Sam Berner on the ethics of machine translation

The rise of technology is always a hot topic at translation conferences, and the AUSIT National Conference 2018 in Adelaide was no exception – Sam Berner, a legal translator based in Queensland, gave a thought-provoking talk entitled: “Ethical Questions for the Age of Intelligent Machines”.

Most such discussions revolve around the question of whether MT will ever replace human translators – which Ms Berner boldly asserted is fundamentally the wrong question, her premise being that we do not, and cannot, know what will happen and how technology will develop in the future. While I don’t personally fully agree with this statement, Ms Berner instead posed a set of questions which are worth thinking about:

  • Is MT being funded? Why, and by whom?
  • Is it being used?
  • Is it a useful and convenient tool?
  • Is it affecting our clients?
  • Is it being hyped?
  • Who is doing the tech research?

While she did not claim to have answers to all of these questions, she did point out a few salient points – funding is being put into machine translation research, including by the very universities and education institutes that teach languages and translation – why? Has nobody considered the conflict of interests? In fact, this is against the backdrop of a widespread defunding of language courses at universities and other education institutions.

While translators are a diverse population of intercultural mediators, representing an ethnically and gender-diverse demographic, who are in the business of solving problems of human relations, Ms Berner pointed out that MT research and development is almost exclusively the domain of young white males with no actual knowledge of languages and cultures – which is alarming, to say the least.

Whether or not machine translation is being hyped is for all of us to judge for ourselves. Nevertheless, it is certainly being used, and its use is becoming more widespread.  And it is in the nature of the industry that freelance translators, as most of us are, have little ability to assert ourselves against the corporate forces that are driving the development of AI. In this context Sam Berner made another bold statement – that, while poor quality is not acceptable to us as translators, it is generally acceptable to the rest of the population. Clearly this is an overstatement, for if it were true, none of us would still be doing what we do. Yet there is an element of truth to it, and it does affect our ability to win over clients who are seduced by the faster and cheaper option – meaning that the onus really is on us to communicate to clients and to the world in general what it is we do, and why it matters.

Ms Berner also undertook a brief review of the existing literature, and pointed out a gaping hole – while much has been written about translation and the ethics of translation, and much has been written about machine translation and NLP (natural language processing), published literature on the ethics of NLP and machine translation is virtually non-existent.

Placing the question of ethics and AI in a broader context, Ms Berner referred to the much-covered case of software developer Ibrahim Diallo, who had the misfortune to be fired by the very system he had helped to develop. In short, the system erroneously identified him as no longer having a current contract, and locked him out. Staggeringly, while his superiors acknowledged that something was wrong with this situation, not one of them was prepared to step in and reverse the decision made by the software. Rather than a story of machines becoming intelligent, this seems to me to be a story of humans abdicating responsibility – which certainly does raise some ethical questions: are machines truly intelligent and capable of making ethical decisions? Clearly the answer is no. To bring this back to the context of translation, there are some ethical questions relating to postediting which are perhaps not adequately addressed in our current Code of Ethics. Who, after all, is ultimately responsible for a text produced by a machine? Especially considering the all-too-common scenario in which a translator is being paid minimum rates to merely “proofread” the document.

This was certainly a thought-provoking talk. And given the absence of the translator’s voice from the forces that are driving this technology and publishing the research on it, I for one hope that some of us will take up the challenge of probing further into the ethics and usability of machine translation, and adding our perspective to the tech discourse.

By Guthrun Love

Photo by Franck V via

Comments are closed.

Proudly powered by WordPress | Theme: Baskerville 2 by Anders Noren.

Up ↑