Language, Translation, and Bias

by | Nov 27, 2020 | ICT4D |

translation bias

It took a day. A whole day. I couldn’t believe it. One day to get through 20-25 questions!

My colleagues brought together the staff who were conducting a survey in the communities with whom we worked. And they were painstakingly going through each question. Discussing it. Describing it. Even debating it at times. Each question was in English, but also translated into the two other main languages of the area. However, many of the members of the communities would want to speak in their own local dialect so the staff would need to translate in real time.

‘We need the staff to fully understand the questions so they can translate them well. We are spending a lot of time debating how to best translate the words as many english words can be translated various different ways. And depending on the choice you make the meaning of the question completely changes.’

I should not have been surprised. Nearly 20 years ago, I learned a language in which water and faeces was spelt the same, the only difference was what part of the word was emphasised. And I almost always got the emphasis wrong.

Language matters. And translation matters too. It’s another way bias can get in. Many software programmes are written in english. Many only have english or roman character compatibility. This causes challenges for Arabic and many asian languages. The staff member capturing the survey responses often needs to translate these responses into a language the digital device will accept. The staff member adds her or his understanding or interpretation to the response. And that filtering can be biased.

So spending a day on understanding 20 questions was time well spent. Even though I didn’t realise it at the time.

Photo by Felicia Buitenwerf


Submit a Comment

Your email address will not be published. Required fields are marked *