28.06.2018

Do the footprints of stereotyping and gender bias follow us in online environments?

by Emel Ince –  Project manager @ cApStAn

Online translation platforms such as Google Translation and Microsoft Translator have fuelled lively discussions. Recently, one such topic on social media has been the alleged gender bias of these platforms: Some users reported that, when translating phrases from gender-neutral languages such as Turkish and Finnish, male pronouns seemed to be associated with certain nouns, professions and adjectives such as ‘programmer’, ‘engineer’, ‘hardworking’ and ‘leader’.  From the range of illustrative examples provided, it seems unlikely that the choice of pronoun in English was random. Female pronouns were associated with ‘lazy’, ‘obedient’ and ‘nurse’.

For Turkish, there is no male or female distinction in the third person pronoun “o”. “O” can be used for females and males alike. Translation of Turkish content results in:

O ilginc.                         He is interesting.

O bir engineer.             He is an engineer.

O bir ahci.                     She is a cook.

O bir doktor.                 He is a doctor.

Are the programmers or the platforms to blame? Or perhaps the enormous text corpora that are being fed into the system upon which the translations are based are at fault? In the case of the corpora, it might be a telling symptom of a general gender bias that exists in the history of written texts of mankind.

The bias seems to be more salient in some languages than others. This is not necessarily a reflection of gender attitudes in the source language’s native culture. The gendered content may simply be the result of different grammar systems in different languages.

The bias is naturally most visible when translating into languages with stronger grammatical distinctions between male and female forms. Languages with unisex pronouns or syntax elements, for instance, remain less prone to the divisions. For example, French has the third person pronoun ‘il’/’elle’ as well as endings of adjectives and past participles to reflect gender, as opposed to the unisex pronoun “o” in Turkish.

In the latter case, with less or no grammatical gender distinction, the bias is unobservable, and even rendered impossible by the grammar. The idea that the structure of the language itself can promote or discourage gender distinction is probably a strong argument for language activists who try to make languages less gender biased. In Sweden, for instance, the first gender neutral pre-school was founded with the aim of freeing kids from social expectations. In this Swedish school, everyone uses “hen”, a genderless pronoun borrowed from Finnish.

While some activists seem to think that gender equality is enhanced by rendering gender specific words as gender neutral, in other instances, the exact opposite is done. The French word “professeur” is the word for “teacher”, and the same word is traditionally used for both male and female teachers. However, some gender conscious language users promote the use of the variant “professeure” (with an “e” added at the end) when talking about a female teacher (following the general French grammar principle where “e” is in many cases added as a suffix to render feminine agreement), to differentiate rather than neutralise.

Whether or not you found this article interesting, note that Google Translate will use the English pronoun “he” when translating that a gender-neutral person talks about boring issues.