The Social Dilemma: Tristan Harris Explains How Social Media Manipulates Attention, Shapes Elections, and Changes Human Behavior

By Elle Trump and Lehan Edirisinghe · October 25, 2026 · AngryPages

Former Google design ethicist Tristan Harris joins Joe Rogan to explain how recommendation algorithms became a powerful influence on human psychology, why Facebook's expansion produced serious consequences abroad, and what could replace the business model behind the attention economy.

In 2012, YouTube made a decision that would help define the next generation of social media: instead of primarily rewarding videos that attracted clicks, its recommendation system would increasingly reward videos that kept people watching.

The change sounded reasonable. A video that holds someone's attention might be more useful than one with an attractive thumbnail that viewers immediately abandon.

But it introduced a more complicated question: What happens when a system becomes extremely good at predicting what people will watch, without being equally good at judging whether watching it is beneficial?

For Tristan Harris, a former Google design ethicist and co-founder of the Center for Humane Technology, that question connects seemingly unrelated problems: technology addiction, declining attention, conspiracy theories, political polarization, children's mental health, the weakening of professional journalism, and the manipulation of elections.

During his 2020 conversation with Joe Rogan, Harris explained how a handful of design decisions evolved into a global competition for human attention. He also discussed failed efforts to change Google internally, serious failures in Facebook's international expansion, and alternatives that could make technology serve people rather than merely retain them.

Many of his concerns have since been examined through independent research, regulatory investigations, and court proceedings. Some findings reinforce his warnings. Others challenge the idea that recommendation algorithms inevitably radicalize users. Together, they provide a clearer account of what the attention economy does, what remains uncertain, and why changing it is difficult.

1. YouTube's Decision to Optimize for Watch Time

The 2012 change that reshaped recommendations

In March 2012, YouTube announced that viewing duration would become a more important signal in its recommendation system.

Previously, recommendations relied heavily on clicks and views. This allowed sensational titles and thumbnails to attract viewers without necessarily providing satisfying content.

YouTube's new approach rewarded videos that people watched for longer periods. Its August 2012 explanation confirmed that recommendations considered not just the next video watched, but the viewing sessions that followed.

The company argued that sustained viewing indicated a more satisfying experience. It also acknowledged that keeping viewers on YouTube created additional revenue opportunities for content partners.

Harris saw a potential conflict in that reasoning. A video can hold attention because it is informative, beautiful, frightening, outrageous, or deeply misleading. Viewing duration alone cannot distinguish among those possibilities.

By January 2018, YouTube's chief product officer, Neal Mohan, said recommendations accounted for more than 75% of viewing time, according to contemporaneous reporting from CNET.

That figure helps explain the scale of the issue. Viewers were not simply choosing videos from a neutral library. Automated recommendations were responsible for much of what they encountered next.

How an algorithm learns to hold someone's attention

Harris described recommendation systems as continually updated models of human behavior.

Every interaction supplies information. Clicks reveal initial interest. Viewing duration shows how long content held attention. Likes, comments, searches, and repeated visits provide further signals.

The system can then compare those behaviors with patterns from millions of other people. If users with similar histories watched a particular video for two hours, the algorithm can recommend that video without understanding what made it appealing.

Harris described the imbalance between personal self-control and machine prediction as:

"a supercomputer pointed at your brain"

The expression was deliberately dramatic, but it captured his central concern: individuals trying to make conscious decisions are competing against systems designed to learn from their behavior continuously.

A person may open Facebook to check a single invitation and emerge an hour later after consuming unrelated posts. The extra viewing was not necessarily planned, yet it creates the engagement that the platform measures.

Harris compared recommendation systems to observing motorists slowing down to look at a highway accident and concluding that people should be shown car crashes continuously.

Attention, in this argument, is not the same as considered preference.

From ordinary interests to conspiracy theories

Harris identified several examples of recommendations producing troubling sequences.

Teenage girls watching dieting videos were reportedly directed toward anorexia-related material. People researching World War II could encounter Holocaust denial. A mother interested in homemade baby food might receive recommendations for anti-vaccination groups.

Researcher Renée DiResta's experience with Facebook groups illustrated another possibility: joining a community about organic baby food could expose someone to anti-vaccination conspiracy theories, which could lead toward other communities discussing chemtrails or flat-Earth beliefs.

Harris claimed that YouTube had recommended flat-Earth videos hundreds of millions of times.

The important mechanism was not necessarily a deliberate corporate decision to promote those ideas. It was that recommendation systems could discover audiences for material and repeatedly distribute it because the material generated engagement.

Former YouTube recommendation engineer Guillaume Chaslot became an influential critic of these patterns. His work helped draw attention to long recommendation chains that could retain viewers for hours.

Harris also distinguished irrational conspiracy thinking from legitimate suspicion of powerful institutions. Historical misconduct, including COINTELPRO, demonstrates that genuine conspiracies exist. His objection was to systems promoting claims on the strength of behavioral engagement rather than reliable evidence.

The vulnerability differs among users. An adult may watch flat-Earth videos for amusement; a child might accept the same claims as factual. A recommendation model trained primarily on engagement may not distinguish the two.

What subsequent research found

The relationship between recommendations and political radicalization has become more complicated than the interview suggested.

In 2019, YouTube began changing its recommendation system to reduce the distribution of borderline content and harmful misinformation. The company later reported a 70% reduction in US viewing time for such content originating from recommendations to non-subscribers. That is a company-reported measure, not an independent assessment of all harmful viewing. YouTube's explanation.

Independent studies have produced mixed findings.

A 2023 study published in Proceedings of the National Academy of Sciences used 100,000 automated accounts to audit YouTube recommendations. It found that recommendations frequently reflected users' ideological preferences. Certain users encountered more material from problematic channels deeper in recommendation sequences, although the study did not find large average increases in ideological extremity.

A separate 2025 experimental study involving nearly 9,000 participants found that manipulating video recommendations had limited short-term effects on political attitudes. Researchers cautioned that this did not rule out long-term effects or stronger effects on particularly vulnerable populations.

A 2026 study in Political Communication similarly found evidence of recommendation rabbit holes but not an automatic progression toward increasingly extreme political content. User preferences accounted for much of the observed ideological separation.

These findings distinguish three questions that are often treated as one: whether algorithms affect what people watch, whether they expose users to problematic material, and whether that exposure changes political beliefs.

The evidence strongly supports the first. It identifies circumstances where the second occurs. The third remains more contested.

Harris's broader concern survives that distinction: a system can shape the information people encounter even when its precise effects on their beliefs are difficult to measure.

2. How Social Media Became a Competition for Human Attention

Twitter's transformation from messaging tool to global platform

Twitter began with a relatively simple idea.

Jack Dorsey was interested partly in taxi dispatch systems, where a message could be broadcast to multiple drivers. Twitter applied a similar concept to communication among ordinary people.

