In society’s journey to be more inclusive of others and mitigate discriminatory behaviour, hate speech has become lost in the moderation as we have started to rely more on technology to monitor comments on social media platforms on the internet. However, this Stanford daily article by Chuying Huo may make us pause as to the accuracy of these programs. Due to the fact that machines learn on the basis of a “single truth-label”, many of these programs miss out on the context in which certain words or phrases are being used, and therefore some comments can be falsely labeled as “hate speech” when in reality, it is not.

This takes me back to the time when the social media site Tumblr went through a scrubbing of adult content and one of my own illustrations (shown below) were flagged as “inappropriate” and not meeting “community guidelines”. Ultimately, context and having a more nuanced touch to labeling and censoring certain content and comments is important.

But I digress, the need for contextualized algorithms must be improved upon in order to avoid any old problems such as words like “yellow” appearing in databases as hate speech in any and all context. And as this Stanford article concludes with this

“Transparency is the key to making platforms more accountable…

The more that can be done to get citizens and everyday people into the

decision-making [process], the more platforms can actually reflect

people’s social values and outlooks.”

Please take the time to read this amazing article for more in depth details into how HAI researchers tackle hate speech moderators!