Listen to This Blog
The rising frequency of discussions about AI has led to much unease among interpreting service providers about the impact of automated solutions on language access and on what that means to the future of their profession. In this blog, we explore the concept of meaningful language access, why automated interpreting struggles with that concept, and how it relates to deciding and defining the role of the interpreter.
What Is Meaningful Language Access?
The concept originated in Title VI of the Civil Rights Act of 1964, landmark legislation that outlawed discrimination based on race, color, or national origin in programs and activities that receive federal assistance. Its reach was wide, extending to other forms of discrimination. The US Department of Justice Language Access Plan most recently defined meaningful access as “language assistance that results in accurate, timely, and effective communication at no cost to the individual with LEP needing assistance. Meaningful access denotes access that is not significantly restricted, delayed, or inferior as compared to programs or activities provided to English-proficient individuals.”
The concern with meaningful language access is the extent and manner in which services should be provided to Limited English Proficient (LEP) or Language Other than English (LOE) individuals to ensure equality and non-discrimination.
Note that for people who are Deaf, the equivalent concept is called “communication access” and is mandated in the US by Section 504 of the Americans with Disabilities Act or Rehabilitation Act of 1973.
Best practices for implementation vary and can clash with some technology solutions. For example, telephone and video remote interpreting (OPI and VRI, respectively) are typically deemed adequate solutions, yet our research shows that end-users fully trust remote interpreters 22% less than in-person ones.
Is remote interpreting a breach of the Department of Justice requirement? Most organizations would claim “no” because they typically consider remote interpreting modalities when they can’t secure an in-person interpreter within a desired timeframe – even if their real reason is that telephone or video interpreting costs a lot less.
Why Would Automated Interpreting Struggle with Meaningful Language Access?
So, let’s add the concept of automated interpreting to the list of alternatives for meaningful access. For spoken languages, it means using artificial intelligence to render a translation in spoken form. For signed languages, it’s either text-to-sign technology using avatars or sign-to-text technology using complex sign decoding systems.
Today, interpreting automation can do a lot, but there is a lot that it cannot capture. For example, its failure to incorporate elements tied to culture, emotional context, and similar elements reduces the quality and impact of the output. In short, absent human sensory organs, machines just receive a fraction of input to process compared to what human interpreters can see, hear, feel, and smell. It’s no surprise that human professionals naturally outperform AI solutions.
This human vs. machine dynamic poses a big question – does automated interpreting for spoken and signed languages qualify as meaningful language or communication access? Two factors seem to point to “not yet”:
- The law. Interpreting services should not be “inferior as compared to programs or activities provided to English-proficient individuals,” yet they likely will be less effective, due to mistakes as AI has not yet reached human parity.
- Trust. The absence of human parity in interpreting takes us again to the trust question we saw earlier. Why is it that remote solutions – which end-users don’t fully trust – are deemed adequate but automated interpreting is not? There is, of course, a difference in quality between a remote interpreter and a machine. But what is not clear is the threshold at which a less optimal option becomes acceptable.
Legal issues or human trust notwithstanding, the fact remains that most organizations feel that remote interpreting and interpreting by non-professionals are permissible, in some situations at least. That reality opens the door to discussing why AI couldn’t be an equally valid option.
What Is the Role of the Interpreter?
The traditional view of interpreting revolved around the concept of a linguistic conduit, meaning interpreters acted as neutral parties who simply conveyed messages from one language to another without adding, omitting, or altering the content. This is what automated interpreting more or less accomplishes today – or strives to.
However, the interpreters’ role has evolved over time to be more than translators of spoken or signed words from one language to another. In this newer “active” model, interpreters play a more involved role in the communication process. They take liberties to ensure that the message is understood correctly by both parties. This can include clarifying meanings, ensuring cultural appropriateness, and sometimes even mediating the conversation. As active participants, their presence and decisions can influence the outcome of the interaction. AI is nowhere close to capable of reaching this level of proficiency.
So, what happens if the less contextually informed, less participating remote interpreter happens to be an AI-bot acting as a simple conduit translating just words but not conveying any of the other signals? Will interpreting still be meaningful if organizations revert to a conduit model due to the adoption of AI interpretation? It will depend on the definition of more than just accuracy and should account for the other responsibilities of interpreters. It may be less of an issue when interpreting for a business meeting or a conference, but it is of vital importance in contexts such as health care, legal, and social services.
Don’t Miss the Report on Perceptions on Automated Interpreting
To learn more about this topic, read the 350-page report available for free on our research platform. The study was conducted for the Interpreting SAFE-AI Task Force as a stepping-stone toward establishing guidelines on safe use of AI in interpreting. This research was not about proving either the AI enthusiasts or AI detractors right. Rather, it was meant to capture a broad, multi-constituent understanding of the current situation and reasoning people have around it, so that sound decisions can be made to leverage benefits but avoid drawbacks.
A summary of the report is also available as well as a replay of a webinar where we presented results of this large-scale study.