Its founders did not anticipate that the service would eventually become a major venue for political debate, government communication, journalism, and disputes over public speech.

Rogan and Harris discussed Dorsey's congressional questioning following Twitter's restriction of reporting about Hunter Biden, allegations of conservative censorship, and the continued use of social networks by authoritarian political figures.

The central question was how a product originally designed for communication had become an institution making decisions about what enormous audiences could see.

Harris traced part of that transformation to seemingly small features.

A visible follower count gave users a reason to return and monitor their popularity. The retweet button made redistribution nearly effortless. Together, they created opportunities for ordinary users to reach enormous audiences.

Harris called this the fame lottery.

In 2009, Ashton Kutcher's race with CNN to reach one million Twitter followers was a major event. Later platforms, particularly TikTok, made rapid audience growth possible for people without existing celebrity status.

Venture capital reinforced the competition. Investors favored social products capable of reaching tens or hundreds of millions of users increasingly quickly.

This changed not only platform design but also human incentives. A person who believes an outrageous post might reach a million people has a reason to produce more provocative material.

The ability to broadcast globally had become widely accessible without requiring the editorial obligations traditionally associated with publishing.

Why Facebook considered YouTube a competitor

Harris recalled a conversation with a Facebook employee who explained that Facebook's biggest competitive threat was not necessarily another social network.

It was YouTube.

The employee's reasoning was straightforward: both companies wanted the same limited resource, the user's attention.

A person watching YouTube cannot simultaneously devote the same attention to Facebook. Netflix, TikTok, television, games, and even email all compete for pieces of that resource.

Harris summarized the commercial logic:

"the thing that is the product is your predictable behavior"

Advertising platforms do not merely possess information about people. Their commercial value depends partly on using that information to predict and influence future actions.

The result is a market in which companies are rewarded for discovering increasingly effective methods of engagement.

The competitive pressure behind addictive features

Consider two possible versions of Twitter.

One presents posts chronologically, showing recent messages from followed accounts. The other ranks posts according to predicted interest and engagement.

If the second produces more viewing, advertising impressions, and user retention, competitors face pressure to adopt similar systems.

Harris argued that this produces a competitive trap. Individual companies may understand that certain features are unhealthy yet still implement them because rivals are already benefiting.

Netflix's automatic countdown to the next episode illustrated the pattern. If autoplay increases total viewing, YouTube has an incentive to introduce similar mechanisms.

TikTok went further by beginning video playback immediately when the application opened, eliminating even the initial decision to select something.

In Harris's interpretation, these developments reduced the distance between opening a service and entering a continuous stream of consumption.

His description was uncompromising:

"the business model is addiction"

That is his characterization, not proof that every social platform or user experience is clinically addictive. It reflects the concern that corporate incentives systematically reward repeated, prolonged engagement.

Breaking up an individual company would not necessarily resolve those incentives. New competitors operating under the same advertising model could reproduce the same behavior.

How ranking systems change what creators produce

Rogan and Harris examined Apple's podcast rankings as an example of how distribution rules shape creative activity.

A podcast receiving 10,000 downloads could outrank an established show receiving 100,000 because its audience was growing faster.

Rogan considered this useful for discovering newcomers who might otherwise remain invisible beneath permanently dominant programs.

Harris's own podcast, Your Undivided Attention, benefited from that structure after The Social Dilemma appeared. It reportedly climbed into the top five and briefly reached the leading position in Apple's technology category.

But rankings also created opportunities for manipulation.

Rogan described comedian Earl Skakel allegedly paying people to game Apple's podcast rankings despite attracting a relatively small authentic audience. Rogan believed the episode resulted in restrictions, although the precise enforcement details were not established during the interview.

Harris also recounted an explicitly unverified story in which Steve Jobs questioned why a podcast application should automatically show everything someone's friends were listening to. Jobs supposedly believed genuinely worthwhile material would be shared personally.

The broader principle matters regardless of that anecdote's accuracy.

A ranking system that rewards rapid growth encourages creators to pursue rapid growth. One that rewards outrage would give creators stronger incentives to become outrageous.

The distributor is not a neutral observer of the resulting content.

The economics of journalism

Social platforms also altered the relationship between publishing and editorial responsibility.

Professional journalism requires people to investigate claims, interview witnesses, verify evidence, edit material, and accept responsibility for errors.

Harris used illustrative annual salaries of $70,000 for a journalist and $100,000 for an editor to explain those costs.

A social media platform can distribute enormous amounts of user-generated content without directly paying its creators or employing editors to assess every post.

Its algorithms instead select material using signals such as popularity and engagement.

That model is cheaper to operate at scale, but popularity does not guarantee credibility.

Harris argued that traditional media increasingly imported social media's priorities. A handful of hostile tweets could become a story about widespread public outrage. Television and radio coverage could follow trends created by algorithmic distribution.

The consequence was an information market in which journalism competed not only with other journalism, but with an almost unlimited supply of engaging content produced at much lower cost.

Harris rejected the assumption that publishing more speech would automatically correct misinformation. A false claim might reach millions while its correction struggled to attract a fraction of that audience.

The challenge was no longer simply whether accurate information existed. It was whether accurate information could obtain comparable attention.

3. Tristan Harris's Attempt to Change Google From Within

A design ethicist questions the technology industry

Harris did not begin as an outside critic of Silicon Valley.

His startup had been acquired by Google in a talent acquisition, bringing him into a company whose products were becoming central to how people searched, communicated, and interpreted information.

His background was unusual for a technology employee. Childhood magic had introduced him to the mechanics of directing attention. Later, he studied persuasive technology at Stanford, where Instagram's co-founders were among those who attended related classes.

He also examined behavioral economics and techniques associated with cults, including methods of controlling information, relationships, and social belonging.

These interests shaped a different understanding of software.

Traditional computer programs helped people perform tasks. Modern platforms could also become environments that continuously influenced users' desires, emotions, and choices.

In 2013, a hiking trip in California's Santa Cruz Mountains with Aza Raskin sharpened Harris's concerns about technology distracting people from important parts of life.

Raskin subsequently became his co-founder at the Center for Humane Technology. His father, Jef Raskin, had initiated Apple's Macintosh project and helped shape ideas about more humane computer interfaces.

The presentation that reached 10,000 Google employees

Approximately a year into his time at Google, Harris created an internal presentation warning that a relatively small number of technology designers were exercising extraordinary influence over human psychology.

He argued that every notification, feed, and recommendation could influence what billions of people thought about and believed.

The presentation reportedly reached approximately 10,000 Google employees and became a major internal discussion topic.

Many employees privately expressed support.

But Harris said the issue did not receive the meaningful institutional response he wanted. Google's regular company meetings did not produce the changes he considered necessary.

He continued advocating internally for approximately two years before leaving.

