Case Study 2: The Like Button — Intent, Effect, and Regret

Background

Few design decisions in the history of digital technology have had greater behavioral consequences than the creation of the Facebook Like button. In the received popular account, the Like button was invented at a hackathon, launched in 2009, and quickly became one of the most recognized interface elements in internet history — a thumb-up icon so ubiquitous that it migrated far beyond its original platform to become a metaphor for social approval itself.

The fuller story is more complicated, more instructive, and more ethically freighted. The Like button was not simply invented and deployed. It was conceived with specific social intentions, deployed with commercial intentions that diverged from those social intentions, and has since become the subject of genuine moral reckoning by some of its principal creators. Understanding this history illuminates not only the specific dopaminergic mechanism of social validation rewards but also the broader gap between intent and effect that characterizes so much of social media's psychological impact.

The Invention: Justin Rosenstein and the "Awesome Button"

Justin Rosenstein joined Facebook in 2007, having previously worked at Google. He was, by all accounts, a skilled engineer with a genuine commitment to building technology that improved human connection. In 2007, in one of Facebook's internal hackathons — intensive periods of unstructured building time that the company used to generate new features — Rosenstein and a small team built a feature they called the Awesome Button.

The original concept was simple and positive in intent. Facebook posts generated comments, but commenting required effort — writing words, navigating a text box. Rosenstein's observation was that many people wanted to express simple appreciation or acknowledgment for a post but lacked an easy mechanism to do so. The Awesome Button would solve this friction: a single click to say, in effect, "I see this, and I appreciate it."

The social vision was benign. Rosenstein has described the feature as intended to "spread positivity and love" across the platform. The friction of commenting, the reasoning went, was suppressing genuine positive social expression. By removing that friction, Facebook could enable people to be more affirming of one another with less effort.

In 2009, after internal deliberation and rebranding (the Awesome Button became the Like button, with a thumbs-up icon), the feature launched publicly. It was an immediate and massive success. Users clicked Like billions of times within days. Advertisers quickly recognized its utility — likes on branded content provided social proof. The feature was exported to other platforms via Facebook's social plugins, eventually appearing across much of the web. The thumbs-up icon became, within a few years, one of the most instantly recognizable symbols in global visual culture.

The behavioral effects of the Like button diverged from Rosenstein's original intentions in several important ways.

Asymmetric Anxiety

The most significant unintended consequence was asymmetric: while the Like button made it easy to give a simple positive signal, it also created new forms of social anxiety for post creators. Before the Like button, a Facebook post existed in a relatively ambiguous social space — people saw it or didn't, responded or didn't, but the absence of visible metrics left the social stakes implicit. The Like button transformed posts into explicitly evaluated content. Every post now had a public score. Absence of likes was visible absence.

This introduced a new form of social comparison and validation-seeking behavior. Users began checking their posts for likes in the minutes and hours after posting — a perfect behavioral loop: post, wait, check, receive variable reward (likes), check again. The anxiety of not checking — the anticipation of unknown social evaluation — was precisely the dopaminergic dynamic described in Chapter 7. The variable ratio schedule operated because likes came in unpredictably over time, and the checking loop was reinforced by intermittent reward.

From Expression to Performance

A second divergence from original intent was the gradual transformation of posting behavior from authentic expression to optimized performance. Once likes became a visible metric, users began, consciously or unconsciously, optimizing their content for likes. Posts that generated strong like responses were repeated; posts that generated weak responses were not. This is operant conditioning: like-generating behavior was reinforced and therefore increased.

The effects were particularly pronounced for younger users, whose identity formation was still in progress. For a 16-year-old posting on social media, the like count on a post is not just a social metric — it is a quantified measure of social acceptance, arriving in the period when social acceptance is developmentally most salient. Research consistently shows that adolescents are more strongly influenced by social reward signals and more sensitive to social exclusion than adults. Placing a public, quantified, real-time social acceptance metric in front of adolescent users produces predictable psychological effects that diverge substantially from "spreading positivity and love."

The Dislike Problem and Emotional Complexity

Rosenstein's original Awesome Button was positive-only by design. The final Like button inherited this property. There was no dislike button, no mechanism for expressing the full range of social responses. This design choice, while defensible in terms of reducing conflict, also meant that the social feedback system was systematically incomplete. Users could express approval with zero friction; expressing anything else — disagreement, concern, sadness — required the higher-friction response of commenting or, eventually, using the more ambiguous emoji reactions.

The result was a social environment optimized for positive validation seeking, which is not the same as a social environment that facilitates authentic connection. The Like button created incentives to post content that generates likes — which tends toward content that is entertaining, aesthetically pleasing, or emotionally resonant in simple positive ways — rather than content that is honest, vulnerable, or complex.

The Regret: Chamath Palihapitiya and the Reckoning

The moral dimension of the Like button's history is sharpened by the subsequent public statements of Chamath Palihapitiya, who served as Facebook's Vice President of User Growth from 2007 to 2011 — the period that included the Like button's launch and explosive growth. Palihapitiya is one of the most prominent former Facebook executives to have spoken critically about the company's practices.

At a Stanford Graduate School of Business event in November 2017, Palihapitiya said: "I feel tremendous guilt. I think we all knew in the back of our minds — even though we feigned this whole line of, 'There probably aren't any bad unintended consequences' — I think in the back, deep, deep recesses of our minds, we knew something bad could happen."

He went on to describe the short-term dopamine-driven feedback loops that Facebook had built, including the like and hearts systems, as undermining the social fabric. Notably, he said he did not allow his own children to use social media — a statement that has been widely cited as a particularly vivid indicator of the gap between the confidence with which the technology was promoted to users and the private assessments of the executives who built it.

