Instagram Now Using AI to Moderate Comments. Going Too Far?

Instagram has announced a new method to filter out inappropriate comments. The Facebook-owned platform will use an artificial intelligence system called DeepText as a monitor. Is this a test run for Facebook? How well can this work? Read more to find if Instagram is going too far?

Instagram Now Using AI to Moderate Comments. Going Too Far?

Previously, Instagram allowed you to turn on a filter to block any content containing pre-selected words or phrases. You could also choose specific words or phrases you didn’t want to see. The new DeepText system has been trained to ‘recognize certain types of offensive and spammy comments.’ However, Instagram has not revealed how the system actually works.

DeepText Overview

The new AI system is turned on by default. However, users can manually turn it off via the settings menu. Once a comment gets flagged no one will be able to see it. The person that wrote it won’t be informed that his/her comment has been blocked.

The DeepText system was first tested to block spam last October. It was a success in eliminating spam in multiple languages. Now, Instagram is using it to target abusive language. The AI algorithm was trained, tweaked and tested by a special team. They went through millions of comments. Different words were sorted into different categories such as bullying or racism. At the moment, it only works in the English language. However, spam can be detected in Spanish, Portuguese, Arabic, French, German, Russian, Japanese, and Chinese.

Also read: How To Post On Instagram From Your Desktop

Pros & Cons

The new system will encourage users to express freely. It offers protection against toxic or offensive comments on posts and in live video. The text classification engine is meant to help machines interpret words as used in context to fight spam.

However, this algorithmic censorship bears the same problem as any kind of stereotyping or profiling does. Devoid of context, a lot of innocent content will be unfairly condemned. Many words and phrases convey a different meaning in different parts of the world. Various slangs in one country are considered hate speech in other nations. Certain songs use offensive language too. How many iterations of the ‘F word’ are there? The seriousness of certain slurs also depend on the shared relationship between users? It is almost impossible to program an algorithm to understand such varied contexts.

Instagram stated that the new system will occasionally get it wrong. Apparently, the app is willing to put up with a certain margin of error. There are all kinds of problems with this attitude, with ‘freedom of speech’ being the first. The Facebook-owned app has faced few controversies in the past in this area. The new system takes us one step closer to an Orwellian future.

Also read: How To Shoot Good Videos On Your Smartphone


Instagram aims to create a safe community by using the new censorship system. However, it comes with its own problems. We will have to wait and see how it turns out.

Since Instagram is already a lot friendlier than most other internet social networks. We might soon see this technology bleed over to Facebook and other social media networks. In response to recent controversies, Facebook hired thousands of additional human moderators. It may not be long before AI can play a role in moderation on Facebook as well. Earlier this month, Facebook CEO, Mark Zuckerberg stated that ‘AI can help keep our community safe and so we have a responsibility to pursue it.’ While the new DeepText system might not be moving to Facebook as yet, the company is surely looking at Instagram as a test case.

More related stories below:

Leave a Reply

Your email address will not be published. Required fields are marked *