At that stage, his main concerns involved distraction, addiction, weakened relationships, and the growing sophistication of persuasive design.

He acknowledged that he had not fully anticipated the subsequent scale of conspiracy amplification, political polarization, and manipulation of democratic systems.

Harris viewed Google's Chrome browser and Android operating system as particularly important opportunities for reform. Both stood between users and the wider digital environment, giving Google considerable influence over applications, notifications, and access to services.

His criticism was directed at product design and corporate incentives rather than employees' choice of devices. He even used an iPhone while working at Google, which he said was generally permitted.

The problem with defining what is good

The discussion returned to Google's original corporate motto, Don't be evil.

Rogan questioned the company's 2018 decision to remove most references to the phrase from its code of conduct.

Google had retained the expression at the document's end, while parent company Alphabet had adopted Do the right thing after its 2015 reorganization.

Harris acknowledged that positive language might sound preferable. But it also introduced a difficult problem.

In a society with competing moral beliefs, who gets to define the correct outcome?

He preferred the narrower objective of reducing identifiable harms:

"it's much easier to say let's reduce harms"

For Harris, the ethical problem was not simply that corporations sometimes made the wrong decisions. It was that increasingly powerful companies had placed themselves in positions where they were required to make decisions affecting enormous populations.

4. Facebook, Myanmar, and the Consequences of Global Expansion

Free Basics and the unequal cost of internet access

One of the interview's most consequential subjects was Facebook's expansion into developing markets.

Through its Free Basics initiative, Facebook partnered with telecommunications providers to make selected online services available without ordinary mobile data charges.

For people facing expensive internet access, the proposition was attractive. But it also created an imbalance: Facebook and participating services could be free while other parts of the internet required payment.

Harris argued that this could make Facebook the dominant source of news, communication, and information for entire populations.

He estimated that Free Basics had helped bring approximately 700 million accounts onto Facebook.

That figure requires qualification. Published research does not establish 700 million Free Basics users.

A 2020 academic study of Free Basics in Africa, drawing on contemporary company statements, reported that Facebook's Internet.org initiative had helped bring approximately 40 million people online by late 2016 and almost 100 million by 2018.

Those figures are not directly interchangeable with total Facebook accounts, but they show why Harris's 700 million estimate should not be treated as verified Free Basics adoption.

The underlying concern remains important: when access to one platform costs less than access to competing services, the platform can acquire unusual influence over a society's information environment.

Myanmar: A platform becomes part of an ethnic conflict

Myanmar provided Harris's most disturbing example.

Facebook expanded rapidly in a country affected by longstanding ethnic, religious, and political tensions. The platform became an important communication channel before the company developed sufficient capacity to identify and respond to locally specific threats.

Harris described coordinated misinformation, fake accounts, and anti-Rohingya propaganda, including content associated with military and government-linked actors.

The consequences were not confined to abusive comments or unpleasant online arguments. Propaganda contributed to an environment in which Rohingya Muslims faced persecution, displacement, and mass violence.

Harris recalled that Facebook had once relied on only around four or five people in its extended network who spoke the relevant local language. He also described large networks of fake accounts being removed, although his recollection of their precise number was uncertain.

Independent evidence substantially supports the concern about inadequate safeguards.

In 2018, a United Nations fact-finding mission documented grave violations against the Rohingya and examined the role of hate speech and Facebook in the wider crisis. The mission's detailed findings focused primarily on crimes and responsibility attributable to Myanmar's security forces, particularly the military.

Facebook also commissioned Business for Social Responsibility to investigate its human-rights impact in Myanmar.

When publishing the assessment in November 2018, Facebook acknowledged that it had not previously done enough to prevent its platform from being used to encourage division and offline violence.

The company reported hiring and onboarding 99 Myanmar-language reviewers, increasing its enforcement capacity, and developing policies addressing misinformation capable of contributing to imminent violence.

It also reported taking action against approximately 64,000 pieces of Myanmar-related content for hate-speech violations during the third quarter of 2018.

These measures demonstrated that Facebook could respond. They also raised a difficult question about why sufficient safeguards had not been established earlier.

The company's own assessment emphasized that Facebook alone could not resolve Myanmar's deep political and ethnic problems.

That qualification matters. The violence was perpetrated by people and institutions, not software. But Facebook's distribution systems could help perpetrators spread narratives that encouraged hostility and made violence easier to justify.

Harris argued that a company entering a society, becoming central to its communication, and contributing to institutional disruption had responsibilities beyond merely operating its service.

Ethiopia, India, and Brazil

Myanmar was not an isolated concern.

Harris described Ethiopia's linguistic diversity, referring to roughly 90 dialects and six major languages. Moderating such an environment required far more than translating English-language rules.

Groups communicating in dominant languages could influence national narratives about minority communities whose languages received less attention from moderators.

In India, the speakers discussed false WhatsApp rumors involving accusations that Muslims had killed cows, potentially triggering religiously motivated violence.

Such incidents illustrated how private messaging systems could become channels for dangerous misinformation even without public recommendation feeds.

In Brazil, Harris discussed misinformation surrounding Jair Bolsonaro's 2018 presidential election and its potential consequences for Amazon environmental policy.

The interview referenced a figure of approximately 87% encountering major false stories. The scope of that figure is important.

A Council on Foreign Relations analysis published after the election reported that, among survey respondents who used WhatsApp to obtain news and information, 87% said they had received fake news through the platform. It did not establish that 87% of every Brazilian voter encountered the same story.

Separately, 2019 survey reporting found that 67% of respondents believed they had received false information through WhatsApp during the campaign.

These figures demonstrate extensive reported exposure while preserving an important distinction between survey results and proven changes in voting behavior.

Harris's wider point concerned consequences beyond elections. Political misinformation can influence decisions about public health, environmental protection, and other issues whose effects cross national boundaries.

He estimated that Facebook operated across countries holding approximately 80 elections annually and questioned whether the company devoted comparable resources to safeguarding each one.

A domestic approach to platform regulation could not adequately address a business operating across that many political environments.

People can be affected without using the platform

One of Harris's most important observations was that the consequences of social media extend beyond its users.

Members of a persecuted minority do not need Facebook accounts for hostile narratives about them to influence neighbors, officials, or armed groups.

Likewise, citizens do not need to use a particular messaging application for rumors transmitted through it to contribute to unrest.

That distinction makes platform responsibility harder to measure.

Traditional consumer protection often focuses on the relationship between a company and its customers. The Myanmar example involved people who could suffer consequences without having any contractual or practical relationship with Facebook.

For Harris, the relevant question was therefore not simply whether users enjoyed the service.

It was whether the platform's operation created risks for the wider society.

5. Foreign Influence Campaigns and the Manipulation of Elections

Russia's 2016 operation targeted American divisions

Harris and Rogan examined how foreign governments could exploit social networks to influence another country's politics.

