How Race is Changing Tech Speak
Master-Slave, Whitelist-Blacklist: Stigmatisation of semantics pops up every now and then. Now with NLP, the consequence is ever more pressing.
Key points:
- Language in Discrimination
- Artificial Intelligence
- Owning the Name
WHEN the #BlackLivesMatter movement began weeks ago, I saw a tweet on Twitter that left me an impression. I could no longer find it, but it was regarding the use of the term “blacklist” in cybersecurity lingo.
Commenters were quick to dismiss the idea, claiming nitpicking or racialisation of innoucous words.
Then a comment in the thread quickly turned the table. That reply was simple, it wrote, “I will leave this here,” followed by a video link:
“And I always asked my mother, I said, ‘Momma, how come is everything white?’ I said, ‘Why is Jesus white with blond hair and blue eyes? Why is the Lord’s supper all white men? Angels are white, the Pope, Mary, and even the angels.’ I said, ‘Mother, when we die, do we go to Heaven?’ She said, ‘Naturally we go to Heaven.’ I said, ‘Well, what happened to all the black angels?’
So anyway, I was always curious. I always wondered why Tarzan is the King of the Jungle in Africa, he was white.
I saw this white man swinging around Africa with a diaper on, hollering. Did you all see Tarzan over here?
And all of the Africans, he’s beating them up and breaking the lion’s jaw, and here’s Tarzan, talking to the animals.
And the Africans have been there for centuries and they can’t talk to the animals. Only Tarzan can talk to the animals. I always wondered why.
And Miss America was always white. All the beautiful brown women in America, beautiful sun tans, beautiful shapes, all types of complexions, but she always was white.
And Miss World was always white, and Miss Universe was always white.
And then they got some stuff called White House cigars, White Swan soap, King White soap, White Cloud tissue paper, White Rain hair rinse, White Tornado floor wax, everything was white.
And the angel fruit cake was the white cake and the devil food cake was the chocolate cake.
I said, ‘Momma, why is everything white?’ I always wondered. And the President lived in the White House.
And Mary had a little lamb with feet as white as snow, and Snow White, and everything was white.
Santa Claus was white and everything bad was black. The little ugly duckling was the black duck, and the black cat was the bad luck. And if I threaten you, I’m going to blackmail you.
I said, ‘Momma, why don’t they call it ‘whitemail’? They lie too.’
I was always curious. And then this is when I knew something was wrong.”
I dropped everything.
Suddenly, the claim that “blacklist” is inherently racist is not so ludicrous anymore. Muhammad Ali’s eloquent discourse on structural racism in our very culture was a turning point for me. And for many others.
It is easy to defend semantics based on their etymology and history, some as long as language itself. Then again, words are anything but unchanging. English words frequently go out of favour. We don’t say “Good dawning to thee, friend” without turning heads.
Fast forward a week, GitHub announced the revision of its naming convention for the master-slave branch. The term “master branch” has hit a raw nerve with the tech-sphere for its reference to slavery.
Many people have began acknowledging it now and followed suit.
Not to mention U.K. cybersecurity agency have already began this movement weeks before with “blacklist” and “whitelist”, leading Github:
The terms “blacklist” and “master branch” are signs of a larger implicit discrimination present in language that cannot be ignored. It is largely invisible to the unaffected, but one can never know the cognitive bias it manufactures and, if left uncorrected, may continue to corrupt our youths.
If there shall be progress for humanity to eventually graduate from this centuries-old plague of racial unrest, then perhaps language is the place to start.
Language in Discrimination
THE role of words in perpetuating stigmas is present in our daily lives. We use words to express anger, insinuate ill-intent, rile and derogate.
As far as utility is concerned, some are unnecessary. As far as meaning is concerned, it’s time they evolve.
Language and power
F-word, S-word, and the A-words come to mind.
In English, the origin of expletives can be boiled down to class difference, particularly back to medieval England when the Normans invaders held higher social status than the Anglo-saxon people. Words from Norman origins eventually became more polite than the uncouth Germanic tongue of the Anglo-Saxons.
In tech space, expletives are powerful and effective utterances to signify to other people that the situation at hand is so bad it is worth spoiling your social conduct for. Especially in the face of a coding bug. F— that God d— bug.
Jokes aside, what’s more important is we know bad words, including those more nuanced than explicit racial slurs, can be traced back to class difference, and therefore power.
Respect in name calling
Some words have deep-seated history and the power to conjure up generations of pain and suffering. Others are archaic that should have long been abandoned, but still exists in subtlety, giving nuanced meaning to objects and people.
