IN THE LONG RUN TRIAL JUDGES MUST BE REPLACED BY EXPLAINABLE ARTIFICIAL INTELLIGENCE

1. Introduction

Currently artificial intelligence (AI) is a ‘black box’, which makes decisions by algorithms. AI is already extensively used in ‘machine learning environments’ where dots are connected from large data sets to make a decision. The application of AI is already extensive in the fields of marketing to medicine and all the way to self-driving cars.

There is also an emerging field of ‘legal tech’ beyond just electronic filings of cases and legal data bases. For example, there are predictive models, which predict the likelihood of success of a litigation or the likelihood of somebody re offending. However, legal tech applications are not without their critiques. Some researchers are questioning whether AI based decisions are leading to the wrong decisions (Selbst, A.D., D. Boyd, A.  Friedler, S. Venkatasubramanian and J.Vertesi, January,2019 ACMFAT Conference Atalanta; Hao K., 21/12019, MIT Technological Review).

One of the fundamental issues with the current AI applications that there is no explanation for a given decision. However, researchers are developing AI systems which can also explain their decisions. The consulting firm Accenture estimates that within three years AI systems will be able to explain their decisions leading to explainable artificial intelligence (EAI) systems (www.accenture.com/au-en/insights/technology/expainable-ai-human-machine) .

My contention is that once EAI technology is well established, trial judges must be replaced by EAI. Furthermore, only appeal judges should remain human judges as a control for EAI errors, if any. I give my reasons for my contention below.

2. Even trial judges are humans

Trial judges are no different to any other human beings. They have good days, they have bad days, they have emotions, they have a home life, they may even have naughty young children and they certainly have political views and biases. No matter how hard they may try, the human factors must influence their judgement, which in turn can have a profound impact on another human’s life.

EAIs do not have such biases. Even if one would run the argument that all cases are unique, hence cases are not comparable, the ‘humanity’ of the judges must influence their decisions. Furthermore, judgements should only be based on facts as presented in the courts and how these facts relate to a specific legislation. Hence, this is an ideal setting for EAI based decisions. Finally, the appeal system with human judges could overcome the ‘machine errors.

3. Judges are badly trained in commerce and psychology

I have reviewed the curriculum of three leading Law Schools; Harvard, Oxford and Melbourne University.

The most surprising element of these law programs are how little Accounting, Economics and Finance options, if any, are available to students during their legal studies. For example, Harvard Law School which seems to me to have  the most flexible program, only offers some of these subjects as a 3-day Section or over a three-Week Section with a few hours per week. May be a did not looked hard enough but, I did not find any offerings in psychology. 

My own experience as an independent expert witness in commercial litigation underpins the proposition that judges are badly trained. It is not unusual for the introductory part of an expert report to include some very basic information on accounting or financial economics issues.

EAIs would not have a problem with understanding the mathematical formulation of even the most complicated accounting or financial economics issues. The role of AI clearly recognised by Oxford as, currently there is a specially funded research project being undertaken jointly by the Faculties of Law, Economics, Computer Science, Education and the Said Business School, which is looking at the role of AI for English law (‘Unlocking the Potential of Artificial Intelligence for English Law’). I wonder if the researchers will look at the role of trial judges versus EAI.

4.The trial process is prohibitively expensive and can take a very long time to conclude

Even before a trial begins, the pre-trial discovery process can be financially exhaustive. Especially, if one of the parties adopt the ‘deep pocket’ strategy to force the other party not to continue with the litigation (A. Cannon, 2011, 37 (1), Monash Law review).

During the trial process, a legal team of two to four legal professionals and a number of expert witnesses are not uncommon. Whilst, I could not get any statistics on the cost of say of an ‘average’ commercial litigation, leading barristers in Australia charge AU$ 20,000.00 per day (B. Walker ,7/6/2019, The Australian). Even the cheapest legal team of one barrister and one instructing solicitor can lead to hundreds of thousands of dollars of litigation costs in Australia. In the US,  it is common to have a contingent based fee structure, where the law firm’s fee is approximately one third of the amount recovered.

The litigation process can take a very long time. The trial date may be delayed by months if not years, the trial itself can be lengthy and the judge may not deliver his/her judgement for months. Palumbo, Giupponi, Nunziata, Sanguinetti provide insightful evidence on the length of civil trials in the Europe (2013, OECD, Economic Policy Papers No5). The average length of trials is 283 days, although in Italy it can be as long as almost 8 years.

I contend that trials by EAIs would be less costly and significantly faster. There would be no need for extended arguments in courts, cross examinations and expensive legal teams, especially costly barristers. EAIs could provide a judgement within minutes if not seconds.

5. Trial judges make mistakes

I could not find in any jurisdiction any statistics on how many judgements are overturned on appeal due to error by the trial judge.

The only information I could get is that the US Supreme Court overturns approximately two thirds of the original judgements as reported by Woodward and M. Sherman (https://www.apnews.com/5407eb459668436abb13be9922d77de2 ). The report does not provide a breakdown for the different reasons why an appeal has been successful. Nevertheless, it is reasonable to assume that at least a number of appeals have been successful due to errors by the trial judge. Furthermore, N. Berkovic reports (The Australian, 19/6/2019) that seventy judgments by an Australian Circuit Court Judge have been overturned by the Appeal Courts over the last five years.

6. Key takeaways

Judgments are biased as they are made by humans.

Judgements are made by judges who are only trained in law, not in commerce or humanity.

Currently the legal system is costly.

Currently the legal process is lengthy.

Explainable Artificial Intelligence could bring less costly, less biased and faster judgments.

The appeal system run by human judges could correct the mistakes, if any, of EAIs.

Leave a Reply

Your email address will not be published. Required fields are marked *