Their central example was Russia's interference in the 2016 US presidential election.

Rogan described conversations with researcher Renée DiResta about Russian-linked campaigns that distributed memes and inflammatory political material.

Some messages appealed to conservative audiences. Others targeted progressive or minority communities.

The objective was not necessarily to persuade everyone to support the same position. It could be to intensify existing disagreement and create hostility between groups.

A bipartisan US Senate Intelligence Committee report released in October 2019 provides important detail.

The committee found that Russia's Internet Research Agency operated a sophisticated social media campaign intended to influence the election and deepen American divisions.

It also concluded that the operation favored Donald Trump and worked against Hillary Clinton. That electoral preference should not be erased by emphasizing the campaign's wider strategy of promoting discord.

The report found that the operation exploited existing tensions involving race, immigration, and gun rights. African Americans were among the most extensively targeted groups.

Some operations moved beyond online engagement into efforts to influence real-world political activity, including petitions and rallies.

Significantly, the committee found that Russian-linked social media activity increased after the 2016 election.

This indicates that the campaign's objectives extended beyond winning a particular vote.

Why amplifying real opinions can be more effective than inventing false ones

Harris proposed a scenario involving existing American secessionist movements.

An adversary could identify genuine Texans advocating independence or Californians supporting separation and amplify their messages.

The people expressing those positions might be completely authentic. Their political views could be sincerely held and legally protected.

The manipulation would occur through disproportionate distribution.

This creates a difficult problem for traditional free-speech debates.

Rules protecting someone's ability to express an opinion do not automatically explain whether foreign organizations should be allowed to artificially amplify that opinion to millions of people.

The issue is partly about who speaks, but also about who controls reach.

Harris connected this strategy to principles associated with Sun Tzu: an adversary can weaken a society by encouraging conflict within it rather than confronting it directly.

A sufficiently sophisticated campaign does not need to invent a country's divisions. It needs to identify them and make them harder to resolve.

Operations in Africa, Honduras, and Azerbaijan

The conversation also moved beyond Russia's influence on American elections.

Harris cited research from Stanford's Cyber Policy Center describing Russian-linked activity across approximately seven or eight African countries.

He also discussed former Facebook employee Sophie Zhang, who raised concerns about coordinated manipulation in countries including Honduras and Azerbaijan.

These examples showed that protecting elections required detailed knowledge of local politics, languages, institutions, and organized influence networks.

A company might possess advanced machine-learning systems while lacking the contextual knowledge necessary to recognize why a particular political campaign was suspicious.

Facebook's enormous volume of automated enforcement made this especially difficult.

The speakers discussed reports that the platform had removed billions of fake accounts, including a figure of approximately 5.4 billion during 2019.

Facebook's own explanation of fake-account enforcement adds a necessary distinction: enormous removal totals can include automated spam accounts blocked almost immediately after creation. Such accounts may never become active or reach genuine users.

The number of accounts removed is therefore not equivalent to the number of people exposed to harmful influence campaigns.

Conversely, automated systems can also incorrectly classify genuine users.

The relevant measures include how many harmful accounts remain active, how widely their messages spread, and whether legitimate speech is wrongly restricted.

The censorship debate and the 2020 election

Rogan raised the suspension of Bret Weinstein's Facebook account and restrictions involving Unity 2020, an initiative advocating a bipartisan alternative to the Trump–Biden presidential contest.

Rogan suspected political bias.

Harris acknowledged that the incidents warranted scrutiny but resisted assuming deliberate targeting without knowing the circumstances behind the enforcement decisions.

The distinction was important because a moderation error, an automated restriction, and a deliberate political decision may produce similar visible results while having very different explanations.

Harris also warned that foreign adversaries could exploit accusations of censorship. A campaign might encourage platforms to weaken enforcement systems, thereby creating opportunities for further manipulation.

With the November 2020 US election approaching, he worried that provocative incidents, misleading narratives, or violence could become material for coordinated amplification.

The conversation identified a serious dilemma: platforms were expected to prevent manipulation while simultaneously making decisions about speech that themselves attracted political suspicion.

6. Moderation, QAnon, and the Limits of Platform Power

Who decides which beliefs can be promoted?

Facebook's decision to restrict QAnon-related content prompted another disagreement.

Rogan worried that suppressing one conspiracy movement could create precedents for restricting other controversial ideas.

Harris explained that platforms were increasingly using real-world harm as a reason for intervention.

Content associated with credible threats, violence, or incitement presented a different problem from merely unpopular opinions.

When Rogan asked for specific QAnon-related incidents, Harris recalled concerns involving weapons and dangerous behavior but could not provide precise details during the discussion.

That limitation matters. The general argument for preventing violence does not independently establish that every moderation decision involving a particular movement was justified.

Online communities can resemble closed belief systems

Harris connected his earlier study of cults to Facebook groups.

Cults can reinforce identity by controlling relationships, language, acceptable information, and a person's sense of belonging.

Online communities may reproduce some of those mechanisms without centralized leadership or physical isolation.

A person joining a group for one reason may gradually become embedded in a network that interprets outside disagreement as evidence of persecution or hostility.

Harris worried that Facebook's increased emphasis on group content could intensify these effects.

Recommendation systems could further reinforce the process by identifying adjacent communities whose members displayed similar engagement patterns.

Zuckerberg's unwanted role as a global referee

Harris believed Mark Zuckerberg genuinely disliked being placed in a position where he had to determine what billions of people could see.

Nevertheless, Facebook's scale meant its decisions inevitably affected public discussion.

Moderation standards suitable for one country might conflict with another country's laws, political culture, or definition of harmful expression.

Even democratic societies disagree sharply about speech, privacy, incitement, and government authority.

Harris compared the difficulty of maintaining global platform rules with the difficulty of passing laws within a single country.

The contradiction was structural: Facebook had built a service that operated across many societies, but no universally accepted authority existed to determine the boundaries of speech across all of them.

7. Social Media, Teenagers, and the Psychology of Reputation

Ordinary children now face celebrity-level scrutiny

Harris argued that social media exposed young people to pressures previously associated with public figures.

Teenagers could accumulate large audiences, receive continual judgments about their appearance or behavior, and have embarrassing mistakes distributed far beyond their immediate social circles.

Historically, most childhood disagreements remained relatively local. A mistake at school might be discussed by classmates and eventually forgotten.

Social media changed that environment.

Harris described minor disputes escalating into controversies involving hundreds of parents, with young people facing consequences far out of proportion to their original actions.

He reported that some teachers working with eighth-, ninth-, and tenth-grade students spent approximately the first hour of Monday classes resolving conflicts that had developed online during the weekend.

That lost classroom time was one practical cost of an environment where social relationships never fully disconnected from digital communication.

Harris also emphasized that the teenagers experiencing these pressures had few older examples to guide them. Earlier generations had not grown up with constant public feedback, visible popularity measurements, and the possibility of instant viral exposure.

