Social Media Algorithms and Communication Theories

YouTube Algorithm Mechanics

• Videos are "pulled" for each user when they visit YouTube, rather than being universally "pushed" or promoted by the algorithm.
• The primary goal of the search and discovery system is to match each viewer with content they are most likely to watch and enjoy, resulting in personalized recommendations for every user.
• Video ranking is determined by hundreds of signals, broadly categorized into two main areas:
  • Viewer Personalization: This category includes signals related to a specific user's preferences, such as: which videos they actively choose to watch, which videos they ignore, which videos they dismiss, and the frequency with which they engage with a particular channel or topic.
  • Video Performance: This measures the effectiveness of the content in engaging viewers when presented to them. Key performance signals include:
    • Whether viewers select to watch the video, ignore it, or click the "not interested" option.
    • The extent to which viewers watch the video, measured by metrics like average view duration and average percent viewed.
    • Viewer satisfaction, which is gauged through various signals, including user surveys and the number of likes a video receives.
• The system itself holds no inherent opinion on the type or format of video content; instead, it strives to follow the audience's demonstrated preferences.
• Creators are advised to focus on understanding their audience and their likes, rather than attempting to become experts in algorithms or analytics.
• Modifying a video's title or thumbnail can influence its ranking in the algorithm. However, this occurs because changes alter how viewers interact with the video, not solely due to the act of alteration itself. If a video is already performing well, it's generally recommended not to change successful elements.
• An individual video's underperformance does not necessarily harm an entire channel. The system relies more on video- and audience-level signals when making recommendations.
• A decline in overall channel views is more likely caused by shifts in audience interest over time or a reduction in how often viewers watch most recommended videos from the channel.
• A video's monetization status (e.g., a yellow icon indicating limited monetization) has no impact on its discovery through YouTube's search and recommendation systems.
• Key Learnings for Creators: Identifying and understanding a target audience (including demographics and psychographics) is paramount. It is more advantageous to cultivate a smaller, dedicated, and loyal audience (e.g., $1,000$ loyal viewers) than a larger, disengaged, and random one (e.g., $100,000$ casual viewers).
• Social media success is often measured by MAUs (Monthly Active Users), where "active" implies meaningful engagement like liking and sharing content.
• YouTube's Core Objective: To maximize user engagement by encouraging clicks and sustained viewing, thereby increasing ad revenue.
• Critical Metrics for Creators: Click-through rate, average view duration, average view percentage, and relative retention.
• MrBeast's Strategic Approach: MrBeast exemplifies a scientific and strategic approach to content creation, meticulously studying the algorithm and trending topics. His early viral content, such as repeating "PewDiePie" a million times, was a deliberate strategy to leverage the popularity of the platform's top creator at the time.

Facebook Algorithm Evolution and Priorities

• Facebook's algorithm is one of the earliest, originating with the News Feed in $2004$.
• The "Like" button was introduced in $2007$.
• The algorithm has continually evolved, including implementing various sorting mechanisms for News Feeds.
• By $2016$, reactions were weighted as more significant signals than simple likes, and the algorithm began prioritizing content based on its substance rather than merely the amount of time a user spent on it. Features like "See First" and "meaningful interaction" updates were also introduced.
• A "news credibility update" was implemented in $2020$.
• Core Priority: The algorithm is designed to prioritize posts from local family and close friends over business-related content, reflecting Facebook's emphasis on "Family" over "Business." Consequently, businesses should seek to establish a personal or familial connection within their content to enhance visibility.
• Definition of "Good Content": For Facebook, good content is engaging, helpful, specific, and inspires user interaction such as liking, commenting, sharing, or saving. An example contrasting a vague post with one offering specific steps and images for protecting Facebook Live videos demonstrated how detailed, actionable content receives significantly higher engagement.
• Facebook is recognized as an innovator in the social media space, investing billions in research and development and often leading industry trends.

TikTok Algorithm

• Key factors influencing TikTok's recommendations include user interaction, video details (e.g., captions and hashtags), and device and account settings (like country and language for geofencing).
• Captions are vital for algorithm performance and to accommodate users who watch videos with audio muted.
• Unlike some platforms, TikTok's recommendations are not directly based on the number of followers a user has or a video's immediate past success.
• The platform primarily values good video content, aiming to provide all creators with an opportunity for their content to be seen by users, friends, and other creators.
• This characteristic makes TikTok a valuable platform for marketers targeting a younger demographic with creative video content.

Twitter (X) Algorithm

• Primary Ranking Factors: Recency, morality (sharing), relevance, geographical location, and personal interest.
• Recency: Given Twitter's origin as a platform for "what's happening right now" and breaking news, the algorithm highly prioritizes new and current information.
• Evolution: Initially, in $2006$, posts were ranked purely by timeline (day and time). By $2023$, the algorithm evolved to consider a broader range of factors, including subject specialization and demographic relevance.

Instagram Algorithm

• Primary Ranking Factors: User interest, relationship with the user, following, session time, time posted, and overall user session time.
• Content Prioritization: Carousels tend to generate $3$ times more engagement than other post types. Reels are actively boosted by the platform to encourage the adoption of this newer feature. Brands are encouraged to experiment with new features across all platforms.
• Key Engagement Metric: "Sends per reach" is a crucial indicator, highlighting that content users share with friends tends to perform best, as it drives significant value and community connection.
• Mandy's Reels Case Study:
  • Successful Content: A clear, obvious cat video resonated with its intended audience, making it easy for the algorithm to find viewers who enjoy cat content.
  • Unsuccessful Content: A vague, talking-head video on "how to get more traffic to your website" lacked on-screen text, visual cues (like a computer), and keywords. This made it difficult for the algorithm to identify its purpose and target audience, leading to low views.

LinkedIn Algorithm

• Distinct Audience and Purpose: LinkedIn serves a professional and business-oriented audience, expecting specific types of content.
• Priorities: The algorithm emphasizes strong, relevant, and engaging content, often favoring longer-format articles or documents over short posts.
• Key Factors:
  • Hashtags: Recommend using between $3$ to $10$ hashtags per post.
  • Video Content: While popular, it is no longer universally prioritized in the feed as it once was.
  • Comments: Hold more weight than likes or reactions in terms of engagement signals.
  • Dwell Time: The duration a user spends watching or reading a post is important.
  • Network Building: Encourages engagement with relevant individuals, active participation from employees, and involvement in LinkedIn groups.
  • Considers LinkedIn ads and analytics for reach and understanding content performance.
• The algorithm aims to promote engagement and prioritize relevant content. Since $2018$, LinkedIn has witnessed a $50\%$ year-over-year increase in viral activity. High-quality, "read-worthy" content is crucial for building a network, even with a limited number of initial connections. Content must be thoughtful and effective, going beyond superficial trends.

Communication Theory 1: Agenda-Setting Theory

• Core Concept: This theory posits that the media, particularly social media, effectively tells us "what to think about" by highlighting certain issues and topics, thereby influencing public perception of their importance. It is not necessarily about telling people what to think, but rather what to think about.
• Application to Social Media: Social media platforms themselves operate with an agenda: to maximize user time spent on the platform. This is achieved through algorithmically determining which topics and content gain prominence for individual users, thus setting their personalized agenda.

Communication Theory 2: Cultivation Analysis Theory

• Core Concept: This theory suggests that prolonged, consistent exposure to the same messages across various media forms can "cultivate" or teach a common worldview, shared roles, and collective values among viewers. It examines the cumulative effects of systems of messages, including cultural themes, images, lessons, and values.
• Application to Social Media: The continuous bombardment of similar messages on social media platforms (e.g., a trending song like Beyonce's, political narratives, or viral memes) can make certain information seem more important or widely accepted. This saturation can alter users' perceptions of reality; for instance, frequent exposure to crime dramas might lead individuals to perceive the world as more dangerous than it objectively is, even if they recognize the fictional nature of the content.
• Political Cultivation Examples: The presentation meticulously details how major news outlets (CNN, MSNBC, Fox News) employ biased language and framing in their social media posts during political events. For instance, headlines about presidential candidates are crafted to cultivate specific biases, annoyance, or support among their respective audiences (e.g., calling a candidate "Crooked Joe" or describing a mugshot as "powerful" versus "greatest humiliation"). This deliberate framing aims to solidify audience loyalty and maintain viewership, illustrating the significant financial stakes (e.g., Fox News being a multi-billion dollar entity).

Communication Theory 3: Uses and Gratifications Theory

• Core Concept: This theory focuses on understanding why and how individuals actively choose specific media to satisfy particular needs and desires. It asserts that people are proactive media users who consume content based on the gratifications they expect to receive.
• Distinction from other theories: While Agenda-Setting and Cultivation theories analyze the effects of media on the audience (media acting upon viewers), Uses and Gratifications examines how the individual selects and uses media (viewers acting on media).
• Application to Social Media: This theory explains why users commit time to social media (e.g., to relax, unwind, or be entertained). Algorithms are designed to learn these individual gratifications to continually provide relevant content, thereby maximizing engagement and time spent on the platform.
• Common Gratifications for Social Media Use: People engage with social media for a variety of reasons, including:
  • $1$. Social interaction
  • $2$. Information seeking
  • $3$. To pass time
  • $4$. Entertainment
  • $5$. Relaxation
  • $6$. Commutatory utility (communication with others)
  • $7$. Convenience utility (easy access via mobile devices, free Wi-Fi)
  • $8$. Expression of opinion
  • $9$. Information sharing
  • $10$. Surveillance (gaining knowledge about others, often referred to as "stalking" exes or checking out new acquaintances).
• Platform-Specific Gratifications:
  • TikTok: Primarily used for entertainment, convenience, social interaction, and mental/physical health influences (e.g., seeking positive content, or as a distraction from personal thoughts).
  • Facebook: Used for relationship creation, escapism, passing time, information seeking, entertainment, disclosure, affection seeking, and social surveillance.

Four Agendas of Mass Media (Revisited in Social Media Context)

• Entertain: Social media unequivocally fulfills this role.
• Educate: Users can acquire knowledge and learn new things through social media.
• Inform: Social media provides details and updates on various topics and events.
• Gatekeep: Despite users believing they see a complete picture, social media acts as a powerful gatekeeper. Algorithms curate content based on individual interests and past behavior, meaning users do not see the entire spectrum of information. For example, personalized feeds deliver content aligned with an individual's existing viewpoints (e.g., political stances), filtering out contradictory or uninteresting perspectives.

General Advice for Social Media Use and Content Creation

• Experimenting with new features on any social media platform is encouraged, as it can boost engagement and visibility.
• Strategic posting times are crucial; content should be published when the target audience is most active (e.g., TikTok users are often active at night, while LinkedIn users are typically active during business hours, such as $9$AM to $5$PM).
• It is essential to be aware of the manipulative potential of social media, as it can significantly influence moods and behavior, capable of evoking both sadness and happiness.