Getting our language right

Getting our language right

I don’t want to call my daughter ‘beautiful’.  I shrink away from calling my son ‘tough’.  Even though both children can be this way, there’s something inside me which rejects using such terms.  It’s heart breaking really because at times these are the sincerest descriptors that well up in my heart, particularly when I see them acting in a way that makes me feel proud.  I also refer to my daughter as ‘tough’ and my son as ‘beautiful’ just as much when describing them.  

But knowing the barrage of gendered messages that they will still, unfortunately, be bombarded with as they grow up in our society, I am determined not to be lazy with my language and unwittingly reinforce, rather than counteract, these messages.  As an example, I add a few ‘she's’ and ‘hers’ into the Three Little Pigs story my daughter loves so much - a tale otherwise dominated by male pigs, a male wolf and male farmers to supply the straw, sticks and bricks.  In a way, I find myself shadow-boxing with society in an attempt to level a skewed playing field.

We have all grown up in a culture where a distinctly different lens is applied when describing men versus women.  A 2019 study  of the language used to describe men and women in over 3.5 million fiction and non-fiction books published from 1900 to 2008, highlights the top 11 adjectives used to describe males and females[1].

Unsurprisingly, it found that the adjectives applied to women overwhelmingly focused on appearance while those for men related predominantly to behaviour and personal qualities; Boys and men are ‘brave’, ‘rational’ and ‘honorable’, while girls and women are about ‘being kind’ and ‘aesthetically appealing’.  And although many of these books were published decades ago in a time when the social mores and gender norms were vastly different to what they are today, these texts still represent a literary canon that many of us have grown up with…and in addition inform many of today’s language learning algorithms.  

These algorithms are fed with data in the form of text material that is found online, hence if the language that we are producing, and publishing has a gendered bias then the algorithm will likewise be so.  An example of this occurred recently in Germany, where Uğur Şahin and Özlem Türeci, a husband-and-wife team from Bio N Tech, made a breakthrough COVID-19 vaccine in partnership with Pfizer (the first to be granted approval for emergency use in the Western world[2]).  Mr Sahin is the company’s CEO and Ms Türeci its Chief Medical Officer.  In addition to being a renowned cancer researcher with 20 years’ experience in her field, Ms. Türeci is also the head of the Clinical Development Department at Bio N Tech, Director of a peak research body funded by the German National Education and Research Ministry, and President of CIMT, the largest European association for cancer immunotherapy development.  

Yet when her name was entered into google.de the Google Knowledge Panel identified her as ‘wife of Uğur Şahin’ (as did the pre-filled response in the Google search bar), while her husband was affiliated first and foremost with his job title of BioNTech CEO.

There was no ‘husband of…’ reference for him.  Furthermore, why was this the case when searching ‘Özlem Türeci’ on the German Google site, but not so on either the US, French or Spanish equivalents? When the Franfurter Allgemeine Zeitung celebrates the ‘Vaccine Inventor’ (male, singular), television channel ‘n-tv’ gushes over the ‘humble visionary’(again, male, singular) and the Bild-Zeitung runs with the headline ‘Father of the German Vaccine Wonder’, we can begin to see why.  

Language matters and always has.  And now with the increases in artificial intelligence and language technologies, we must be aware that algorithms are not editors which can critique the language in use or its societal consequences.  Algorithms identify patterns and form systems presenting gender stereotypes and prejudices as true, if we present them as such by the language that we use.  

So can we change it?  Of course. Through sustaining an increased awareness that gender bias does exist in our language, when building or programming these machine-learning models. These models can be directed to ignore or counteract bias or being intentionally given less biased text to learn from.  

And if machines can learn through exposure, can’t human brains be similarly influenced?  A German study in 2018 tested the abilities of different groups of people to solve the ‘Specialist Riddle’ after reading a text on expertise written using either gender-neutral language, or male forms/pronouns as the generic.  The Specialist Riddle is the one where the father and son are in a car accident, the father dies and the son is taken to hospital where the specialist in the operating theatre says “I can’t operate on him, he is my son”, and we are asked how can that be.  Disqualifying those who were already familiar with this pickle, the remaining respondents who had read the gender-neutral text before being asked the riddle were about a third more likely to identify that the specialist could be a woman (the mother) than those who had read the text with male generics[3].  

Could addressing the subconscious gender biases, which exists in all of us, be as simple as being presented with the alternative, over and over and over again?  Could ‘you can’t be what you can’t see’ even apply to text? Is the increased language policing of gendered terms completely overblown?  Or a part of a crucial redefinition?  Undoubtedly, we will create systems and processes around us which will reflect our selves and our fundamental take on life back at us. Skewing the system will skew the result, and we would be wise to keep this in mind if we are to consciously choose the future we are creating, rather than deliver a shallow echo of opinions from days long past.

Because undoubtedly the ultimate language-learning machine is us.  

So, I can keep applauding my daughter’s toughness and strength, and continue championing my son when I see him express kindness, gentleness or vulnerability.  And I can keep changing up the genders of those damn three little pigs.  Maybe there’ll be a female version one day.  

[1] https://www.futurity.org/adjectives-gender-descriptions-books-2143682-2/

[2] https://www.spiegel.de/netzwelt/web/biontech-impfstoff-wie-google-eine-impf-forscherin-zur-ehefrau-macht-a-e8ace050-8329-4706-9543-8a4e3ff6cd34?sara_ecid=soci_upd_wbMbjhOSvViISjc8RPU89NcCvtlFcJ

[3] https://www.frontiersin.org/articles/10.3389/fpsyg.2018.00985/full

‍