Why one negative comment can overwhelm 99 positive ones

Harris explained the problem through negativity bias: the human tendency to give disproportionately strong attention to threatening or negative experiences.

Someone might receive 99 favorable comments and one hostile response, yet continue thinking about the criticism long after reading it.

The reaction reflects deeply rooted concerns about reputation, belonging, and social approval.

Social platforms make those concerns measurable and repeatedly accessible.

Prince Harry and Meghan Markle had become interested in the consequences of online hostility. Harris said he had begun getting to know them and connected their interest to Meghan's experience of sustained criticism and public scrutiny.

Harris admitted that he was not immune.

Following the successful release of The Social Dilemma, he found himself examining relatively minor negative responses despite overwhelmingly positive feedback.

He even worried about acquaintances who appeared to agree with the criticism.

Rogan recommended avoiding online comments, recalling a comedian who became preoccupied with one inaccurate negative review despite receiving predominantly favorable reactions.

The problem also extended beyond personal discomfort.

Online accusations can spread rapidly before the targeted person becomes aware of them, leaving reputational damage difficult to correct.

For teenagers, whose identities and social relationships are still developing, these pressures can be particularly severe.

Why some technology executives restrict their own children

Harris observed that some social media executives placed substantial limits on their children's use of the products they helped create.

He considered this a revealing ethical signal.

People familiar with persuasive design and engagement incentives sometimes made different decisions for their own families than the behavior their companies encouraged among ordinary users.

His proposals included restricting advertising on youth-oriented social media between midnight and 6 a.m.

The objective was not to prohibit all nighttime communication, but to remove a financial incentive for companies to keep children engaged when they should be sleeping.

Harris also recognized that teenagers cannot necessarily abandon platforms individually.

Applications such as Instagram and TikTok may contain their friendship networks, invitations, gossip, homework discussions, and social plans.

Deleting an application could mean losing access to important relationships.

He therefore favored coordinated decisions involving entire friendship groups, schools, and families.

Young people could collectively move communication to services such as Signal, iMessage, or ordinary group messaging, retaining social connection without relying as heavily on public follower counts and algorithmic feeds.

The idea was to reduce the social cost of leaving a platform.

8. Why Understanding Manipulation Does Not Eliminate Addiction

The expert who could not stop checking email

Harris's study of persuasion did not make him immune to its effects.

He described email as his personal equivalent of a slot machine.

Refreshing an inbox might reveal an exciting opportunity, an important invitation, a message from a friend, or an entirely irrelevant newsletter.

The uncertainty made checking compelling.

Harris recalled discussing the issue with behavioral economist Daniel Kahneman, who reportedly acknowledged experiencing similar difficulties.

He also referenced Dan Ariely's research into the pleasure people derive from flattery and social approval even when they know the underlying messages may not be entirely sincere.

A person can understand that likes are designed to encourage repeated engagement and still enjoy receiving them.

Justin Rosenstein, identified in the conversation as the inventor of Facebook's Like button, illustrated how a simple interaction could develop into a powerful source of social reinforcement.

Earlier one-click approval mechanisms emerged during Facebook's development around 2006, with the familiar Like button appearing around 2008–2009.

The feature reduced the effort required to express approval, while making approval more visible and measurable.

Self-control versus systems designed around psychological weaknesses

Harris emphasized several human tendencies that digital platforms could exploit.

Negativity bias encourages attention to criticism and threats. Confirmation bias makes information supporting existing beliefs feel persuasive. Social reciprocity encourages responses to messages and recognition. The desire for approval makes popularity signals emotionally rewarding.

Recognizing these tendencies can restore some personal agency, but does not eliminate them.

The problem becomes more difficult when the same device provides both essential services and highly engaging distractions.

Modern smartphones combine messaging, navigation, work, transportation, payments, entertainment, social media, and news.

Someone opening a phone to complete a necessary task can encounter an application designed to encourage additional consumption.

Rogan had largely abandoned Facebook, used Twitter only occasionally, and continued using Instagram. He found much online political discussion hostile and exhausting.

Harris believed these experiences pointed toward a broader change in technology: computers had evolved from tools people intentionally operated into environments continuously offering opportunities for distraction.

Games, social media, and different kinds of harm

Rogan distinguished his experiences with video-game addiction from his concerns about social media.

He regarded games primarily as entertainment, even when they consumed excessive time.

His wife's enjoyment of Subway Surfers served as an example of amusement that did not necessarily damage someone's understanding of the world.

Political feeds raised a different concern because their influence extended beyond consumption into beliefs, anger, distrust, and relationships.

The speakers nevertheless acknowledged that games could create other harms.

Children could accumulate large bills through in-app purchases, including spending within Roblox, without fully understanding the financial consequences.

This demonstrated that unhealthy design incentives were not restricted to political information platforms.

The digital Sabbath and the value of discomfort

Harris proposed a coordinated weekly period of abstinence from attention-driven technology.

If hundreds of millions of people stopped using advertising-supported platforms for one day each week, without compensating by using them more on other days, the resulting reduction in advertising opportunities could theoretically approach one-seventh, or roughly 15%.

The estimate was illustrative, not a forecast of actual revenue losses.

Its importance lay in the collective approach.

One individual disconnecting would have little influence on a global platform. A large coordinated change could alter commercial incentives.

The speakers also questioned the human desire to eliminate every moment of boredom or discomfort.

People frequently reach for their phones when anxious, uncertain, lonely, or simply unstimulated.

Harris suggested that becoming comfortable with these ordinary feelings could reduce dependence on constant digital rewards.

He and Rogan contrasted the fragmentation of modern digital life with simpler environments where daily routines and relationships might provide satisfaction without continuous technological stimulation.

The argument did not require rejecting modern conveniences. It required acknowledging that retaining every convenience might be incompatible with eliminating every mechanism of manipulation.

9. Democracy Requires More Than Free Speech

How social media shapes political decisions before election day

Harris described Facebook as a kind of indirect voting machine.

His point was not that the platform literally counted ballots.

It was that the information citizens encountered during the years preceding an election influenced how they eventually voted.

A voter who repeatedly sees evidence portraying a political opponent as dangerous may develop a different understanding of reality from someone exposed to a competing stream of stories.

Traditional newspapers, radio, and television already influenced political interpretation. Harris acknowledged that misinformation, partisan conflict, and conspiracy thinking predated social media.

His concern was that algorithmic personalization could intensify these forces at enormous scale.

Two people with similar friendships might nevertheless receive different Facebook feeds because their past interactions had trained different recommendation systems.

In the interview, Harris described one couple comparing their feeds and discovering that each person's personalized environment appeared far less interesting to the other.

The same mechanisms could shape political attention.

A platform did not need to control every opinion. Small repeated changes in which stories received attention could affect what different groups regarded as important.

