How Instagram is Implementing Machine Learning to Help W/ Cyber Bullying

Instagram strives to irradiating cyber-bullying by leveraging machine learning.

I have never been an advocate of social media. I grew in the Bronx, New York and have observed many members of my community chasing what I call the “social media fix”, trying to create a false perception of their financial wealth or gauging their personal success on the postings of other social media members. More importantly, I have witnessed people being bullied on social media platforms; pictures of people being posted with captions such as “who thinks he or she is ugly”. This form of cyber bullying has led to many people becoming depressed and/or committing suicide1.

This is a serious problem since many children are now growing up in the “cyber era”, spending a vast amount of time on the internet and social media platforms. Subsequently, it is important that industry leaders such as Instagram and Facebook are commit to irradiating this problem.

Instagram and other social media sites have received criticism in the past for not addressing bullying. In 2017, a studied conducted by Ditch the Label showed that Instagram was worst social media network for cyber bullying2. The study examined 10,000 Instagram users and revealed that 714 were bullied. Many anti-social media groups have formed urging parents to not allow their children to join social media3. As Instagram strives to grow their community it is essential that they foster an inclusive community to ensure their members safety.

Instagram, under public scrutiny and having several members of their community complain about bullying, has decided to act4. Instagram recently introduced a machine learning tool to help them detect bullying. This will be done by scanning captions and pictures for potential signs of bullying. Instagram pictures and captions that are flagged will be reviewed by an Instagram employee5. If it is determined that bullying is taking place, the post will be removed, and the culprit’s user account will likely be deleted. This process will be refined as information is obtained. As Instagram employees review posting that are flagged by the machine learning filter, they will be able to adjust the filter’s algorithm to better identify bullying patterns6. This is a form of supervised machine learning, which I learned during a Bridgewater case discussion in my Leadership course (supervised learning is when data is inputted and compared to a data set for matches. The inputted data and search criteria can be refined) 7. This will increase efficiency helping Instagram to more efficiently capture bullying.

Previously, Instagram relied on community members to self-police the platform8. This process proved to be antiquated as many community members failed to report teasing and the defamation of one’s character. Many members often encouraged bullying by “liking” (endorsing a post by clicking the like button on a photo) postings that teased someone. By leveraging machine learning, Instagram will be able to better protect their community from bullying. The machine learning filter will be able to scan Instagram’s entire platform for bullying. This initiative symbolizes Instagram’s commitment to protect their members and is essential to Instagram’s business as they look to grow their client base.

While implementing machine learning to identifying bullying is a good step, more is needed to eradicate bullying. Instagram currently only requires an email address or phone number to create a user account. Since an email address is relatively simple to create and a Google Voice phone number is easy to obtain, cyber bullies can create an unlimited amount of Instagram accounts. This means that if a members account is deleted from Instagram, he or she can recreate a new account in minutes, by leveraging a different email address or phone number. This makes it extremely hard to keep bullies off of Instagram. In order to ensure that bullies don’t create a new account and stay off of Instagram there needs to be additional account creation verification measures. This can include a potential user providing personal information such as their name, birth date and home address. This will help improve processes. First, Instagram could run this information through a public database to verify the person’s identity. Secondly, Instagram can scan their member database to ensure that the user has never created an account that was associated with bullying.

I commend Instagram for their commitment to anti-bullying. Their ability to leverage machine learning to create a better community for their members highlights the value of machine learning. I wonder how artificial intelligence will affect the workforce. According to the Harvard Business Review, there is a “growing range of new job opportunities in the fields of big data analysis, decision support analysts, remote-control vehicle operators, customer experience experts, personalized preventative health helpers, and online chaperones”9 to support cyber bullying and harassment. I hope that Instagram’s usage of machine learning will decrease bullying and lead to more jobs. (793)

 

Sources:

  1. Young N. (2017). Mum puts daughter, 10, on ‘suicide watch’ after bullies used her picture on ‘Ugly or Not’ Instagram poll, Whimn (Kidspot), Retrieved From: https://www.kidspot.com.au/parenting/real-life/in-the-news/mum-puts-daughter-10-on-suicide-watch-after-bullies-used-her-picture-on-ugly-or-not-instagram-poll/news-story/c29c3d030c28343d5d2aaab23ee81b04
  2. Gibbs C. (2017). Instagram is the worst social network for cyber bullying: study, Daily News, Retrieved From: http://www.nydailynews.com/life-style/instagram-worst-social-network-cyber-bullying-study-article-1.3339477
  3. Sundstorm (2015). Experts to parents: Don’t let kids use Instagram before 13, Sunshine Coast Daily, Retrieved From: https://www.sunshinecoastdaily.com.au/news/age-limit-for-social-media-users-backed/2562977/
  4. Lorenz T. (2018). Instagram Has a Massive Harassment Problem, The Atlantic, Retrieved From: https://www.theatlantic.com/technology/archive/2018/10/instagram-has-massive-harassment-problem/572890/
  5. Constine J. (2018) Instagram now uses machine learning to detect bullying within photos, Tech Crunch, Retrieved From: https://techcrunch.com/2018/10/09/instagram-bullying-photos/
  6. Holsen L. (2018) Instagram Unveils a Bully Filter, New York Times, Retrieved From: https://www.nytimes.com/2018/05/01/technology/instagram-bully-filter.html
  7. Hume K. (2016) How to Spot a Machine Learning Opportunity, Even If You Aren’t a Data Scientist, Harvard Business Review, Retrieved From: https://hbr.org/2017/10/how-to-spot-a-machine-learning-opportunity-even-if-you-arent-a-data-scientist
  8. Instagram (2018). Reporting harassment or bullying on Instagram. Instagram, Retrieved From: https://help.instagram.com/547601325292351
  9. Pistrui J. (2018) The Future of Human Work Is Imagination, Creativity, and Strategy, Harvard Business Review, Retrieved From: https://hbr.org/2018/01/the-future-of-human-work-is-imagination-creativity-and-strategy

 

Previous:

DeliveryHero: Managing over 150,000 restaurants in 40 countries using machine learning

Next:

JBS and Robot Butchers

Student comments on How Instagram is Implementing Machine Learning to Help W/ Cyber Bullying

  1. This is a very interesting use of supervised machine-learning in screening social media! I wonder how effective this technology will be though in eradicating cyber-bullying. My sense is that bullies will always try to find new ways of taunting their peers through social media and there aren’t enough consequences in place for bullies to stop this type of behaviour. Moreover, it also can become very subjective as to what is considered “bullying” and what is not, especially to a machine or person behind a machine that does know the relevant parties. How can they be sure it is not just two friends creating friendly banter vs. something more serious?

  2. This is a great example about how machine learning can help to solve some deep social problems. Cyber bulling is unfortunately a growing phenomena. One example is that, as Tynes points out: “race-related online victimization is on the rise, with 50 percent of adolescents of color reporting these incidents in 2013, up from 32 percent in 2011” [1]. This increase has partly been driven by the increased use of content sharing platforms that enable users to share content freely, such as Instagram. Therefore, I believe these content sharing platforms are liable to prevent this to happen.

    However, I have some concerns regarding the use of machine learning to eradicate this problem:
    – As united22 points out, will these processes limit freedom of speech for Instagram users?
    – In social networks, a lot of content is ironic and has double meanings. Will machine learning algorithms be effective in overcoming this complexity?

    Finally, I think that a corporation like Facebook Inc., which owns large assets (financial assets and data), should do more to fight bullying:
    – Could identification of cyber bullies lead to prevention of bullying in the offline world?
    – Could Instagram partner with police to prosecute cyber bullies?

    [1] Tynes, Brendesha. “Cyberbullying Is a Bigger Problem Than Screen Time Addiction”. The New York Times. August 24, 2016. https://www.nytimes.com/roomfordebate/2015/07/16/is-internet-addiction-a-health-threat-for-teenagers/cyberbullying-is-a-bigger-problem-than-screen-time-addiction, accessed November 14, 2018.

Leave a comment