AI translations of British Sign Language ‘dangers new obstacles’ for deaf individuals

Charities and MPs raised concerns about the use of AI to translate BSL, with Labour MP Jen Craft explaining deaf people were being brought in to road test the AI too late

View 2 Images

British Sign Language is a first language for many (Image: Getty Images/iStockphoto)

AI translations of British sign language risk spreading misinformation and making disabled access worse, ministers have been warned.

Charities and MPs raised concerns about the use of AI to translate BSL, which for many is a first language. Labour MP and chair of the APPG for BSL Jen Craft explained deaf people were only being brought in to road test the AI, by which time it was too late.

She said: “The NHS has already had to withdraw an AI generated BSL video about measles because the grammar was incorrect, and the meaning therefore became entirely misleading.

“People who speak BSL as a first language were getting inaccurate information which can actually have really dangerous consequences. If you don’t talk to the people in the first instance who it’s supposed to be for, it’s really at odds with how we improve disable access to services, and it also risks wasting quite a lot of government money. It can also feel like taking something away that is theirs, and making money off it.”

READ MORE: ‘I’m deaf and shouting at people with hearing loss doesn’t help – do this instead’READ MORE: All 17 changes Universal Credit claimants need to report or risk losing payments

Rebecca Mansell, Chief Executive of the British Deaf Association, suggested AI could be helpful, but that quality control was essential. She said: “We can’t allow technology to simplify or distort BSL, or to put people off learning it. BSL translations should be universally available, but it is vital that they are accurate and of high quality. The signing community must make final decisions on what is acceptable.

“There are places where human judgment and trust are essential to effective translation. “Without quality control by independent deaf signers, it would be unthinkably risky for a deaf witness in court to rely on a translation avatar to give evidence, or for someone to receive a cancer diagnosis from a computer-generated signer. That’s why the Minderoo Centre for Technology and Democracy at Cambridge University has set out a deaf-led framework for government spending on AI-BSL — putting deaf signers in the driving seat, where they belong.”

There were also warnings from Victoria Boelman, Director of Insight and Policy at the Royal National Institute for Deaf People. She said: “AI has the potential to transform inclusion for British Sign Language (BSL) users, but without meaningful involvement from BSL users at the very beginning, it risks creating new barriers.

Article continues below

“More than 150,000 people in the UK are BSL users, 87,000 of whom are deaf and 21,000 use it as their main language. Without input from this community as well as robust regulation, there is a danger of BSL users being given incorrect information, which is particularly concerning in vital areas like healthcare, education and news.

“RNID is working alongside other charities in the sector to make sure AI-driven BSL is developed in a fair, ethical and inclusive way to ensure it benefits BSL users across the UK.”

DisabilityMeaslesNHSPoliticsUniversity of Cambridge