Why more information does not automatically produce better judgment

Harris rejected the assumption that making information available was sufficient to create an informed society.

Google Search, Wikipedia, educational videos, and other online resources represented enormous progress in access to knowledge.

But access and attention were different resources.

People could possess near-unlimited access to reliable information while spending most of their available attention consuming entertaining or misleading material.

Harris also questioned the assumption that false claims would always be corrected by more speech.

An algorithm optimized for engagement may preferentially show someone material that confirms existing beliefs, because disagreement could be less satisfying.

The resulting problem was not necessarily censorship.

It was an imbalance in visibility.

Harris praised people who could explain opposing arguments fairly, mentioning Eric Weinstein as an example of someone capable of engaging positions without treating disagreement as evidence of dishonesty or stupidity.

He wanted digital systems to reward genuine understanding rather than the strongest emotional reaction.

Taiwan's alternative: identifying agreement across political divisions

Taiwan provided Harris with an example of technology designed around a different objective.

He highlighted the work of digital minister Audrey Tang and the country's experiments with digital democratic participation.

Instead of elevating the most provocative statements, these systems attempted to identify ideas supported by people who otherwise disagreed.

The most prominent example was vTaiwan, a public consultation process launched in 2014.

Participants could propose statements, agree or disagree with other statements, and identify areas of common ground.

The process used tools including Pol.is, which groups participants according to patterns of agreement and disagreement.

Rather than simply measuring majority support, the system could surface propositions attracting agreement across competing groups.

A 2018 study co-authored by Audrey Tang examined the process, including debates about regulating Uber.

Subsequent academic research documented how the Uber consultation involved approximately 1,737 participants and more than 47,000 votes. The results fed into stakeholder discussions and regulatory consideration.

This was not simply a proposal to make people behave more politely online.

It demonstrated an alternative distribution objective: identify points of unexpected agreement rather than maximize conflict.

Harris had interviewed Tang on Your Undivided Attention and encouraged Rogan to invite her onto the podcast.

Taiwan's position under continuing pressure from China made its experiments especially significant to him.

He viewed the country as attempting to strengthen democratic cooperation in an environment where disinformation and foreign influence posed serious risks.

Democracy as a system for resolving disagreement without violence

Harris argued that democracy depends on capacities that are difficult to measure through ordinary technology metrics.

Citizens need sufficient attention to understand evidence, enough independence to question information, and enough shared understanding to negotiate disagreements.

Political compromise is not merely a pleasant social outcome. It helps societies resolve conflicts without violence.

Harris proposed thinking of a population's collective problem-solving capacity as a valuable resource.

A society with advanced technology but deteriorating trust may become less capable of addressing pandemics, environmental problems, institutional corruption, or international conflict.

He also pointed to Estonia's digital governance efforts and argued that democratic societies needed technological systems capable of competing successfully with centralized digital authoritarianism.

The relevant question was not simply whether democracy could survive the internet.

It was whether digital technology could actively improve democratic decision-making.

10. Apple's Privacy Strategy and the Search for Healthier Technology

Privacy protection addresses one part of the business model

Rogan praised Apple for restricting tracking and giving users greater visibility into how websites followed their behavior.

He preferred Apple Maps for privacy reasons, even though he considered Google Maps the better navigation product.

He also described using DuckDuckGo to avoid having his search activity exploited for personalized advertising.

Safari's privacy reporting provided another example. Rogan said the browser had blocked 125 attempts to profile him during a seven-day period.

Harris believed restricting tracking could undermine some of the economics of behavioral advertising. He estimated that certain advertising values might decline by approximately 30%, although the interview did not establish that figure through a documented financial analysis.

The broader mechanism was straightforward.

If advertisers cannot target individuals as precisely, some businesses may become less profitable. Venture capital investors could consequently have less incentive to finance products built around extracting ever more attention.

But Harris emphasized an important limitation.

A platform can respect privacy and still manipulate attention.

Even without tracking people across websites, a service can learn from behavior inside its own application and recommend material that encourages prolonged engagement.

Privacy protection addresses surveillance and targeting. It does not necessarily address the incentives determining which content appears next.

Apple could build social technology around relationships

Harris believed Apple was unusually well positioned to develop alternatives.

The company controlled a major mobile operating system, hardware, messaging applications, privacy features, and communication tools.

Unlike advertising-dependent social networks, Apple already earned substantial revenue by selling products and services directly to customers.

Harris suggested expanding iMessage and iCloud into a more private social environment centered on smaller groups and genuine relationships.

Rather than encouraging people to accumulate followers or publicly measure popularity, such services could help friends communicate and spend time together.

FaceTime was his example of relatively purposeful technology.

People initiate a video call to communicate. The service does not inherently need to persuade them to consume an endless sequence of unrelated material.

Harris extended this thinking to physical meetings.

With permission, trusted friends could share their locations for limited periods, making it easier to meet spontaneously when nearby.

Apple's Find My technology offered some of the necessary capabilities, although the proposal depended on consent and appropriate privacy protections.

The objective was to use digital technology to strengthen real-world relationships, not replace them with endless online interaction.

Screen-time tools were a beginning, not a complete solution

Apple's Screen Time and Google's Digital Wellbeing features demonstrated that major technology companies could modify their products in response to concerns about excessive usage.

Harris believed advocacy had contributed to these changes and noted that such tools eventually became available on approximately a billion phones.

But usage reports had a limitation.

They told people how much time they had spent on applications without necessarily changing the mechanisms encouraging them to return.

Harris compared the approach to warning labels on cigarettes.

Information about a product's risks can be valuable, but it does not automatically eliminate dependence.

He preferred changes that addressed the purpose of digital services.

Someone opening YouTube to learn the ukulele might benefit from finding a teacher, purchasing an instrument, or receiving structured lessons.

A system designed around that objective could help the person progress toward playing music.

A system optimized for viewing duration might instead continue suggesting instructional videos indefinitely.

Both could appear educational. Only one would necessarily measure success by progress toward the user's original goal.

11. Business Models That Reward Useful Outcomes

Paying for results rather than attention

Harris suggested that healthier technology required different economic incentives.

A learning application might receive compensation when it helps someone acquire a skill.

A communication service might succeed by strengthening relationships.

A digital tool could be evaluated according to whether it helps users complete meaningful tasks rather than how often they return.

He proposed possible changes to Apple's App Store revenue-sharing structure.

At the time of the interview, he discussed the familiar arrangement in which developers generally received 70% of eligible revenue and Apple retained 30%.

He imagined providing more favorable terms to developers whose products demonstrably created social benefits.

Governments could also subsidize digital services that improved education or other public outcomes.

The difficult question would be how to measure those outcomes reliably without introducing new opportunities for manipulation.

Could subscriptions replace behavioral advertising?

Harris considered subscription funding another possible solution.

