Robots in the Court?

It is no secret that certain technology companies have set their sights on revolutionising entire industries. From driverless cars to automatic medical diagnosis, some experts in Artificial Intelligence have even gone as far as to predict that within a few decades, no humans will be needed in the workplace. The precise effects of technological change on the justice system remained relatively unexplored until the current Lord Chief Justice, Lord Burnett of Malden, delivered the Sir Henry Brooke lecture this year. In his lecture, he predicted that computers would put court interpreters out of a job ‘within a few years.’ In this article, I will take a look at the case for and against his assertion.

There are two main arguments used to support the idea that human court interpreters will soon go the way of duelling. In fact, they are basically the same arguments used to predict or justify replacing any professionals with technology. The first of these is simply that professionals are expensive and so some way must be found to reduce the bill for their services.

Everyone involved in the legal profession will be familiar with that particular argument. It is the argument used whenever government cuts are made. It is the argument behind every drive for efficiency and, ironically, it is the argument used to justify spending millions on new systems, which, we are told, will lead to incredible leaps forward in simplicity and reduce waste.

It is indeed true that court interpreters do cost a lot of taxpayer money. The total value of the contract awarded by the Ministry of Justice in 2016 to cover interpreting and translation was £232.4 million, with around 90% of that expected to be spent on interpreting ( “And So It Begins: thebigword Takes Over From Capita TI”, 30th November, 2016). It makes sense to look for efficiency savings with a budget that large, especially since the languages covered range from Acholi to Zulu.

The case for looking for financial savings is clear. What is not clear is how to save money today without having to spend more money later. Faced with the choice of two interpreters, with the second charging £10 per hour more than the first, many buyers might decide to pick the cheapest. Yet the cost and wasted time of finding that the cheaper interpreter was simply not able to handle the work will easily outweigh any upfront savings. Which barrister would want to find out that they lost a case due to trying to save a few tenners by booking a cheaper interpreter for initial consultations?

If barristers are to be able to do their work and if courts are to run well, quality of interpreting should always be as important as price, if not more so. Talk of cost savings and high fulfilment rates mean absolutely nothing if the interpreters who do turn up are not properly trained and experienced or even if, for some unfathomable bureaucratic reason, they do not have the right languages for the case at hand.

In addition to the cost argument, many promoters of machine interpreting will also make the argument that machines will soon manage to match or even outdo the performance of professional interpreters. Even if performance might not be perfect at the moment, they argue that machine interpreting is a problem on the way to being cracked. If the explosion in the number of companies now offering machine interpreting solutions is anything to go by, it will soon be as easy to get flawless interpreting on a smartphone as it is to order an Uber or find a nearby restaurant.

It is true that machine interpreting has suddenly become a well-funded and popular technological venture. From tech giants such as Microsoft and Google to new players such as Waverly Labs and ili, it is hard to miss their large-scale marketing and PR efforts. It is even harder not to get swept up in the hype and to start predicting, like the Lord Chief Justice, that the combined technical expertise set on the problem will obviously lead to a solution within a short space of time.

The case for the replacement of human court interpreters with machines is definitely strong. It would be completely naïve to pretend that courts will not be affected by technology and that interpreting in court is not ripe for change. Yet even with that in mind, there is a case for seeing human court interpreters as having a bright future ahead of them.

The clearest answer to those who predict the demise of human court interpreters is the one given by the Institute of Translation and Interpreting in response to the Lord Chief Justice. Despite the best efforts of the engineers and marketers in technology companies great and small, it remains true that there is no machine interpreting that comes close to even basic human interpreting competence.

A few recent examples will suffice. Proud of their own efforts to produce flawless machine interpreting and awed by the high accuracy levels reported by their engineers, Chinese technology company Tencent recently decided to showcase their own machine interpreting solution. They allowed the machines to take the place of humans at the 2018 Boao Forum, their corporate forum for leaders in academia, business and government. The results were less than compelling.

For example, when one speaker spoke about China’s “belt and road initiative” to build infrastructure abroad, the machine called it “a road and a waistband” (“AI-powered translation still needs work after errors mar debut at Boao Forum”, South China Morning Post, 16th April, 2018). At one point, the same technology simply repeated the word “for” over and over again (“AI Interpreter Fail at China Summit Sparks Debate about Future of Profession”,, 17th April 2018).

Despite encouraging laboratory results, real-world trials of machine interpreting have all followed a much similar trend of high expectations followed by poor results. Users who have taken Google’s Pixel Buds for a trial in cafés or meetings have posted reviews complaining about the time taken to get any interpreting at all, an inability to deal with poor internet strength and the raw nature of the output. When one Russian expert tried out Microsoft’s Skype Translator with a call to her husband, she found that it couldn’t tell the difference between “cattle” and a “kettle” (“How I tested Skype Translator in Russian”,, 8th November, 2016).

All this demonstrates the difference between laboratory work with controlled data, ideal conditions and repeated tests and real-life, with its inherent unpredictability. This is before anyone even considers the fact that court interpreters must quickly change between styles of interpreting as the requirements of consecutive interpreting used in cross-examinations are entirely different to those necessary when interpreting simultaneously for participants who do not speak English. Indeed, interpreters who do not adjust to these different requirements have been found to impede understanding and hence justice (Speak English or What?, Philipp Angermeyer).

Perhaps the main reason why there is such a strong belief that human interpreters can be replaced is that there is so little understanding of the skills involved. Just as reading a few law books and being able to speak in public is not sufficient to make a barrister, knowing two languages is not nearly enough to become an interpreter. There are almost always several possible ways to interpret what has been said, without losing any accuracy. The skill of interpreting then is not simply knowing the French for “manslaughter” or the Hindi for “murder” but in understanding how to communicate what was said as fully and exactly as possible while taking account of cultural, grammatical, and linguistic difference. Far from being robots who are entirely interchangeable, court interpreters must be highly culturally and legally aware as their decisions can shape the outcome of cases (see for example The Bilingual Courtroom, by Susan Berk-Seligson).

For the kind of interpreting that saves money in the long-term by providing a service that allows barristers and judges to do their job effectively, there simply is no replacement for well-trained, adequately paid, monitored human interpreters. Anything else is simply a recipe for frustration, wasted time and ultimately, miscarriages of justice.

Dr Jonathan Downie

Dr Jonathan Downie is a consultant conference and business interpreter and interpreting researcher, sitting on the Board of the Institute of Translation and Interpreting.




Share this post