May 2, 2023

AI Translators Are Putting Afghan Asylum Claims At Risk

The errors and cultural blind spots introduced by machine translation can lead to confusion and rejected asylum claims, Rest of the World has found.

Rise in demand for translators

In 2021, the Taliban returned to power in Afghanistan, displacing million of Afghans. As Afghan refugees seek asylum around the world, there is a growing demand for translators.

Some translation providers have turned to machine learning models to cut costs and save time, but refugee advocates say this risks introducing error into asylum claims.

AI translation models are more accurate for ‘high-resource’ languages like English and Mandarin, than Afghan languages like Pashto and Dari—languages native to tens of millions of Afghans around the world.

Machine translations of Dari and Pashto have led to confusion and the rejection of at least one asylum claim, Rest of the World has found.

Errors and cultural blind spots

Tarjimly,  a nonprofit startup that connects refugees and asylum seekers with human volunteer translators and interpreters, reported a fourfold increase in requests for Afghan languages in 2022.

“Machine-learning translations are not yet in a place to be trusted completely without human review,” Tarjimly’s COO Sara Haj-Hassan, told Rest of World.

“Doing so would be irresponsible and would lead to inequitable opportunities for populations receiving AI translations, since mistranslations could lead to the rejection of cases or other severe consequences.”

“I really feel that if we rely on machines, I [am] expecting at least 40% of our decision-making on the asylum applications for refugees would be incorrect.”

Muhammed Yaseen, a member of the Afghan team at Respond Crisis Translation, added that machine tools have struggled to translate terms for some relatives in Dari dialects, and specialized words like military ranks that can be vital to the asylum applications of former US-allied soldiers.

Moreover, cultural blind spots and failures to understand regional colloquialisms can introduce inaccuracies when machine translation is used to draft documents.

“If we use machines for Afghans, I think we would be unfair to them,” Yaseen told Rest of the World. “I really feel that if we rely on machines, I [am] expecting at least 40% of our decision-making on the asylum applications for refugees would be incorrect.”

Rise in AI translators

According to ProPublica, in 2019, US Citizenship and Immigration Services officers were instructed to use Google Translate to vet asylum seekers’ social media accounts.

Moreover, federal immigration agencies have millions of dollars worth of contracts with several major translation companies that often advertise machine translation in their services.

Several AI translation providers are also actively pitching refugee organizations to integrate their services into their work, Rest of World reports.

The stakes are high

OpenAI recently updated its user policies, prohibiting the use of the AI chatbot in “high-risk government decision-making,” including work related to migration and asylum.

Similarly, advocates for refugees caution against relying solely on machine translations in high-stakes scenarios like asylum claims.

Read: GPT Detectors Biased Against Non-Native English Writers, Study Finds

“One of the things that we see frequently is pointing to small technicalities on asylum applications,” Ariel Koren, the founder of Respond Crisis Translation, told Rest of World.

“That’s why you need human attentiveness. The machine, it can be your friend that you use as a helper, but if you’re using that as the ultimate [solution], if that’s where it starts and ends, you’re going to fail this person.”

Samara Linton

Community Manager at POCIT | Co-editor of The Colour of Madness: Mental Health and Race in Technicolour (2022), and co-author of Diane Abbott: The Authorised Biography (2020)