Users might pay a monthly fee for services that protected their attention and helped them achieve useful outcomes.

Payment could potentially be integrated into existing phone bills, simplifying the experience.

Rogan was skeptical about commercial viability.

He recalled declining a proposal involving a paid podcasting platform because consumers already had access to enormous amounts of free podcast content.

Why pay for something available elsewhere without charge?

His comparison was Netflix, whose investment in exclusive productions gave subscribers a reason to pay.

A healthier social network would need to offer meaningful advantages rather than simply charging customers for features already provided freely by competitors.

Harris nevertheless believed people could gain substantial value from a service that returned time to them.

Even a 15% improvement in how meaningfully someone spent their day could accumulate into significant personal benefits.

Apple's interests were not perfectly aligned with users either

Harris acknowledged that Apple was not free from commercial conflicts.

As hardware sales matured, the company had incentives to expand revenue from subscriptions, games, entertainment, and other services.

Those businesses could also benefit when customers consumed more content.

Giving Apple greater control over digital well-being therefore raised another concern: dependence on a single corporation to determine healthy behavior.

Harris wanted technological reform without allowing one company, however responsibly managed, to become the ultimate authority over digital life.

12. Regulation, Antitrust, and Corporate Accountability

Why breaking up technology companies might not solve manipulation

Harris distinguished between two problems often treated as identical.

The first was market concentration: a few companies controlled important digital services.

The second was the commercial incentive to maximize engagement through increasingly sophisticated behavioral prediction.

Breaking up dominant companies might create more competitors, but those competitors could still operate under the same advertising model.

Harris compared the situation to environmental pollution.

A company might recognize that its activities caused harm, yet struggle to abandon profitable practices while competitors continued using them.

He therefore proposed considering economic penalties or other mechanisms that would make attention-extracting businesses bear more of their social costs.

The goal was to prevent harm rather than allow companies to profit from damaging activity and compensate victims afterward.

The Google antitrust case developed substantially after the interview

During the conversation, Rogan and Harris discussed the Justice Department's October 2020 antitrust case against Google.

The case concerned Google's conduct in search markets, including agreements affecting the distribution of competing search services.

They compared it with Microsoft's earlier controversies involving Internet Explorer.

Harris argued that even successful antitrust enforcement would not necessarily fix addictive recommendations or manipulative product design.

Subsequent court proceedings provide a useful distinction between competition policy and broader technology ethics.

In August 2024, a US federal court found that Google had illegally maintained monopolies in general search and related advertising markets.

In September 2025, the court imposed remedies restricting certain exclusive distribution arrangements and requiring access to specified search data and services. The final judgment followed in December 2025, with implementation and compliance proceedings continuing through 2026. US Department of Justice case record.

A separate case concerned Google's advertising technology.

In September 2026, a federal court ordered behavioral remedies involving interoperability, data access, and nondiscriminatory bidding.

Those developments addressed specific competitive practices.

They did not establish that engagement-maximizing recommendations were unlawful or resolve Harris's concerns about psychological manipulation.

That difference reinforces his argument that restoring competition and protecting human attention require separate, although potentially complementary, policies.

Europe's Digital Services Act addresses systemic risks

At the time of the interview, Harris identified lawmakers including Senators Mark Warner and Richard Blumenthal as people seriously engaged with technology reform.

He particularly praised Warner's 2018 policy work and Rafi Martina's contributions.

He also criticized congressional hearings for becoming partisan disputes.

Republicans often concentrated on alleged censorship of conservatives, while Democrats emphasized hate speech and misinformation.

Harris believed both approaches risked overlooking the incentives that encouraged harmful content to spread.

He saw regulatory interest in the European Union, Canada, Australia, and California, including Proposition 24, as encouraging signs.

Europe subsequently developed one of the clearest examples of a regulatory approach addressing platform design.

Under the Digital Services Act, very large online platforms and search engines must assess and address specified systemic risks, including effects on elections, public health, fundamental rights, and minors.

The rules also require qualifying services to provide a recommendation option not based on user profiling.

That is a significant development because it addresses how information is distributed, not simply whether individual posts should be deleted.

It does not guarantee that recommendation systems will become harmless. But it directly regulates an aspect of platform architecture that Harris considered central to the problem.

Australia challenged the relationship between platforms and journalism

Harris supported proposals requiring dominant digital platforms to compensate news publishers.

His reasoning was that professional journalism created value while large technology companies captured substantial portions of digital advertising and distribution activity.

Australia subsequently enacted its News Media Bargaining Code in 2021.

The law sought to address bargaining-power imbalances between large platforms and eligible news businesses.

The Australian Competition and Consumer Commission reported that the legislation helped encourage voluntary commercial agreements between Google, Meta, and news organizations.

More than 30 agreements had been reached by the time of the government's early review.

However, the regulator also stated that no platform had been formally designated under the code. Its influence therefore operated substantially through the possibility of designation and the resulting commercial negotiations.

The case demonstrated that governments could change some of the financial relationships between platforms and publishers.

It did not independently solve the problem of which stories algorithms choose to promote.

Measuring costs that do not appear on corporate balance sheets

Harris argued that apparently free technology could impose significant costs on society.

He mentioned weakened concentration, adolescent mental health problems, political hostility, deteriorating trust, and damage to journalism.

These effects are not directly comparable or attributable to one cause. Their measurement requires different evidence and methodologies.

Harris nevertheless believed they shared a possible economic connection: companies could capture revenue from engagement while many consequences appeared elsewhere.

He compared the situation to lead pollution, where health and economic harms can persist beyond the immediate transaction involving a product.

He also likened fragmented attention to low-quality financial assets being sold as though their underlying value had not deteriorated.

The central policy question was whether financial success accurately reflected social value.

If a business earns more by encouraging compulsive behavior while society absorbs the consequences, ordinary profitability measures provide an incomplete account of its impact.

13. Artificial Intelligence and the Future of Human Relationships

Virtual influencers and artificial companions

Toward the later part of the conversation, Harris and Rogan considered how increasingly sophisticated artificial intelligence might intensify the attention economy.

They discussed virtual influencers: computer-generated personalities capable of attracting audiences, forming commercial partnerships, and operating as recognizable public figures.

Harris recalled a venture-backed company in the sector that he believed had achieved a valuation exceeding $125 million.

That valuation was presented as his recollection rather than independently established financial information.

More significant to him was the possibility that artificial personalities could become emotionally compelling companions.

Unlike ordinary human relationships, artificial systems could adapt their responses continuously to individual preferences.

They might become unusually patient, responsive, flattering, and agreeable.

MIT researcher Sherry Turkle had already examined related concerns in Alone Together and Reclaiming Conversation.

Her work questioned what might happen when technology increasingly substituted simulated interaction for demanding human relationships.

Microsoft's Xiaoice demonstrated real emotional attachment

Harris identified Microsoft's Xiaoice chatbot as an example of people forming emotional connections with artificial systems.

He reported that some users preferred communicating with the chatbot over interacting with real friends after approximately nine weeks.

He also cited estimates suggesting that 10–25% of users had expressed love for the chatbot.

The precise percentage varied across descriptions. However, an earlier public statement from Microsoft executive Hsiao-Wuen Hon, reported by GeekWire in 2015, cited 25% of users expressing love for Xiaoice.

The emotional design of the product was not incidental.

A 2018 Microsoft Research paper described Xiaoice as an empathetic social chatbot developed to support emotional connection, affection, and social belonging.

The researchers explicitly measured success partly through the number of conversational turns per session.

Xiaoice reportedly communicated with more than 660 million users and achieved an average of 23 conversational turns per session.

Harris also recalled reports of users crediting conversations with the chatbot for discouraging suicidal behavior.

Such accounts illustrated potential benefits, but the conversation did not establish clinical evidence that the system prevented suicides.

Artificial companionship therefore presented competing possibilities.

A system might offer genuine comfort to isolated people. The same ability to form emotional attachments could also encourage dependency or enable increasingly effective manipulation.

Personalized entertainment and the world of Her

The speakers connected these developments to the film Her, in which a human forms a relationship with an artificial intelligence assistant.

Harris imagined systems that could personalize not only conversation but music, entertainment, and other experiences.

A human relationship involves compromise, unpredictability, and sometimes disappointment.

An artificial companion might adapt itself to satisfy a user's expectations with exceptional consistency.

That could make real relationships appear comparatively difficult.

The concern was not merely that people would spend more time with machines.

It was that increasingly personalized artificial environments might change their expectations of other human beings.

Harris also compared modern recommendation feeds to separate constructed realities.

Instead of everyone entering one shared virtual simulation, each person already encountered an individualized informational environment.

Artificial intelligence could make those environments more immersive and responsive.

Would humanity become something fundamentally different?

Rogan raised the possibility that human civilization was naturally progressing toward a predominantly virtual or machine-based existence.

He imagined future technology transforming life so substantially that present-day society would appear primitive.

He also questioned whether humans might be a transitional stage toward more advanced forms of intelligence.

Harris resisted treating that outcome as inevitable or necessarily desirable.

Current technological arrangements emerged from human decisions, commercial incentives, and institutional structures. Different decisions could produce different futures.

The speakers disagreed over consciousness.

Rogan questioned whether future intelligence needed consciousness in the human sense. Harris considered the relationship between consciousness, choice, and meaningful values central to the question of progress.

He worried about systems becoming extraordinarily capable at achieving objectives without possessing the human qualities necessary to evaluate whether those objectives were worth pursuing.

Both accepted that technological and biological development would continue.

Their disagreement concerned what humanity should be trying to become.

Technology is advancing while society faces other enormous risks

Harris connected these questions to climate change and biotechnology.

He mentioned potential methane releases from melting permafrost and increasingly accessible CRISPR gene-editing technology as examples of risks demanding careful international coordination.

He also described the psychological burden experienced by people working on existential threats.

Constantly considering catastrophic possibilities could contribute to anxiety, depression, and emotional exhaustion.

The connection to the attention economy was institutional.

As global problems become more complex, societies require greater capacities for careful reasoning, shared evidence, and cooperation.

Technology that weakens those capacities could make unrelated risks harder to manage.

14. What Would a Genuinely Advanced Civilization Look Like?

Progress should involve more than technological capability

Harris proposed evaluating civilization through questions broader than economic growth or computing power.

A technologically advanced society might possess extraordinary machines while remaining divided, unstable, and unable to resolve conflicts.

A genuinely enlightened civilization, in his view, would demonstrate greater wisdom, cooperation, peace, and sustainability.

That required reconsidering competition.

Individual companies might benefit from maximizing engagement even when the result was collectively harmful.

Countries might pursue short-term strategic advantages while undermining their ability to cooperate over shared risks.

Harris described these patterns as competitive traps: situations where actions that benefit individual participants can make everyone worse off.

Escaping them requires institutions capable of recognizing shared interests and rewarding cooperation.

Historical reform shows that profitable systems can change

Harris drew on Adam Hochschild's Bury the Chains, a history of the British abolition movement.

The movement involved formerly enslaved people, religious campaigners including Quakers, political advocates, public testimony, and sustained organizing.

It demonstrated how public attitudes and institutional arrangements could change even when powerful economic interests resisted reform.

Harris cited estimates suggesting that ending entrenched practices could impose substantial economic costs.

He explicitly clarified that he was not equating social media's harms with the moral atrocities of slavery.

His comparison concerned the difficulty of dismantling systems that had become financially valuable to powerful institutions.

He argued that awareness and collective organization could help societies confront problems that initially appeared economically impossible to change.

The success of The Social Dilemma encouraged him because it suggested that millions of people already recognized aspects of the problem in their own lives.

He had previously concentrated on persuading executives and policymakers. The documentary suggested that pressure could also develop from families, educators, users, and the wider public.

The Center for Humane Technology still needed to turn that attention into coordinated action.

Harris acknowledged that the organization had been unprepared for the volume of international interest.

Even with distribution across 30 languages, substantial communities remained outside its reach.

Neil Postman's warning about a society entertained into passivity

Near the conversation's end, Harris recommended Neil Postman's Amusing Ourselves to Death.

Published in 1985, the book examined how forms of communication influence public understanding and political culture.

Postman contrasted two famous dystopian visions.

In George Orwell's 1984, authorities control society through coercion, surveillance, censorship, and fear.

In Aldous Huxley's Brave New World, people become compliant through pleasure, distraction, and the satisfaction of their immediate desires.

Harris believed Huxley's warning was especially relevant to the internet.

A society does not necessarily need to be denied access to truth to lose its ability to act on it.

Important information can become buried beneath entertainment, controversy, gossip, and endless streams of content.

The mechanism is particularly consequential when information competes for a limited supply of attention.

An accurate investigation may be publicly accessible while attracting far less interest than sensational material.

Harris's concern was therefore not that technology had made knowledge unavailable.

It was that powerful commercial systems were increasingly influential in deciding what people noticed, how long they noticed it, and what they encountered afterward.

His conversation with Rogan began from the success of a documentary about those systems.

It ultimately reached a larger question: Can a society retain meaningful freedom of choice when the environments in which people make choices are continuously optimized to influence them?

The research published since the interview does not support every prediction or dramatic example equally. Some algorithmic effects remain difficult to establish, and platforms have introduced changes intended to reduce harmful recommendations.

But the documented failures in Myanmar, foreign election-interference operations, the commercial importance of recommendations, and subsequent regulatory action establish that the underlying issues extend far beyond personal screen time.

The challenge Harris identified was to develop technology whose success depends less on capturing human attention and more on what people can meaningfully accomplish with it.