In respectful disability language, we have “person with mobility” and “cognitive disability” over stigmatised “handicapped and “mentally retarded”.
Whether in race or disability, respectful language is established to diminish the characterisation of the marginalised persons as their stigma and as inferior to the others without. Don’t go around calling people using terms such as “that girl in the wheelchair”.
Psychology in words
Ask a German and a Spanish speaker to describe an inanimate object like a bridge, and you can expect a gender prejudice.
The word “bridge” in German is feminine, while in Spanish masculine. Germans used words like “beautiful” and “elegant”, while the Spanish used “strong” and “dangerous”.
Psychology and linguistics researchers has long documented this effect on cognitive processing, and they agree that language can influence on our thoughts, hide stereotype in supposed post-stereotype world, and potentially bring the worse out of us.
Artificial Intelligence
IN the advent of artificial intelligence, Natural Language Processing has betrayed our cognitive bias.
Race and Language is so intertwined that Artificial intelligence are learning bias in Natural Language Processing. An example is the association between African-American sounding names with high score of criminality than European sounding names.
Anthropologist Samy Alim highlighted in his book, Raciolinguistics: How Language Shapes Our Ideas About Race, how former President Barrack Obama would change his speech pattern depending on the racial demographic of his audience to either maximise or minimise his African-American descent.
A.I. would have picked that up like a baby.
Racist bots
A notorious chatbot called Tay, known infamously as the Racist bot after an incident, was launched by Microsoft on Twitter in 2016. Tay was designed to organically learn speech patterns off real conversations with users on Twitter, with each interaction reinforcing learning towards a more sophisticated chatbot.
However, the experiment did not last a day when Tay began to spout racist slurs and make bold tweets that include denying the Holocaust.
The ironic tale exposes the inherent prejudice within colloquial lingua franca, and arguably demonstrated nurtured racism. In a killer-bot fashion, Tay learnt the ways of the human to hurt humans. Even now as a legend, the moral of the story still stings, the blame lands squarely on us. And our language.
How to change
To prevent chatbots from using offensive language, NLP systems usually use bla— blocklists to remove undesirable words that could possibly habour nuanced racial underpinnings.
In other cases, it is not that easy to completely remove the NLP bias due to the relationships between words in language; word embeddings, examples include “black” and “blacklist”, and “master” in “master thread”. Word vectors are integral building blocks of NLP which give numerical values to this relationships.
Ishant Juyal has great articles on NLP word vectors. In it, he discusses various models and vectors including, one-hot vectors, Word2Vec, CBOW, and Skip-gram.
In an increasingly A.I. dominant age, sorting out the information integrity of our vernacular is becoming urgent. Unlike mathematics which has rigorous interpretation that cannot be refuted, language is fraught(rich) with multiple interpretations. On the other hand, linguists and poets will bemoan with legitimate reasons the lack of subtlety in literature and prose if the richness of language were to dilute further in the digital age. Perhaps in the future, a new prototype of language will emerge, characterised by vocabulary whose ascribed meanings are methodical and absolute, used to almost exclusively communicate with our gadgets.
A kind of tech speak if you will, like a good version of George Orwell’s newspeak, dull but kinder in the necessary movement towards racial equality.
Owning the Name
“I am a mudblood, and I’m proud of it!”
Old Problem
The call for semantic changes because of equality issues is not new. To list a few: Gender, disability, age, race, etc; “man-hours” verus “work-hours”, “accessible” versus “disabled”, — you have seen enough.
“The Angel Food cake was the white cake, and the Devil Food cake was the chocolate cake” — Muhammad Ali
Now, the protest only proves the seething grievance that was left unseen for too long.
New Hope
As a double-edged sword, language can both capture and set free. Take a look at celebrated fiction.
In the book “1984”, we saw doublespeak used to ignite an insurgence against an oppressive power. Not far from the real world, people use the power of words to pressure and exact change.
With the revived activism, global rallying, and young people galvanised, there is an opportunity for social reform. It is too precious to miss.
“As far as meaning is concerned, it’s time they[words] evolve.”
One day against all odds, people would stand up like Hermoine, and declare: “I am a mudblood, and I’m proud of it!”
Reference:
https://drive.google.com/file/d/1bOyCvqHScML8xHri-KICkXdXzL88C4ne/view
https://www.pcmag.com/news/uk-cybersecurity-agency-drops-blacklist-and-whitelist-terms-over-racial
https://www.bbc.com/news/technology-53050955
Feel free to drop by, connect, or say Hello on my Medium and Twitter for more tech perspectives.
Or, read more:
