Instagram Tries It's Hand at Refereeing Derogatory Comments

Instagram and Facebook are using DeepText to change the way they monitor SPAM.

Instagram and Facebook are using DeepText to change the way they monitor SPAM.


Freedom of speech is a basic tenet of American democracy. This right, which is protected by our first amendment, is acknowledged as one of the greatest aspects of our nation. However, it can sometimes be a double-edged sword.  On the one hand, we can express any idea or notion in any form we see fit. On the other hand, this expression can take the form of hatred and negativity.

We know this all too well, especially on the world wide interwebs. The internet is the undefeated champion of hate-fueled rants, perverted ideologies and social bullying. On the internet - no one is safe. Or, so we thought. Facebook and Instagram believe they can change the world’s dialogue for the better. With DeepText, a natural language algorithm, the platforms attempt to combat destructive speech by scanning and flagging endless amounts of web text, spam, and derogatory comments. The impact of this algorithm can change how users engage. But, will it be for the better?

DeepText was announced June of last year to scour scores of text with almost human-like accuracy. The process was extremely involved. It turns out to achieve human-like accuracy one has to recruit well...humans. Instagram utilized a diverse panel of researchers to score comments for context, content and intent. The goal was to discern whether comments were spam or not.  

After a thorough review, the Instagram engineers developed a baseline to create an algorithm to learn how humans understand the context of these comments, judging what people actually mean. Instagram wants the algorithm to evolve as language evolves. But, will the “human-like” algorithm ever pick up on the nuanced complexity that language can offer, especially for those subcultures that thrive in its subversion?

Tastemakers could be at risk by what DeepText might bring. They are ever-necessary purveyors of cultural growth, creating new meanings, idioms, and expressions of life, art and human emotion. However, seldom are they understood by the class creating these “intelligent” algorithms. Like it or not, the engineers creating methods like DeepText have their own set of beliefs and biases. And these engineers are bound to find data that are incongruent with their set of experiences. Think about it. A young, Black man from New Orleans tweets about their latest playlist feature on his social media accounts. Will the engineer in Silicon Valley know the slang terms he use? Probably not.

Beyond social media’s mainstream, there’s a thriving subculture of expression, which could be wrongly interpreted as derogatory. But, there comments are far from demeaning. The terms are often a way for people to both understand and entertain. But, will the algorithm, with all its learning abilities, be able to decipher the growing complexity of language that even we as humans misunderstand? While noble in their intention, censorship can be devastating for communities traditionally misunderstood. Attempts to combat slander should be met with equal attempts to combat cognitive bias. That way, we ensure social media platforms are not only safe, but more equitable, too.


Have you or someone you know been the victim of Instagram spamming or unwanted derogatory comments on your Facebook or Instagram account? Weigh-in and tell us about how you feel about DeepText and its implications in the social media universe.

In the meantime, be on the lookout to have DeepText and Instagram deleting your spam or unwarranted messages in the near future.