What role, if any, can interpreting studies play in an era in which machines interpret at human — or even super-human — levels of accuracy?
To answer this question, one must first accept a premise that many within the field still resist: machine interpreting will become extremely capable. As a researcher in interpreting studies and a developer of interpreting technologies, I consider this not a speculative fantasy but a realistic, and frankly plausible, scenario. Not in some distant future, but in the near term. I have often suggested 2030 as a rough horizon for parity.
Let’s be clear, by “parity” I do not mean that artificial interpreters will replicate human interpreters in every respect (read here an article about the Turing Test of Speech Translation). The processes will remain fundamentally different. But the observable outcome — the interpreted speech as experienced by users — will increasingly be comparable. This distinction matters. It is also where much of the current debate falters. Many scholars still regard such a scenario as impossible; others concede it may one day occur, but only in a remote future. To reflect seriously on the future role of interpreting studies, however, requires a modest leap of faith: to treat this scenario as not only possible, but likely.
Machines will change the interpreting ecosystem
If that happens, the first consequence will be ubiquity. Interpreting will no longer be an exceptional service, delivered in carefully delimited settings. It will become a feature. You will put on earbuds, glasses, or some yet-to-be-invented device, and you will gain real-time access to speech across languages. Interpreting will be woven into everyday communication, much as spell-checking or navigation software is today.
A second consequence is visibility. Unlike written translation — which thrives on invisibility, to the point that many readers forget they are reading a translated book or website — interpreting cannot disappear. It unfolds in real time. It makes linguistic difference perceptible rather than erasing it. And for a long time, it will remain imperfect, because language in real contexts is irreducibly complex. Users will know they are encountering mediated speech; translation will be present, audible, and consequential.
Third, in sensitive or high-stakes contexts, the choice between human and artificial interpreters will not simply dissolve. There remains a threshold beyond which people prefer a service over a feature. In such domains — legal proceedings, healthcare, diplomacy — the demand will not merely be for performance, but for guarantees. Certification will matter. This is not new. Technical devices are already regulated in this way, as are human interpreters, who must demonstrate eligibility through credentials that vary widely in rigor. The same logic will apply to artificial interpreting systems, particularly in regulated environments.
Machine Interpreting and Interpreting Studies
Against this backdrop, interpreting studies faces a rare opportunity. It can offer insights that are both socially valuable and capable of enhancing the discipline’s, let us be honest, rather modest impact on the interpreting ecosystem. While computer science builds highly capable systems, interpreting studies brings something else: a long tradition of examining interpreting as communication in action, embedded in context, norms, and expectations.
Its most important potential contribution concerns quality evaluation. What makes an interpretation successful? How can success be measured? Here the discipline’s historical record is mixed. Even for human interpreting, quality has resisted precise, operational definition. Yet the ingredients are now in place to revisit the question with renewed urgency. This will not be easy. Moving beyond common-sense judgments and broad assertions toward metrics that are scientifically motivated and empirically traceable is a demanding task. But the stakes are high. Artificial interpreting systems will soon need to be certified, or explicitly excluded, from high-risk uses. If standards remain vague, adoption will default to convenience rather than appropriateness, with serious implications for fairness and safety. Expanding language access is a worthy goal. Doing so at the cost of accountability is not.
A second, more modest but still meaningful contribution lies in comparative analysis. Artificial and human interpreters may converge in performance while diverging radically in how that performance is achieved. Exploring these differences, such as how machines “decide,” process, and produce interpretations, opens new interdisciplinary terrain between interpreting studies and computer science. At the same time, this contrast sharpens our understanding of human interpreting itself, not as a benchmark to be defended, but as a phenomenon newly illuminated by comparison. Other disciplines of the mind have been revitalized in precisely this way; neuroscience’s encounter with computing offers a telling example.
Finally, interpreting studies may serve a broader public function: fostering literacy about interpreting, and especially about artificial interpreting. For millions of users, AI-mediated interpreting will soon be their primary point of contact with multilingual communication. Helping society understand what interpreting is, what it can and cannot do, and what is at stake in its automation is no small task, but it is a necessary one (for an introduction to Machine Interpreting, see my chapter The Routledge Handbook of Interpreting, Technology and AI1).
Abandoning a (human) interpreter-focused tradition
What would it take for the discipline to move in this direction? First and foremost, recognition, still largely absent, that artificial interpreting agents are legitimate objects of inquiry. This requires a shift away from the field’s long-standing fixation on the human interpreter and back toward the interpreting event itself, regardless of who or what performs it. The risk, of course, is that AI will be approached primarily in opposition to humans, a bias that is understandable given the discipline’s history and interests. But approaching it with intellectual openness would be far more rewarding, for researchers, for the field, and for society at large.
Interpreting studies can either observe this transformation from the margins or help shape how it unfolds. This choice is rapidly ceasing to be theoretical. Taking the latter path would also allow the discipline to finally confront an object of inquiry that was already anticipated in 1995, when machine interpreting was placed — well ahead of its time — on the conceptual horizon of interpreting studies itself1.
- Fantinuoli, C., 2025. Machine Interpreting, in: Braun, S., Davitti, E., Korybski, T. (Eds.), The Routledge Handbook of Interpreting, Technology and AI. Routledge. ↩︎
- Salevsky, H., 1993. The Distinctive Nature of Interpreting Studies. Target 5, 149–167. https://doi.org/10.1075/target.5.2.03sal ↩︎