if released on bail as an influence in their final ruling. In Virginia6, almost twice as many defendants as usual were released on bail with zero increase in pre-trial crime. Similarly, New Jersey7 saw an almost 16 percent drop in its pre-trial jail population with no increase in crime. This example serves as resounding evidence for AI’s ability to help protect the rights of defendants and ensure greater judicial equality. AI may also serve to decrease time to trial resulting in shorter remand times for those who don’t get bail. More recently, a software app developed in Argentina was used to generate draft rulings in place of lawyers and paralegals. Those rulings so far have had a 100 percent approval rate from judges and lawyers and are clearing through six months’ worth of cases in six weeks8.
While these results are all promising, there is still the issue of whether we can truly trust artificial intelligence to decide the outcome of a court case instead of a human judge or give us legal advice instead of a lawyer. There is less of an issue in terms of using technology for administrative purposes like e-discovery and document review, but more is seemingly at stake when it comes to artificial intelligence taking on the role of an advisor or implementer of justice. We know that algorithms are not perfect, and that the data fed into an algorithm can be flawed and potentially result in biased outcomes. From a wider perspective, there are also issues of transparency in decision making and proper regulation of technology. If we are not provided with an explanation as to how an algorithm came to a decision, it will be near impossible for unsatisfied parties to examine and challenge the decision. Algorithm-based decisions could, therefore, effectively render the long-standing grounds of judicial review, such as procedural fairness and error of law, inoperable. Similarly, the current absence of proper and independent regulation of the use of algorithms leads to a complex discussion of the ethical and potentially fiduciary responsibilities of those using the algorithms – particularly for those who may not completely understand the technology that they use -- as well as the companies who design them. All of these are valid concerns and should be examined carefully. In fact, it is likely that the rapid development in legal technology will be raising more questions at a quicker pace than the legal industry, and quite possibly the world at large, can answer.
We should, however, not refrain from using something which has been proven to be beneficial simply because we do not fully understand it. There was a time when communicating with clients via email was relatively unreliable and seen as ground-breaking feat - yet we now cannot fathom business without it. Obviously, we must remain vigilant about how technology could undermine our access to legal justice and actively seek to address the current ethical debates. In paraphrasing a lawyer who lost to AI in a recent contract review challenge, however, the future of the rule of law will no longer be human versus computer. Instead, it will be human and computer versus another human and computer. Either one of them working alone will be inferior to the combination of both.9