How Instagram is Using Machine Learning to Stop Bullying
Instagram announced on Tuesday that they are bolstering their efforts to combat bullying by using machine learning technology. Advances in computer vision allow Instagram to optically scan photos posted on the platform to detect bullying and alert Instagram’s community moderators to review the post. In his first public statement as the newly appointed Head of Instagram, Adam Mosseri states “I’m proud to build on our commitment to making Instagram a kind and safe community for everyone”. The updated bullying campaign is designed with a particular focus on Instagram’s younger users, as Mosseri explains, “it will also help us protect our youngest community members since teens experience higher rates of bullying online than others”.
Earlier this year, Instagram increased its efforts in machine learning based caption filtering with its Bullying Comment Filter, which scans captions for words and phrases that have the potential to cause harm. With this feature, users have the option to opt out and hide offensive comments as well as words they choose themselves. This feature has now been extended to comments in live videos.
The use of machine learning has allowed Instagram to detect bullying en masse, which goes beyond the use of words and phrases. Instagram says their algorithms can detect attacks on personal appearance, character and threats to their users’ well-being and health. Seemingly unique bullying tactics such as comparing users with a split screen, rankings, ratings and text imposed on photos can be detected in stories and the feed. Human intervention is still required as each photo marked as suspicious will be inspected by an image moderator who will take action, including banning the bullying accounts if necessary.
As organisations move in a direction of having their own AI strategy we will see more examples of how machine learning can assist human roles such as Instagram’s moderators. Using Computer Vision, Deep Learning and Natural Language Processing (NLP) to detect bullying is a indication of the vast potential that AI has to promote wellbeing and a safer online community. At Northraine, we strongly believe in the potential of AI to help create positive changes in society, as reflected by our application of machine learning in aged care, mental health and disability support.
