United States immigration authorities use artificial intelligence-powered translation applications without understanding the limitations of these automated systems, leaving people seeking asylum in their country without a real means of communication and even jeopardizing their ability to enter a foreign country The guard.
This was the case for Carlos, an Afro-Indigenous originally from Brazil who decided to flee to the United States along with his sister and two of his nephews after a local gang murdered his son in front of him.
Upon arrival, Carlos was separated from his family and held in an Immigration and Customs Enforcement detention center, where he spent six months without knowing the whereabouts of his loved ones and without being able to communicate with anyone because he spoke Portuguese using an AI-based language translation tool to interpret what Carlos was saying, but the system didn’t recognize his accent and didn’t understand him.
While he was at the immigration center, he didn’t really know why he was being held, and when he sought medical attention for high blood pressure and COVID issues, the nurses had trouble understanding and caring for him. The same thing happened when filling out his asylum application: the language barrier was one of the biggest challenges in his search for refuge in the United States. “I never thought it would be the worst,” Carlos said through a translator.
This is not an isolated incident, as the US immigration system has stated that it will provide migrants with a human interpreter if they need one. However, the reality is that many do not have access to one. Instead, authorities – and even some aid organizations – are increasingly relying on AI-powered translation tools such as Google Translate and Microsoft Translator.
According to Damian Harris-Hernández, executive director of the Refugee Translation Project, a group that helps refugees with translations, AI-powered translation tools are particularly unreliable in languages that are significantly different from English or are less documented.
Problems with translation tools occur throughout the asylum process, from border posts to detention centers to immigration courts. Some volunteers in the support group say they have experienced cases where asylum applications were rejected because the translation tool misinterpreted some wording
For example, translation issues were particularly concerning for people from Afghanistan. For example, major machine translation systems like Google Translate do not offer translations in Dari, one of the country’s two official languages.
A similar phenomenon is also occurring in universities, where text recognition systems generated by artificial intelligence are biased towards people whose native language is not English.
Without cultural context, machine translation systems will continue to prioritize a Western view of the world, making it almost impossible to correctly interpret the nuances of most non-English languages, which is essential to avoid discrimination and bias against people in disadvantaged situations.