Chapter 10: Social Rewards and the Approval Economy — Why Likes Feel Like Love

Introduction: The Button That Changed Everything

The like button was introduced to Facebook on February 9, 2009. In the fifteen years since, it has become one of the most psychologically consequential design decisions in the history of digital technology — a small thumbs-up icon that has fundamentally restructured how billions of people relate to social approval, self-worth, and the act of sharing.

Before the like button, social feedback on digital platforms was qualitative: you left a comment or you didn't. After the like button, social feedback became quantitative: your post received a number. That number told you not merely that someone responded, but how many people responded, and by implication, how much your post — and by extension, you — were valued. The number was public. It was persistent. It was updating in real-time.

The designers who built the like button were, almost universally, trying to add something positive to the world: a quick, frictionless way to express appreciation without the effort of writing a comment. They were not trying to create an anxiety machine or an approval-seeking compulsion engine. But they were, in retrospect, doing exactly that — or at least providing the infrastructure for it. Many of them have said so publicly.

This chapter examines why a small thumbs-up icon carries such neurological weight. The answer requires going back much further than February 2009 — back to the evolutionary origins of social reward in the human nervous system, and forward through the specific ways that digital platforms have exploited those origins with precision and scale that no previous technology approached.

The core argument is this: platforms didn't invent social approval needs. They digitized them, quantified them, and weaponized them. Understanding how requires understanding both what social approval means to the human brain and what the like button actually did to it.

Social Reward in the Evolved Brain

Humans are, by any reasonable measure, the most socially complex animals on Earth. This is not merely a cultural observation — it is a biological one. The human brain devotes an extraordinary proportion of its architecture to social cognition: understanding other minds, tracking social hierarchies and relationships, predicting social consequences of actions, and managing social reputation. Neuroscientist Matthew Lieberman, in his 2013 book Social: Why Our Brains Are Wired to Connect, argues that the primary evolutionary driver of human cognitive expansion was not tool use or language per se, but the demands of navigating increasingly complex social environments.

For most of human evolutionary history — the two to three million years during which the brain's fundamental architecture was shaped — social acceptance and rejection were not merely pleasant or unpleasant. They were survival-relevant. An individual who was excluded from their social group faced dramatically increased mortality risk: reduced access to food, less protection from predators, fewer allies in conflict, lower reproductive opportunities. An individual who was socially included benefited from all of the above. Social standing was, in the literal sense, life or death.

This evolutionary history left deep marks on the brain's reward and threat detection systems. The neural response to social acceptance and rejection involves structures that are shared with responses to physical pleasure and pain — not metaphorically, but literally. Research by Naomi Eisenberger and Matthew Lieberman at UCLA, published in Science in 2003, used functional MRI to demonstrate that social rejection activates the dorsal anterior cingulate cortex (dACC) — the same region activated by physical pain. The colloquial experience of emotional pain as feeling "like a punch in the gut" reflects a neurological reality: the brain processes social rejection through many of the same circuits it uses to process physical injury.

The mirror of this finding is equally striking: social acceptance activates the brain's reward system — specifically, the ventral striatum and the medial prefrontal cortex — with the same kind of dopaminergic signaling associated with other primary rewards like food, sex, and warmth. Being liked, accepted, and valued triggers a genuine neurological reward signal. The social approval is not a secondary consequence of a more fundamental reward; it is a primary reward, evolved over millions of years of selection for social animals.

This is the neurological foundation on which the like button was built: a species for whom social acceptance is among the most powerful rewards available, and social rejection one of the most potent threats. The button did not create these responses. It created a new delivery mechanism for them, at unprecedented scale and frequency.

Eisenberger's fMRI Research: Pain That Isn't Physical

The 2003 Eisenberger, Lieberman, and Williams study deserves closer examination, because its findings have been replicated and extended in ways that bear directly on the social media experience.

The original study used a paradigm called "Cyberball" — a simple computer game in which participants played a ball-tossing game with what they were told were two other participants connected via the internet. In fact, the "other players" were controlled by a computer program. After an initial period of inclusion, the program began excluding the participant, with the two computer players throwing the ball only to each other and ignoring the actual participant.

This exclusion — from a trivial computer ball-tossing game with strangers who weren't even real — produced measurable activation in the dorsal anterior cingulate cortex, the region associated with physical pain experience. The participants weren't experiencing physical harm. They weren't losing anything of objective value. They were being socially excluded from a meaningless game. And their brains registered it with the same circuitry activated by a real physical injury.

Subsequent research has refined and extended these findings. A 2004 follow-up by Eisenberger and colleagues found that the right ventral prefrontal cortex, a region associated with emotional regulation, showed greater activation for participants who reported less distress during exclusion — suggesting that this region is involved in regulating the social pain response. A 2010 study by Ethan Kross and colleagues at the University of Michigan found that viewing a photograph of a former romantic partner who had rejected them activated the same regions (dACC and anterior insula) in participants who had recently experienced unwanted breakups as placing a heated probe on participants' forearms in a pain study.

The relevance to social media is direct. When Maya posts a photo and receives significantly fewer likes than she expected — when the platform delivers a quantified verdict of comparative social underperformance — the dACC regions associated with social pain experience do not distinguish between exclusion from a Cyberball game and exclusion from the digital approval economy. The circuits are the same. The response, while varying in intensity, is neurologically continuous with the physical pain response that evolved to protect organisms from injury.

Lieberman's broader research program adds a complementary perspective. His social brain hypothesis proposes that the default mode network — the brain regions most active when we are not engaged in specific tasks, previously thought to be the brain "idling" — is actually engaged in social cognition: thinking about other people, their minds, their relationships, their likely responses to us. The human brain, Lieberman argues, is never truly at rest. When given free time, it thinks about people. The social monitoring function is not a special-purpose system activated for specific social situations; it is the brain's resting state.

This makes the social reward signals of digital platforms particularly potent. They are not activating a specialized social-processing module that operates separately from other cognition. They are engaging the brain's default cognitive activity — its constant background social processing — with new, high-frequency, quantified inputs. The like button plugs directly into what the brain is already doing most of the time.