Palihapitiya's statement was followed by similar, if somewhat more qualified, public reflections from other former Facebook figures. Sean Parker, in his 2017 Axios interview, described the like button explicitly as a "dopamine hit" and characterized the platform's development as a conscious project of exploiting psychological vulnerabilities.

Justin Rosenstein himself has also expressed regret. In interviews following the release of the documentary "The Social Dilemma" (2020), in which he appears, Rosenstein described the like button as one example of the broader unintended consequences of technology design, and has become an advocate for more thoughtful, humane technology design practices.

Analysis Using Chapter Concepts

The Like Button as Variable Ratio Reward Delivery

The Like button's behavioral power derives entirely from its variability. If posts reliably received the same number of likes — or if likes arrived on a predictable schedule — the checking loop would not have the same compulsive quality. The power of the system comes from not knowing: how many likes? From whom? By when? Each check is a pull-to-refresh, a slot machine lever. Sometimes the reward is greater than expected; sometimes it is less; sometimes the notification says "47 people liked your photo" and that number activates the nucleus accumbens in exactly the way Lauren Sherman's research documented.

The Like button is, in this analysis, a variable ratio reward delivery system for social validation, built directly into the social content sharing interface. Its brilliance, from an engagement design perspective, is that it required no additional engineering to produce the variable ratio effect. Social behavior is inherently variable — you cannot predict how many people will like your post any more than you can predict when the slot machine will pay out. The platform simply had to make that variability visible, quantified, and easily checked.

The Intent/Effect Gap

The history of the Like button is a vivid illustration of what this book calls the gap between intent and effect. Rosenstein's intention — enabling easy positive social expression — was genuine and benign. The effect — creating quantified social metrics that drove compulsive validation-seeking, social anxiety, identity disturbance in adolescents, and the optimization of authentic expression toward engagement performance — was neither intended nor, in many cases, foreseen.

This gap has important implications for how we evaluate the moral responsibility of designers and platforms. It suggests that good intentions are insufficient, and that the behavioral consequences of design decisions require the same systematic evaluation as any other engineering decision with significant safety implications. It also suggests that the moral analysis of platform harms cannot be resolved by asking whether designers meant to harm users. The relevant question is what they knew, what they should have known, and what they did when the effects became clear.

What Changed When the Effect Became Clear

A crucial element of the moral evaluation of the Like button is the timeline of knowledge. By 2012–2013, research on the effects of social media on adolescent mental health was beginning to accumulate. By 2015–2017, there was substantial public discussion and emerging evidence linking social media use to anxiety, depression, and social comparison effects. The Like button — with its quantified, public, variable social feedback — was repeatedly identified as a key mechanism.

Facebook's response during this period included conducting internal research on the platform's effects on well-being (some of which later became public through leaked documents), engaging with external researchers under terms that limited the external researchers' ability to publish negative findings, and making some design changes (including the 2019 rollout of hidden like counts in some markets) while resisting more fundamental changes that might affect engagement.

The decision to hide like counts in some markets in 2019 — making the count visible only to the post creator, not to all viewers — is itself significant. This change directly addressed one of the most documented harms (the visibility of social comparison metrics to adolescents and others) while preserving the dopaminergic reward structure for the creator (who still received the notification and could see their own count). It is a partial measure that illustrates the difficulty of redesigning platforms that are commercially dependent on engagement metrics that are themselves driven by the dopamine loops these features create.

What This Means for Users

The Like button case illuminates several dimensions of the power asymmetry between platforms and users.

First, users interacting with the Like button and the validation-seeking behavior it generates are responding to design decisions they did not make, were not informed about, and could not meaningfully consent to. The behavioral effects documented by Sherman and others — the nucleus accumbens activation, the social comparison pressure, the compulsive checking — were not disclosed to users as properties of the system they were adopting.

Second, the gap between the intentions of designers and the effects of their designs illustrates a structural feature of the attention economy: the optimization pressures that shape platform design consistently favor engagement over well-being, regardless of individual designers' values. Rosenstein's good intentions did not prevent the Like button from becoming a social anxiety engine. This is not because Rosenstein was naive or negligent; it is because the design decision was evaluated and deployed primarily in terms of engagement effects rather than well-being effects, and those two sets of effects point in different directions.

Third, the subsequent expressions of regret by Palihapitiya, Rosenstein, Parker, and others — while genuine and valuable — are insufficient as a social response to the harms documented. Individual moral reflection by engineers and executives, however sincere, does not change the structural incentives that produced the design decisions in the first place. Addressing those structural incentives requires changes at the level of business model, regulatory framework, and platform governance — not simply better individual decision-making.

Discussion Questions

Justin Rosenstein's original intent — enabling easy positive social expression — was genuinely benign. Does this intent mitigate his moral responsibility for the Like button's documented harmful effects? What standard should we apply when evaluating the moral responsibility of designers for the unintended behavioral consequences of their designs?
Chamath Palihapitiya's statement that "we knew something bad could happen" suggests that at least some knowledge of potential harms was present at Facebook during the period when the Like button was being built and deployed. What obligations does partial foreknowledge of harm create? How does this change the moral evaluation compared to a situation of genuine ignorance?
Facebook's 2019 decision to hide like counts for post viewers (while preserving them for post creators) has been described both as a meaningful harm-reduction measure and as an insufficient half-measure. Evaluate this design change using the chapter's framework. What harm does it address? What harm does it leave in place? Is it sufficient?
The Like button created a quantified, public, real-time social approval metric in an environment where social approval is developmentally critical (particularly for adolescents). Should platforms that serve adolescents be held to different design standards than those serving only adults? What would those standards look like?
If you were advising a technology company on how to prevent the kind of unintended-consequence story described in this case study, what process changes would you recommend? What institutional structures, review mechanisms, or design principles might have led to a different outcome?