The Addictive Algorithm
I wake up, and I open Instagram. I walk to class, and I’m scrolling through Instagram. I finish a homework problem, and I’m back on Instagram! Being a meme junky myself, the application provides some amazing content. At times, however, I feel enslaved to the dominant social media platform. It feels as if I’m in an endless cycle of continuous scrolling, liking, and sharing posts. Hours pass and I can’t seem to avoid the need to open the application once more.
Unfortunately, this is common among young adults worldwide. As of the year 2023, there are 4.8 billion social media users globally, representing about 60% of the human population (University of Maine, “Social Media Statistics Details — Undiscovered Maine — University of Maine”). To give a rough idea, this is equivalent to about 410,000 new social media users daily, and about 4.7 new users every second! Instagram alone boasts a whopping two billion users itself (Dollarhide, “Social Media: Definition, Importance, Top Websites & Apps”). Such extraordinary numbers can only suggest one thing — is this a new addiction?
To answer such questions, one must first consider what addiction is. It is the condition of being addicted to a particular substance, activity, or thing. It consists of three or more of the following symptoms: tolerance, withdrawal, difficulty controlling usage of addictive activity or substances, negative consequences such as a decline in mood and relationships, neglecting or postponing activities, spending significant time or emotional energy, and a desire to cut down on the particular activities or substances (Boyle, “Hijacking with Dopamine”). Usually, many only consider the most common forms of physical addiction such as substance abuse or behavioral addiction such as gambling, which both have prevailed for many decades. However, social media usage is a fairly new form of addiction that is characterized by the use of a variety of technologies to share information in real-time (Dollarhide, “Social Media: Definition, Importance, Top Websites & Apps”). As such, many may not even consider the fact that they may be addicted to social media themselves. How does one become addicted? Social media applications, such as Instagram, are specifically engineered to keep users engaged with endless scrolling and never-ending interactions.
To understand how social media addiction is formed, one must first understand how social media platforms, like Instagram, produce users’ endless results with each scroll. Have you ever scrolled on Instagram and noticed a trend? A new post that is very similar, or even better, than the last post you liked or scrolled past? The answer? A multi-stage recommendation system that produces fast, highly relevant media results in real-time. This machine learning algorithm is divided into 4 stages: 1) retrieval of media posts, 2) first-stage ranking of media posts, 3) second-stage ranking of media posts, and 4) final reranking of media posts.
In the retrieval stage, hundreds of relevant items are gathered based on heuristics such as trending posts and users’ recent interactions. A Two Towers neural network model is then deployed to generate user-media embeddings. Features, such as visual (i.e. image features), textual (i.e. comments), social (i.e. users’ followers, following, number of likes), temporal (i.e. posting time, engagement time), and metadata (i.e. demographics), are gathered from these embeddings. The model then predicts whether a user will positively interact with a specific media post or not.
After retrieval, similarity functions are employed to find items/posts similar to those interacted with by a user, creating a pool of recommended rankings. The first-stage ranking is less precise but scales for a large pool, while the second-stage ranking is more precise and produces the top 100 recommendations using a more sophisticated neural network model.
The final reranking calculates the probability of user interactions and employs a formula involving weights for various engagement types (Vorotilov and et al., “Scaling the Instagram Explore Recommendations System”). The system thus provides immediate recommendations in real-time, creating a continuous and highly personalized experience for users consisting of the most probable media posts.
Through the rapid delivery of personalized engaging content, users can grow evermore addicted to this application. Instagram and similar social media platforms also incorporate “sophisticated neuroscience knowledge to add addiction and habit-forming features” that manipulate users to remain engaged with the applications longer (Boyle, “Opinion: Wish the Law Could Protect Teens on Social Media? It Can — but It Needs an Update”). The anatomy of addiction and the “reward” system can help explain this phenomenon. From a habituation perspective, the human brain is structured to slowly form associations, with the help of the basal ganglia, between stimuli and response (Boyle, “Hijacking with Dopamine”). That is, we unconsciously associate a specific action taken with a response or reward. In the case of Instagram and its infamous recommendations, we associate liking, sharing, and scrolling past a specific post with getting even better relevant posts.
The “reward” system — consisting of the amygdala, nucleus accumbens, and ventral tegmental area (VTA) — plays an even bigger role in habit formations and addiction. These areas of the mesolimbic system are responsible for producing and sending dopamine to the prefrontal cortex — the area of the brain responsible for decision-making and problem solving (Boyle, “Hijacking with Dopamine”). As such, when we positively interact with a media post and scroll onto a similar yet better post, we unconsciously form an association and experience a dopamine hit, producing excitement. Manipulated social media users also earn small random prizes that reward their brains with more dopamine hits (Boyle, “Opinion: Wish the Law Could Protect Teens on Social Media? It Can — but It Needs an Update”). Through our prefrontal cortex, we then decide to keep scrolling and seeking rewards to experience the same feelings of excitement. This eventually leads us to scroll even more, forming a new behavioral addiction to social media.
Another study, by Wolfram Schultz, also found that the learning mechanism of the human brain pushes us to want more, but never less (“Dopamine reward prediction error coding”). It has served us well in human evolution, but can potentially harm us in terms of addictions and stimuli-reward-response associations. More specifically, the responses to rewards disappear if the rewards are predicted; but if more rewards are given than predicted, the dopamine neurons show more significant responses (Schultz, “Dopamine reward prediction error coding”). This ultimately affects our behavior and unconscious associations as we begin to expect more, using the previous rewarding event as a comparison baseline. According to Schultz, “the previously larger-than-predicted reward becomes the norm and no longer triggers dopamine prediction [error] surge,” indicating our endless cycle of wanting, and our growing dissatisfaction with what we currently have. As a result, our “reward” system is one of the root causes of physical and behavioral addictions as it becomes overstimulated, almost as if it were hijacked.
As we scroll endlessly through Instagram, the brain overstimulates our dopamine neurons as we associate our actions on the social media platform with receiving great, rewarding content. This leads us to continue scrolling without consciously deciding to stop until, what could possibly be, several hours later. This overstimulation from social media will eventually lead to feelings of depression, anxiety, low self-esteem, poor body image, and many more negative feelings (Boyle, “Hijacking with Dopamine”). Despite such negative emotions, we still search for that improved feeling of excitement as we humans aim for more, but never for less.
This never-ending cycle is obviously an addictive mess. So, how can we combat this? The first step, just like any other addiction, is to consciously realize we have a social media addiction. The next step is to understand that addiction develops from a combination of an individual’s biological mechanisms, the social environment surrounding them (i.e. interacting with social media platforms), and factors outside of their control such as Instagram’s highly accurate recommender system (Griffiths, “Why anything can be addictive”). We can then counteract our current biological “reward” system with better incentives and rewards for not utilizing social media for a certain time period. We can also consciously limit the time we spend on social media, such as once a month, to ultimately delay the immediate gratification we receive from the consistent dopamine hits (Hafner, “To Deal With Obsession, Some Defriend Facebook”). The idea behind these countermeasures is to combat our need to keep scrolling through Instagram; and limit the overstimulation of our dopamine neurons that produce a comparison baseline for wanting more, by replacing our bad addictions with good habits.
In the end, we are all human with the ability to consciously decide whether to enslave ourselves to social media, or break the chains of this addictive algorithm curated from machine learning models and neuroscience knowledge. In order to improve as individuals, we must choose the latter. We must continue to evolve and seek great rewards, but ensure that we are associating their responses with positive, healthy stimuli — not with enslaving social media platforms.
Works Cited
Boyle, Mary. “Hijacking with Dopamine.” 2024, UC San Diego Department of Cognitive Science.
Boyle, Mary. “Opinion: Wish the Law Could Protect Teens on Social Media? It Can — but It Needs an Update.” Tribune, San Diego Union-Tribune, 29 Dec. 2021, https://www.sandiegouniontribune.com/opinion/commentary/story/2021-12-28/social-media-addiction-teens-communications-decency.
Brailovskaia, Julia, et al. “Tell Me Why Are You Using Social Media (SM)! Relationship between Reasons for Use of SM, SM Flow, Daily Stress, Depression, Anxiety, and Addictive SM Use — an Exploratory Investigation of Young Adults in Germany.” Computers in Human Behavior, Pergamon, 30 July 2020, https://www.sciencedirect.com/science/article/abs/pii/S0747563220302636?via%3Dihub.
Dollarhide, Maya. “Social Media: Definition, Importance, Top Websites & Apps.” Investopedia, 31 Aug. 2023, https://www.investopedia.com/terms/s/social-media.asp.
Geyser, Werner. “How the Instagram Algorithm Works in 2024.” Influencer Marketing Hub, 30 Jan. 2024, https://influencermarketinghub.com/how-instagram-algorithm-works/.
Griffiths, Mark. “Why Anything Can Be Addictive.” BBC News, BBC, 25 Nov. 2011, https://www.bbc.com/news/health-15723834.
Hafner, Katie. “To Deal with Obsession, Some Defriend Facebook.” The New York Times, The New York Times, 21 Dec. 2009, https://www.nytimes.com/2009/12/21/technology/internet/21facebook.html.
Maria Hasprova, Stefan Zak, et al. “SHS Web of Conferences.” SHS Web of Conferences, https://www.shs-conferences.org/.
Schultz, Wolfram. “Dopamine reward prediction error coding.” Dialogues Clin Neurosci. 2016 Mar;18(1):23–32. doi: 10.31887/DCNS.2016.18.1/wschultz. PMID: 27069377; PMCID: PMC4826767.
University of Maine. “Social Media Statistics Details — Undiscovered Maine — University of Maine.” Undiscovered Maine, 29 Dec. 2023, https://umaine.edu/undiscoveredmaine/small-business/resources/marketing-for-small-business/social-media-tools/social-media-statistics-details/#:~:text=There%20are%204.8%20billion%20social,increase%20year%2Dover%2Dyear.
Vorotilov, Vladislav, et al. “Scaling the Instagram Explore Recommendations System.” Engineering at Meta, 8 Aug. 2023, https://engineering.fb.com/2023/08/09/ml-applications/scaling-instagram-explore-recommendations-system/.