The Adolescent Brain and Heightened Social Reward Sensitivity

The social reward circuitry described above is active throughout life, but it is particularly sensitive during adolescence. Research on adolescent brain development has consistently found that the reward circuits associated with social evaluation — the striatum, the medial prefrontal cortex — are hyperresponsive during adolescence compared to both childhood and adulthood.

A 2016 study by Eva Telzer at the University of North Carolina found that adolescents showed greater activation of the ventral striatum in response to positive social feedback than adults did, and greater deactivation in response to negative social feedback. This heightened sensitivity is not pathological — it reflects the developmental logic of adolescence, a period during which building and managing social relationships outside the family becomes a central developmental task. The brain is tuned, during this period, to make social feedback feel particularly salient and significant.

Research by Sarah-Jayne Blakemore at University College London has mapped the specific neural developmental trajectory underlying this heightened sensitivity. The medial prefrontal cortex — critical for self-evaluation in social contexts, for thinking about how others perceive us — undergoes substantial structural and functional maturation during adolescence, with this maturation continuing well into the mid-twenties. During adolescence, this region is particularly responsive to social feedback, particularly status-relevant feedback about how one is perceived by peers. The brain is, in a developmental sense, doing something entirely rational: it is allocating maximum attentional resources to the social information that is most relevant to the developmental task at hand.

What makes the intersection of adolescent neurobiology and social media design so consequential is that these two things — a brain maximally tuned to social approval signals and a platform maximally engineered to deliver those signals — have been brought into contact without any design consideration for the developmental implications. The adolescent brain is not merely a smaller adult brain with the same social reward sensitivity. It is a qualitatively different neural system, at the specific developmental moment when social approval is most neurologically potent, interfacing with a system designed to exploit social reward circuits with laboratory precision.

Neuroimaging studies by Lauren Sherman and colleagues at UCLA, published in 2016, directly connected adolescent brain response to social media feedback. Using a simulated Instagram paradigm, Sherman's team showed adolescent participants photos and manipulated the number of "likes" each photo appeared to have received. Photos appearing to have many likes showed greater activation in the ventral striatum and medial prefrontal cortex — the social reward circuitry — than the same photos appearing to have few likes. The number, not the content, drove the neural response. And the response was stronger in adolescents than in adults.

What this means for the like button in the context of a seventeen-year-old like Maya is that she is not merely an adult experiencing social reward on a digital platform. She is a person whose brain is in the developmental phase for which social approval is maximally salient, interacting with a system specifically designed to deliver social approval signals in the most engagement-maximizing way possible. The intersection is not incidental. Platforms know their core demographic is young. The neurological vulnerability of that demographic is not a problem to be solved; it is a feature to be exploited.

The Quantification of Approval: What Changes When Social Acceptance Becomes a Number

Human beings have always sought social approval. The desire to be liked, admired, and respected is as old as the species. What the like button introduced was not the desire — it was the number.

Before quantified social feedback, social approval was received as a qualitative, contextual experience. You noticed whether people seemed pleased or displeased with you. You tracked your social standing through conversations, body language, inclusion or exclusion from activities, and the general quality of your relationships. This feedback was rich, nuanced, and embedded in specific social contexts. It was also relatively private: you knew roughly how you stood with the people around you, but that standing was not summarized as a public integer.

The like count changed this. For the first time, a person's social value — or at least a proxy for it — was expressed as a public number. The number was attached to specific pieces of content. It was visible to both the creator and anyone else who viewed the content. It was updated continuously. And crucially, it enabled direct comparison: your post got 47 likes; hers got 312. The comparison was inescapable, because it was made explicit in numerical form.

Leon Festinger's social comparison theory, developed in 1954, proposed that humans have a fundamental drive to evaluate their own abilities and opinions, and that in the absence of objective standards, they do so by comparing themselves to others. We naturally look around to see how we measure up. This tendency toward social comparison is normal and in many contexts adaptive — it helps individuals calibrate their performance and standing. But it can produce negative outcomes when the available comparison targets are systematically non-representative (e.g., everyone appears to be doing better than you) or when the comparison metric is something intrinsically subjective that comparison makes falsely objective (e.g., social desirability expressed as a number).

The like count creates both problems simultaneously. The available comparison targets are systematically non-representative: the posts that accumulate the most likes are more likely to appear in others' feeds (via algorithmic amplification), creating an environment where the typical visible post appears more liked than the typical post actually is. And the like count makes the intrinsically subjective — how much people value your specific contribution to their social environment — appear as an objective quantity, susceptible to the arithmetic of comparison.

How Numbers Changed the Subjective Experience of Approval

The transformation of qualitative social approval into a quantitative public metric produced changes to the psychology of social approval that go beyond mere convenience. The number does not simply measure social acceptance — it fundamentally alters the experience of it.

In pre-digital social environments, receiving appreciation for something you shared — a story you told at a party, a meal you cooked — was distributed across time, was variable in expression, and was embedded in relationship context. Ten people might have enjoyed your dinner party; you would know this through laughter during the meal, compliments at the end, invitations from those same people to their own events subsequently. The approval was real, but it was woven into the fabric of ongoing relationships and lacked a numerical summary.

The like count collapses this rich, distributed experience into a single integer. It strips away the relationship context (the 47 people who liked your post exist as a count, not as a set of specific relationships with specific histories and dynamics), it removes the temporal distribution (you know immediately, precisely, how many people responded), and it creates an artificial zero-point (zero likes is a quantitatively defined failure in a way that no equivalent social experience was previously definable).

Research on the hedonic psychology of quantification suggests that converting subjective experiences into numbers does not merely measure those experiences — it changes them. A 2008 study by Christopher Hsee and colleagues at the University of Chicago on "evaluability" theory found that when people evaluate things that are naturally continuous (how much do I enjoy this?), they are more satisfied when the quality is high and more dissatisfied when it is low after quantification than they would be with the same objective quality absent quantification. Making things measurable makes the differences between values feel larger. A post with 12 likes feels not just somewhat worse than a post with 120 likes — it feels definitionally worse, because numbers permit the kind of precise comparative evaluation that qualitative approval experience does not.

The implication is that quantifying social approval through the like count likely amplified both the highs and the lows of social feedback in ways that its designers did not anticipate. The person whose post does unexpectedly well does not merely feel approximately as good as they would have after a qualitative expression of appreciation — they feel the specific, precise surplus of their performance over expectation. And the person whose post does unexpectedly poorly does not merely feel the ambient disappointment of a lukewarm reception — they feel the quantified deficit, concrete and unarguable.

Research on social media use and wellbeing is substantial and consistent. A 2018 study by Jean Twenge and colleagues, published in Psychological Science, found that heavy social media use among American teenagers correlated with increased depression and loneliness — with the largest effects appearing around 2012 to 2013, precisely the period during which Instagram (launched 2010, acquired by Facebook 2012) became ubiquitous among teenagers. The correlation is not proof of causation, but the timing is striking, and subsequent experimental studies have shown that even brief interventions reducing social media use improve wellbeing outcomes.

For Maya, the like count is not an abstract social metric. It is, on the days she posts, one of the primary dimensions along which she evaluates herself. A post that does unexpectedly well produces a genuine emotional high — a feeling of social validation that translates into energy, confidence, and a willingness to post again. A post that does worse than expected produces anxiety, rumination ("What was wrong with it? Was it the photo? The caption? The time I posted?"), and sometimes the decision to delete. These are not trivial emotional experiences. They are the experience of social acceptance and rejection, mediated through a number and delivered with laboratory precision.

The Like Button's Origins and Evolution

The story of the like button begins in 2007, not 2009.

Justin Rosenstein was a 24-year-old engineer on Facebook's platform team when he proposed what he called the "awesome button" — a one-click way for users to express approval of a post without having to type a comment. The idea was simple: sometimes you want to indicate that you've seen and appreciated something, but writing a comment is too effortful for a casual positive response. A single click, publicly visible, would fill this gap.

Rosenstein's proposal was debated internally for approximately two years. There were genuine concerns about what a quantified public approval mechanism would do to user behavior and emotional wellbeing. Would it make users post more compulsively to accumulate likes? Would it make low-like posts feel stigmatizing? Would it change what people were willing to share? Facebook's leadership ultimately concluded that the benefits — higher engagement, more positive signaling, a new mechanism for social expression — outweighed the risks.

The "awesome button" launched on February 9, 2009 as the "Like" button, and within weeks it had transformed Facebook's engagement metrics. Users were posting more, spending more time on the platform, and checking back more frequently to see how their posts were performing. The quantified approval mechanism had created exactly the engagement loop its proponents hoped for.

Leah Pearlman, a product manager who worked on the like button's implementation and is often cited as one of its co-creators, has spoken publicly and with considerable candor about what happened next. In a 2017 The Guardian interview and in subsequent public statements, she described having developed her own problematic relationship with the like button she helped build. "I was at a point where I needed external validation to feel good about myself," she said, describing checking her posts' performance multiple times a day and experiencing genuine distress when posts received fewer likes than expected. She noted the irony: she had helped build a machine designed to generate the need for approval she now found herself caught in.

Rosenstein himself has been similarly candid. In a widely-cited 2017 The Guardian profile by Paul Lewis, he described the like button as something that was "bright, fun, and joyful" in conception, but acknowledged the evidence that it had contributed to dynamics he found troubling. He now limits his own social media use substantially, citing his awareness of the behavioral patterns it activates.

The like button spread rapidly beyond Facebook. Twitter introduced the "favorite" button (a star icon) in its early years and converted it to a heart icon in November 2015 — a change motivated partly by research showing that hearts produced stronger engagement responses than stars. Instagram, from its launch in 2010, included a heart icon for post appreciation. TikTok, launched in 2016, uses a heart system for videos. The specific form varies, but the core function — a one-click public approval mechanism attached to a visible count — has become the near-universal standard for social content platforms.

The like button is no longer a feature. It is the grammar of social media.

Variable Social Reward: The Unpredictable Heart of Engagement

Every post is a small gamble.

Maya posts the coffee shop photo with Priya. She hits share. And then she waits. She doesn't know how many likes it will get. It might get 14 — which happened, and which felt okay. It might have gotten 3 — which would have produced a quiet, unspoken anxiety and the background processing of what went wrong. It might have gotten 80 — which would have produced a genuine high, a screenshot-worthy moment, proof that something clicked.

This uncertainty is not incidental to the engagement power of the like button. It is central to it.

The reward prediction error mechanism, examined in Chapter 8, predicts that dopaminergic responses to social rewards will be amplified when the reward is uncertain at the time of posting and when the actual reward exceeds prediction. This is precisely the structure the like button creates. Every post is a behavioral gamble: the creator invests social and creative capital (the decision to share something about themselves), and the return is uncertain until the engagement accumulates.

Research specifically on social media post outcomes has confirmed this pattern. A 2020 study published in Social Cognitive and Affective Neuroscience used fMRI to examine brain responses when participants received social feedback on posts they had created. The dopaminergic response to social approval was stronger when the approval was unpredictable — when the participant was uncertain how the post would be received — than when approval was expected. The uncertainty is not a neutral condition to be tolerated before the reward arrives. It is itself part of the reward mechanism, priming the system for a heightened response.

This variable social reward structure has a documented behavioral consequence: it makes posting behavior more compulsive and harder to control than it would be if social feedback were predictable. When you know a post will get 50 likes, the engagement curve is flat — you post, you get 50 likes, done. When you don't know whether it will get 3 or 80, every posting decision carries the possibility of a high-magnitude reward, and every negative outcome is processed as more significant because you had been hoping for the high outcome. The gambling nature of the interaction — I don't know how this will land — is a design consequence, and it is the design consequence that most directly drives compulsive posting behavior.

Social Comparison on Steroids: Instagram as the Comparison Machine

If Facebook's like button created the basic structure of quantified social approval, Instagram perfected it as a comparison medium.

Instagram launched in October 2010 with a specific aesthetic premise: beautifully composed photographs, presented in a clean, consistent interface. The visual quality standard was higher than Facebook's messy mix of text, photos, and links. The platform quickly developed a culture of aspirational visual presentation — carefully composed photos of food, travel, fashion, fitness, and social life, filtered and captioned to convey a coherent personal aesthetic.

The comparison dimension was built in from the start: every post displayed its like count publicly, every account displayed its follower count, and the interface was designed to make direct comparison between accounts easy and natural. The "Explore" page showed you posts that were performing well — by definition, posts with high engagement relative to their account size. The feed showed you content from accounts you followed, but ranked, from 2016 onward, by predicted engagement rather than chronological order — meaning the posts most likely to generate high engagement floated to the top of the experience.

The result was an environment that systematically presented users with a non-representative sample of social reality. Festinger's social comparison theory predicts that people will compare themselves to the available targets — and on Instagram, the available targets were systematically better: more beautiful, more well-traveled, more stylishly dressed, living seemingly more enjoyable lives. Research on upward social comparison (comparing yourself to people who appear to be doing better than you) consistently finds that it decreases wellbeing, self-esteem, and mood.

The research on Instagram and depression, particularly in adolescent girls, is substantial and has been substantially amplified by internal Meta documents that became public in 2021. A 2020 study by Amy Orben, Andrew Przybylski, and colleagues at Oxford and the University of Cambridge found statistically significant negative associations between Instagram use and life satisfaction, with larger effects for girls than boys. A 2019 internal Facebook study, later reported by The Wall Street Journal, found that Instagram's own research had documented that 32 percent of teenage girls said that when they felt bad about their bodies, Instagram made them feel worse. The document noted: "We make body image issues worse for one in three teen girls."

The phrase "social comparison on steroids" is not metaphorical. Instagram combined three amplifying factors that make comparison effects more powerful than ordinary social environments produce: (a) universal visibility of approval metrics, making relative social standing explicitly quantifiable; (b) the algorithmic curation of high-quality, high-engagement content as the baseline of the experience, making it appear that everyone else's life is more interesting and their approval higher; and (c) the reach to thousands of followers for popular accounts, making comparison targets with numbers far exceeding anything achievable through ordinary social networks readily available and visually salient.

For Maya, the comparison is constant and textured. She does not merely notice that her post got 14 likes while someone else's got 312. She notices what was different about those two posts — the lighting, the composition, the caption, the person featured, the time of day, the hashtags used — and she builds a mental model of what gets approved and what doesn't. This model shapes her future posting decisions. The algorithm of her own mind has been trained, through repeated social feedback, on what performs. What she posts next will reflect what this training has taught her about what is worth sharing — which is to say, what is most likely to be liked.

The Comment and Reply as Higher-Order Reward

Within the hierarchy of social feedback signals, comments occupy a qualitatively different position from likes.

A like requires no cognitive investment from the person giving it. It is a single tap, a one-bit response: positive. Comments require more: they require the person to formulate a thought, type it out, and attach their name to it. The effort asymmetry makes comments a stronger signal — someone who comments on a post has invested meaningfully more in their response than someone who liked it.

This is reflected in both the neurological response and the behavioral engagement patterns. Research by Dar Meshi and colleagues at Michigan State University, published in Frontiers in Human Neuroscience in 2015, found that receiving comments — particularly substantive, personal comments — produced stronger activation in the ventromedial prefrontal cortex and the nucleus accumbens than receiving likes alone. The higher-effort signal carries higher reward weight.

Platforms have designed their notification and display systems to reflect and amplify this hierarchy. Comment notifications are typically distinguished from like notifications both in display prominence and in notification priority. Instagram notifies users of comments in real-time, with the commenter's name (unlike the vague "someone" of like notifications), because a named comment carries higher personal relevance. Platforms also surface comments prominently in post displays, placing them immediately below the image, while burying like counts in a less prominent position.

The comment response cycle — someone comments, the original poster responds, the commenter responds again — is the engagement pattern platforms most want to generate, because it produces extended platform time, high emotional involvement, and strong social bonding that creates retention. The design of comment systems is oriented toward generating this cycle: short reply threads are easy to generate, notifications pull both parties back when the other responds, and the visible exchange demonstrates social connection to other viewers, potentially drawing them into the conversation.

For Maya, the two comments on her coffee photo carry disproportionate psychological weight relative to the 14 likes. Priya's "love this" is a warm confirmation from a close friend. The classmate's "cute" is something to be analyzed: Does she mean the photo? The setting? Is she being sincere or slightly dismissive? The comment from someone she doesn't know well opens a question about what that relationship is or could be. These are genuine social dynamics, not trivial responses to an app feature. They are the platform successfully delivering social reward of the quality that actually matters to human beings — personal recognition from specific individuals.

Hiding Like Counts: The 2019 Experiment in Depth

In April 2019, Instagram began a series of tests in which the public like count was hidden from posts in selected countries: Canada, Australia, Brazil, Ireland, Italy, Japan, and New Zealand. The change was framed as an effort to reduce social comparison pressure and make the platform a less anxious experience for users who felt judged by their like counts.

The creator could still see their own like count. But visitors to the post could not — they could see who had liked it, but not how many. The number was hidden.

The experiment attracted intense attention from researchers, journalists, mental health advocates, and creators. The stated rationale — that visible like counts drive unhealthy social comparison — was credible and grounded in research. The question was whether hiding the count would actually change the psychological dynamics, and what effect it would have on platform behavior.

Methodology and What the Researchers Found

Evaluating the Instagram like-hiding experiment properly requires attention to its methodological limitations. Because Instagram did not conduct the experiment as a formal randomized controlled trial with pre-registered hypotheses, published protocols, and open data, the findings that emerged came primarily from independent researchers who worked with survey data and secondary measures rather than platform behavioral data.

Amy Orben and Andrew Przybylski conducted one of the more rigorous analyses of the hiding period. Their approach compared wellbeing self-reports and platform-use patterns among Instagram users in countries that implemented the hiding feature against users in countries where it was not implemented. Their analysis found modest reductions in self-reported social comparison anxiety among users in hiding-condition countries, with effects that were statistically detectable but small in absolute terms. Critically, the effects were more pronounced among users who identified social comparison as a primary motivation for their Instagram use, suggesting that the intervention was most effective for those it was most designed to help.

A complementary study by Tran and colleagues at the University of British Columbia examined whether the hiding feature changed actual posting behavior rather than merely self-reported anxiety. Their findings were less encouraging for the hiding hypothesis: posting rates, timing patterns, and the types of content shared showed little meaningful change between hiding and non-hiding conditions. Users continued to seek approval through their posting behavior; they simply had less direct numerical feedback on whether they had received it. The demand for approval, in other words, appeared to be more fundamental than the particular feedback mechanism through which it was satisfied.

The most important finding to emerge from the experiment, however, was methodologically indirect: studies that attempted to measure social comparison using alternative metrics — comment counts, follower counts, story view counts — found that these metrics became more salient to users when like counts were hidden. Users adapted by using comment counts and follower counts as proxy metrics — the comparison impulse remained, it simply found new numbers to attach to. For creators with large followings and monetization goals, the hidden like count was a significant problem: brands and sponsors use like counts as metrics for influencer campaign effectiveness, and creators couldn't easily demonstrate their reach without the visible number.

Why the Experiment Rolled Back

The creator backlash was substantial and commercially significant. Professional creators and influencers — the class of users who generate the most engaging content and who attract the advertising revenue that funds the platform — loudly objected. Without visible like counts, they argued, their ability to demonstrate value to brands was impaired, their audience couldn't see social proof of their content quality, and the social dynamics of the platform felt fundamentally altered.

Instagram began rolling back the change in most markets by late 2019 and through 2020. By 2021, the company had settled on a compromise: users could choose whether to hide their own like counts and whether to see like counts on others' posts, but the public default was restored to visible counts.

What the Experiment Revealed

What the experiment revealed — particularly in the way it was ultimately resolved — was that the business model's dependence on social reward visibility is structural, not incidental. The like count is not just a psychological mechanism; it is an economic mechanism. The visible number tells creators how they're performing, tells brands how many eyeballs a creator reaches, and tells users which content has been socially validated as worth their attention. Hiding it was an experiment in removing a feature that the platform's economic architecture had been built around.

The partial reversal of the hiding experiment was commercially necessary. But the episode produced valuable evidence: when researchers found that hiding like counts did not dramatically improve wellbeing outcomes — partly because users adapted by finding other comparison mechanisms — this was used to justify restoring the visible counts. What the argument missed is that the fundamental dynamic (quantified public social approval visible to the creator and their audience) was never fully removed. Even with hidden counts, Instagram was still a machine for delivering and measuring social approval. The number being visible or invisible changes the specifics of the comparison dynamic without changing its existence.

The experiment also highlighted an important distinction between proximate and distal causes of the psychological harms associated with social media. Visible like counts are a proximate cause of social comparison distress; the distal cause is the platform's construction of social interaction around quantified approval. Removing the proximate mechanism while preserving the distal structure produces modest and unstable effects, because users will reconstruct the comparison apparatus from available materials.

Cross-Platform Comparison: How Different Platforms Structure Social Rewards

The like button is not implemented identically across platforms, and the differences in implementation produce meaningfully different psychological dynamics. Understanding the social reward architecture of different platforms requires attending to the specific design choices each has made.

Facebook maintains the most complex social reward structure, having evolved from a simple like button into a "reactions" system (like, love, haha, wow, sad, angry) introduced in 2016. The reactions system represents an attempt to provide richer feedback signals — to distinguish between approval and empathy, between appreciation and amusement. Neurologically, the richer feedback creates slightly more complex reward signals: knowing that something you shared made people laugh (haha reaction) is a differently textured reward than knowing they approved of it (like). Research on reactions use has found that emotional content — content that provokes strong emotional responses — receives disproportionately more non-like reactions, consistent with Facebook's internal research (discussed in Chapter 5) showing that emotional content drives engagement.

Twitter/X has historically structured social reward around two mechanisms: likes (formerly favorites) and retweets. The retweet is a qualitatively distinct reward: it represents not merely approval but endorsement sufficient to share with one's own audience. The retweet count carries a different valence than the like count — being retweeted means someone thought your contribution was worth distributing, not merely worth acknowledging. Research by Jonah Berger and colleagues on what makes content shareable found that content that evokes high-arousal emotions — awe, anger, anxiety, and amusement — is shared more than content that evokes low-arousal emotions like sadness or contentment. Twitter's architecture, in which sharing is a distinct and rewarded behavior, has shaped the kinds of content that perform well on the platform in ways that favor emotional intensity.

TikTok operates a social reward structure notably different from its predecessors. While hearts (likes) and comments are present, the platform's distinctive social reward mechanism is the duet and stitch system, which allows users to create content in direct response to other users' content. This creates a social reward architecture in which the highest form of recognition is not a count but creative engagement — someone thought your content was worth building on. The system creates a more complex and arguably more meaningful form of social reward, while also creating a distinctive anxiety: your content might be responded to by a much larger account, bringing unexpected attention that is itself variable in valence.

Snapchat uses "streaks" — the count of consecutive days on which two users have exchanged snaps — as a social reward structure that is explicitly relationship-maintaining rather than content-quality-rating. A streak of 200 days with a friend is a number that represents the persistence and reliability of a relationship rather than the quality of any individual piece of content. Research on Snapchat use among adolescents has found that streak maintenance is a significant source of both social obligation and anxiety — teenagers report stress about maintaining streaks and distress when streaks break. The streak mechanism turns reciprocity into a metric, with consequences for how relationships are experienced and maintained.

LinkedIn applies the social reward structure to professional rather than personal identity, creating a distinct psychological profile for its approval dynamics. Professional validation — the endorsement of skills, the response to a thought leadership post — is explicitly status-relevant in a domain where status has material consequences for career and income. Research on LinkedIn use finds that the platform produces stronger self-presentation concerns and performance anxiety than platforms explicitly framed as personal or social, because the approval being sought is professional rather than merely social.

The cross-platform comparison reveals that while the like button in its original form established the basic grammar of quantified social approval, different platforms have implemented that grammar in ways that emphasize different dimensions of social reward: breadth (number of approvers), depth (quality of engagement), reciprocity (relationship maintenance), virality (reach amplification), or professional standing (status-relevant endorsement). Each emphasis creates distinctive psychological dynamics, and the typical user navigating multiple platforms simultaneously is managing multiple distinct social reward economies simultaneously.

The "Approval Addiction" Individual Difference Research

Not everyone responds to social media approval cues with the same intensity. Research on individual differences in social media engagement has identified a cluster of psychological characteristics that predict stronger approval-seeking behavior on social media and stronger emotional responses to approval and disapproval outcomes.

Sara Konrath and colleagues at the University of Michigan examined the relationship between narcissism (particularly the grandiosity and entitlement dimensions) and social media use. Their research found that individuals higher in narcissism were more likely to use social media for self-promotional purposes, to post more frequently, and to check their post performance more obsessively. However, they also found that the relationship ran in both directions: heavy social media use was associated with increases in narcissistic traits over time, suggesting that the platform's reward structure for self-presentation may cultivate narcissistic orientation, not merely attract it.

Complementary research by Ethan Kross and colleagues examined the relationship between attachment style and social media use patterns. Individuals with anxious attachment styles — characterized by preoccupation with relationship security and fear of rejection — showed stronger negative emotional responses to low-engagement posts and stronger positive responses to high-engagement posts than those with secure attachment styles. The approval economy, for anxiously attached individuals, mirrors and amplifies the approval-seeking and rejection-sensitivity dynamics that characterize their broader relational life.

Research specifically on "approval addiction" as a construct — the tendency to regulate self-esteem primarily through external social approval signals — has found this trait to be particularly strongly associated with problematic social media use patterns. A 2019 study by Andreassen, Pallesen, and Griffiths, using the Bergen Social Media Addiction Scale, identified approval-seeking as one of the strongest predictors of social media addiction symptom scores, including the inability to reduce use despite desire to, withdrawal symptoms when unable to access platforms, and continued use despite negative consequences.

The approval addiction framework is clinically useful because it connects social media behavior to a well-understood psychological construct with established therapeutic approaches. Cognitive behavioral therapy for approval addiction focuses on building internal sources of self-evaluation that are less dependent on external feedback, recognizing cognitive distortions in interpreting social feedback (all-or-nothing thinking about post performance, mind-reading about what low engagement means), and developing behavioral experiments that test the actual consequences of posting less strategically. These approaches have documented effectiveness in the general clinical context; their application to social media-specific approval-seeking is an active area of therapeutic development.

Maya's Extended Narrative: The Posting Cycle

Maya does not think of herself as approval-seeking. She would describe her relationship with Instagram as "just posting photos" and "staying connected with friends." These descriptions are partly true and partly a defense against a more uncomfortable analysis.

Before: The Decision to Post

The uncomfortable analysis looks like this: Maya makes posting decisions that are substantially shaped by her mental model of what will be liked. She chooses photos based on lighting and composition criteria she has developed by observing which of her photos have performed well. She times posts for 6 to 9 PM on weekdays — she knows, without having formally analyzed it, that posts at this time perform better for her audience. She writes captions that are brief, warm, and slightly self-aware, because she has observed that earnest or too-long captions on her kind of content underperform. She has never articulated any of this as a strategy. It is embedded in her choices.

There is a photo she took last week — a candid shot of her little sister laughing at the kitchen table, light coming in at a perfect angle, genuine and unposed. Maya loves the photo. She spent ten minutes looking at it on her phone, feeling something warm about it. But when she considered posting it, she felt a different thing: a calculating hesitation. Her audience responds best to photos of her, not photos of her sister. The lighting is good but the composition isn't the kind of thing that gets shared. The caption would be hard — what do you say about a photo like that without it seeming precious? She didn't post it. Instead she posted a mirror selfie with better light and a simpler caption. The mirror selfie got 47 likes. The photo of her sister exists only on her camera roll, unshared, a private record of a moment that was apparently not worth posting.

She does not fully register this pattern when she's inside it. She just feels like she's making natural choices about what to share. The pattern becomes visible only in retrospect, and even then it takes a deliberate kind of analysis she rarely applies to her own social media behavior.

The Two-Hour Window

When Maya posts, there is a two-hour window that operates differently from the rest of her day.

She posts the photo. Within the first minute, she refreshes the app twice to see if any likes have appeared. There's something about the first minute that feels critical — if someone sees it immediately and likes it, that's a signal that the content connected instantly. By five minutes in, she has a first-read on the trajectory: if she has four or five likes, the post is probably fine. If she has one or zero, something's wrong.

She puts her phone down. She picks it up again forty-five seconds later. She knows this is happening; she watches herself doing it with a kind of bemused awareness that doesn't produce change. She tries to distract herself. She texts Priya about something unrelated. She half-watches a TikTok. But the awareness of the post is a background presence — a kind of cognitive attention that doesn't fully release even when she's doing something else.

Around the forty-five-minute mark, the monitoring frequency usually decreases. If the post is performing well, she has proof; the anxiety resolves into something that looks like satisfaction. If it's performing poorly, she has data; the anxiety doesn't resolve but it flattens into a dull acknowledgment. By two hours in, the acute phase is over. The post has found its level.

The two hours she spends in this monitoring state are not comfortable hours. They are not hours she would choose to spend this way if she were fully conscious of the choice. They are hours in which a significant portion of her cognitive and emotional resources are allocated to processing the incoming social verdict on whether something she made was good enough. She does this for approximately thirty percent of her waking days, corresponding to approximately how often she posts.

When a Post "Fails"

Maya has a private threshold that she has never stated aloud but that shapes her experience of every post she makes: fifteen likes in the first hour is the soft floor. Below it, something has gone wrong. Above it, things are okay. Above thirty, things are good. Above sixty, something exceptional has happened.

When a post finishes below fifteen — and this happens — the mood effect is subtle but real. It doesn't announce itself as despair or distress. It arrives as a low-key coloring of the rest of the evening. She's slightly more irritable. She's quicker to find fault with the homework she's trying to do. She's more likely to end up in a negative spiral on TikTok, watching content that makes her feel worse. She doesn't connect these effects to the post. The connection is there, but it operates below the level where she typically applies the analysis.

What she does connect to the post is a particular flavor of self-doubt that attaches to the specific content that failed. If she posted a photo she thought was particularly good and it underperformed, the doubt isn't just about the photo — it's about her judgment. If she misjudged how good that photo was, what else is she misjudging? The approval verdict doesn't merely rate the content. It rates the part of her that thought the content was good enough to share.

Twice in the past year, she has deleted posts that underperformed within the first hour. She feels slightly ashamed of having done this — she knows it's not who she wants to be, doesn't want to be someone who curates her social media presence that anxiously. But the desire to remove the evidence of the failed post, to erase the record of the low number before more people see it, has twice been stronger than the competing desire to be authentically unbothered.

She has, also twice in the past year, decided not to post something she genuinely liked — a photo she found beautiful, a caption she thought was funny — because she couldn't shake the sense that it wouldn't do well. Both times, she felt the loss of not sharing. She had something she wanted to share, and she chose not to, because she was afraid of the approval verdict. This is the approval economy functioning at its most revealing: the user editing her authentic self-expression to conform to the projected preferences of an audience, to avoid the social pain of a low number.

The Creator Economy's Approval Trap

For users like Maya, the approval economy is a matter of self-esteem and social positioning. For users who earn income from their social media presence, the approval economy acquires an additional dimension: economic survival.

The creator economy — the ecosystem of individuals who produce content for platforms and earn income through advertising revenue share, brand partnerships, and direct fan support — has grown to include an estimated 50 million people globally, according to SignalFire's 2023 creator economy report. For professional creators, like counts, follower counts, and engagement rates are not merely social metrics; they are the data that determines their income. A drop in engagement is a drop in revenue. A viral post is a business windfall.

This economic dimension does not merely add a new layer of stress to the approval dynamics described in this chapter. It fundamentally changes the relationship between the creator and their audience, because the creator's authentic self-expression is now filtered through the economic calculus of engagement optimization. Content is chosen not because it reflects what the creator most wants to share, but because it is most likely to generate the engagement metrics that produce income. The creator becomes, in a very real sense, an employee of the algorithm — producing the content the algorithm rewards, in the formats the algorithm amplifies, on the schedules the algorithm prefers.

We will examine the creator economy and its approval trap in detail in Chapter 34. The point to establish here is that the social reward system described in this chapter — the neurological machinery of approval-seeking activated by the like button — is not a recreational dynamic for professional creators. It is their economic reality, and the platforms' capture of that reality through the approval economy represents one of the most sophisticated labor extraction mechanisms in the history of digital capitalism.

Velocity Media Sidebar: The Real-Time Like Count Debate

In one of the more philosophically charged internal debates at Velocity Media, Marcus Webb and Dr. Aisha Johnson found themselves on opposite sides of a design question with significant consequences: should the platform show creators their like counts in real time, or should there be a delay?

The proposal on the table was Dr. Johnson's. She had been reading the research on variable reward and its role in compulsive checking behavior, and she had connected it to internal data showing that Velocity's creators checked their post analytics an average of eleven times in the two hours after posting. Her proposal was modest: instead of updating like counts in real-time as each like arrived, delay the display by thirty minutes and show creators a batched total. The content would still receive the likes as they arrived; the creator would simply see the data on a delay.

"We're not hiding the information," she said at the product review. "We're changing the cadence at which it's delivered. The research on variable reward is clear that what makes checking compulsive is the unpredictability of the update — you never know if there's been a new like since you last checked, so you check constantly. A thirty-minute batch would eliminate the unpredictability while preserving all the information."

Marcus Webb's objections were layered. The first was competitive: "If we delay like counts and our competitors don't, creators will choose competitors. Real-time data is a feature, not a bug. Creators need to know when their content is performing well so they can engage with comments at peak momentum."

The second objection was more interesting: "We'd be making a paternalistic decision about how creators should receive information about their own content. If a creator wants to check their analytics every minute, that's their business. Our job is to give them the best tools, not to manage their psychological relationship with those tools."

Dr. Johnson's response to the paternalism objection drew on behavioral economics. "We already make this choice. Every choice about what information to display, at what frequency, in what format is a choice that shapes how creators relate to their analytics. There is no neutral default. The current real-time update was chosen — explicitly or by convention — without any research on its behavioral effects. I'm proposing we choose differently, based on evidence about what that choice does to people."

Sarah Chen, in a characteristic compromise, proposed a test. The thirty-minute delay would be offered as an option in the creator dashboard settings — not a default change, but an available alternative. Creators who found their analytics checking compulsive could opt into the delayed view.

The test ran for three months. Adoption was 3.7% — very few creators voluntarily chose the delay. But among those who did, creator satisfaction scores were notably higher, and the creators who adopted the delay showed lower rates of what the platform called "reactive behavior" — deleting posts within two hours, significantly altering their posting strategy after a single underperforming post, or spending more than thirty minutes per day in the analytics dashboard.

The debate remained unresolved in terms of platform default. Real-time counts remained the standard. But the test data gave Dr. Johnson a foothold: a small population of creators who chose less compulsive feedback monitoring reported better experiences. The question of whether to extend that finding to all creators — through default change rather than voluntary adoption — was deferred to a future product cycle that had not yet arrived.

Conclusion: Digitized, Quantified, and Weaponized

The desire for social approval is not a product of the smartphone. It is among the most ancient and durable features of the human nervous system, shaped over millions of years of evolution in which social standing was directly correlated with survival. The brain's reward circuitry responds to social acceptance with the same signals it uses for food and warmth. Social rejection activates the same circuits as physical pain — a neurological fact demonstrated by Eisenberger's fMRI research and replicated across multiple laboratories. Lieberman's social brain hypothesis adds that this social processing is not a special-purpose function but the default cognitive activity of the human brain — the thing the mind does when it has nothing else to do.

What the like button did — and what the broader social reward architecture of modern platforms has done — is take this ancient, profound, and neurologically deep human need and build a delivery system for it that is:

• Quantified: Making social approval into a number that can be compared, tracked, and optimized.
• Public: Making that number visible not just to the creator but to anyone who sees the post, creating permanent public social verdicts.
• Variable: Distributing approval unpredictably, creating the dopaminergic urgency of an uncertain reward.
• Persistent: Making the approval record permanent, accessible, and scrollable — a continuous social ledger of your standing.
• Scaled: Extending the potential audience from dozens to thousands or millions, making the approval machine dramatically more potent than any previous social environment.
• Algorithmic: Distributing approval not by democratic social interest but by algorithmic amplification, rewarding certain types of content and certain types of presentation over others — training creators, consciously or not, to produce what the algorithm rewards.

The adolescent brain's hyperresponsive social reward circuitry makes young users like Maya not just typical consumers of this system but its most neurologically vulnerable. At the developmental moment when social approval is maximally salient — when the ventral striatum fires most strongly in response to peer approval, when the medial prefrontal cortex is most engaged by questions of how others perceive us — these same users encounter a platform architecture that delivers quantified social approval signals with laboratory precision, variable enough to sustain compulsive checking behavior, and persistent enough to maintain the social ledger across years.

Platforms didn't invent social approval needs. They gave those needs a new shape — a number, a heart, a count — and then they engineered the experience of receiving that number to be as compelling as possible. The like button is not a neutral tool for expressing appreciation. It is a neurological lever, pulled billions of times a day, connected to the most powerful reward circuitry the human nervous system possesses.

Chapter 11 examines what happens when the approval and engagement systems described in Chapters 9 and 10 combine with a specific emotional mechanism: the fear of missing out. FOMO is not merely a casual anxiety about social media. It is a specific neurological and psychological response that platforms have learned to cultivate systematically. How that cultivation works, and what it costs, is the subject of the next chapter.

Next: Chapter 11 — Fear of Missing Out: The Engineered Anxiety

References and Chapter Notes

On social pain and rejection neuroscience: Eisenberger, N.I., Lieberman, M.D., & Williams, K.D. (2003). Does rejection hurt? An fMRI study of social exclusion. Science, 302(5643), 290-292.

On the neural regulation of social pain: Eisenberger, N.I., Taylor, S.E., Gable, S.L., Hilmert, C.J., & Lieberman, M.D. (2007). Neural pathways link social support to attenuated neuroendocrine stress responses. NeuroImage, 35(4), 1601-1612.

On romantic rejection and physical pain neural overlap: Kross, E., Berman, M.G., Mischel, W., Smith, E.E., & Wager, T.D. (2011). Social rejection shares somatosensory representations with physical pain. PNAS, 108(15), 6270-6275.

On social reward neural correlates: Lieberman, M.D. (2013). Social: Why Our Brains Are Wired to Connect. Crown Publishers.

On adolescent social reward sensitivity: Telzer, E.H., et al. (2016). Social influence on positive youth development: A longitudinal study of neural response to social rewards and risk taking. Journal of Neuroscience.

On adolescent brain development and social cognition: Blakemore, S.J. (2012). Imaging brain development: the adolescent brain. NeuroImage, 61(2), 397-406.

On Instagram-specific adolescent social reward: Sherman, L.E., Payton, A.A., Hernandez, L.M., Greenfield, P.M., & Dapretto, M. (2016). The power of the like in adolescence. Psychological Science, 27(7), 1027-1035.

On Festinger's social comparison theory: Festinger, L. (1954). A theory of social comparison processes. Human Relations, 7(2), 117-140.

On quantification and hedonic psychology: Hsee, C.K., & Zhang, J. (2010). General evaluability theory. Perspectives on Psychological Science, 5(4), 343-355.

On the like button's origins: Lewis, P. (September 2017). "Our minds can be hijacked: the tech insiders who fear a smartphone dystopia." The Guardian. Pearlman quotation from same source and subsequent interviews.

On Instagram, social comparison, and adolescent wellbeing: Orben, A., et al. (2020). The association between adolescent well-being and digital technology use. Nature Human Behaviour; Wells, G., Horwitz, J., & Seetharaman, D. (September 14, 2021). "Facebook Knows Instagram Is Toxic for Teen Girls, Company Documents Show." The Wall Street Journal.

On the Instagram like-hiding experiment: Tran, L.T.N., et al. (2021). How does hiding the "Like" count affect Instagram users? Social Media + Society.

On social reward fMRI responses: Meshi, D., Morawetz, C., & Heekeren, H.R. (2013). Nucleus accumbens response to gains in reputation for the self relative to gains for others predicts social media use. Frontiers in Human Neuroscience.

On what makes content shareable: Berger, J., & Milkman, K.L. (2012). What makes online content viral? Journal of Marketing Research, 49(2), 192-205.

On narcissism and social media: Konrath, S., et al. (2014). Changes in dispositional empathy in American college students over time: A meta-analysis. Personality and Social Psychology Review.

On attachment style and social media: Oldmeadow, J.A., Quinn, S., & Kowert, R. (2013). Attachment style, social skills, and Facebook use amongst adults. Computers in Human Behavior, 29(3), 1142-1149.

On approval addiction and social media: Andreassen, C.S., Pallesen, S., & Griffiths, M.D. (2017). The relationship between addictive use of social media, narcissism, and self-esteem: Findings from a large national survey. Addictive Behaviors, 64, 287-293.

On the creator economy: SignalFire (2023). Creator Economy Market Map. San Francisco, CA.

On social media and adolescent depression: Twenge, J.M., et al. (2018). Increases in depressive symptoms, suicide-related outcomes, and suicide rates among U.S. adolescents after 2010 and links to increased new media screen time. Clinical Psychological Science.