Group items tagged

Lately, it has also become one of the most controversial. The algorithm has been found to be promoting conspiracy theories about the Las Vegas mass shooting and incentivising, through recommendations, a thriving subculture that targets children with disturbing content

One YouTube creator who was banned from making advertising revenues from his strange videos – which featured his children receiving flu shots, removing earwax, and crying over dead pets – told a reporter he had only been responding to the demands of Google’s algorithm. “That’s what got us out there and popular,” he said. “We learned to fuel it and do whatever it took to please the algorithm.”

academics have speculated that YouTube’s algorithms may have been instrumental in fuelling disinformation during the 2016 presidential election. “YouTube is the most overlooked story of 2016,” Zeynep Tufekci, a widely respected sociologist and technology critic, tweeted back in October. “Its search and recommender algorithms are misinformation engines.”

Those are not easy questions to answer. Like all big tech companies, YouTube does not allow us to see the algorithms that shape our lives. They are secret formulas, proprietary software, and only select engineers are entrusted to work on the algorithm

Guillaume Chaslot, a 36-year-old French computer programmer with a PhD in artificial intelligence, was one of those engineers.

The experience led him to conclude that the priorities YouTube gives its algorithms are dangerously skewed.

Chaslot said none of his proposed fixes were taken up by his managers. “There are many ways YouTube can change its algorithms to suppress fake news and improve the quality and diversity of videos people see,” he says. “I tried to change YouTube from the inside but it didn’t work.”

Chaslot explains that the algorithm never stays the same. It is constantly changing the weight it gives to different signals: the viewing patterns of a user, for example, or the length of time a video is watched before someone clicks away.

The engineers he worked with were responsible for continuously experimenting with new formulas that would increase advertising revenues by extending the amount of time people watched videos. “Watch time was the priority,” he recalls. “Everything else was considered a distraction.”

Chaslot was fired by Google in 2013, ostensibly over performance issues. He insists he was let go after agitating for change within the company, using his personal time to team up with like-minded engineers to propose changes that could diversify the content people see.

He was especially worried about the distortions that might result from a simplistic focus on showing people videos they found irresistible, creating filter bubbles, for example, that only show people content that reinforces their existing view of the world.

“YouTube is something that looks like reality, but it is distorted to make you spend more time online,” he tells me when we meet in Berkeley, California. “The recommendation algorithm is not optimising for what is truthful, or balanced, or healthy for democracy.”

YouTube told me that its recommendation system had evolved since Chaslot worked at the company and now “goes beyond optimising for watchtime”.

It did not say why Google, which acquired YouTube in 2006, waited over a decade to make those changes

Chaslot believes such changes are mostly cosmetic, and have failed to fundamentally alter some disturbing biases that have evolved in the algorithm

It finds videos through a word search, selecting a “seed” video to begin with, and recording several layers of videos that YouTube recommends in the “up next” column. It does so with no viewing history, ensuring the videos being detected are YouTube’s generic recommendations, rather than videos personalised to a user. And it repeats the process thousands of times, accumulating layers of data about YouTube recommendations to build up a picture of the algorithm’s preferences.

Each study finds something different, but the research suggests YouTube systematically amplifies videos that are divisive, sensational and conspiratorial.

When his program found a seed video by searching the query “who is Michelle Obama?” and then followed the chain of “up next” suggestions, for example, most of the recommended videos said she “is a man”

He believes one of the most shocking examples was detected by his program in the run-up to the 2016 presidential election. As he observed in a short, largely unnoticed blogpost published after Donald Trump was elected, the impact of YouTube’s recommendation algorithm was not neutral during the presidential race: it was pushing videos that were, in the main, helpful to Trump and damaging to Hillary Clinton.

“It was strange,” he explains to me. “Wherever you started, whether it was from a Trump search or a Clinton search, the recommendation algorithm was much more likely to push you in a pro-Trump direction.”

Trump won the electoral college as a result of 80,000 votes spread across three swing states. There were more than 150 million YouTube users in the US. The videos contained in Chaslot’s database of YouTube-recommended election videos were watched, in total, more than 3bn times before the vote in November 2016.

“Algorithms that shape the content we see can have a lot of impact, particularly on people who have not made up their mind,”

“Gentle, implicit, quiet nudging can over time edge us toward choices we might not have otherwise made.”

But what was most compelling was how often Chaslot’s software detected anti-Clinton conspiracy videos appearing “up next” beside other videos.

I spent weeks watching, sorting and categorising the trove of videos with Erin McCormick, an investigative reporter and expert in database analysis. From the start, we were stunned by how many extreme and conspiratorial videos had been recommended, and the fact that almost all of them appeared to be directed against Clinton.

“This research captured the apparent direction of YouTube’s political ecosystem,” he says. “That has not been done before.”

There were too many videos in the database for us to watch them all, so we focused on 1,000 of the top-recommended videos. We sifted through them one by one to determine whether the content was likely to have benefited Trump or Clinton. Just over a third of the videos were either unrelated to the election or contained content that was broadly neutral or even-handed. Of the remaining 643 videos, 551 were videos favouring Trump, while only only 92 favoured the Clinton campaign.

The sample we had looked at suggested Chaslot’s conclusion was correct: YouTube was six times more likely to recommend videos that aided Trump than his adversary.

The spokesperson added: “Our search and recommendation systems reflect what people search for, the number of videos available, and the videos people choose to watch on YouTube. That’s not a bias towards any particular candidate; that is a reflection of viewer interest.”

YouTube seemed to be saying that its algorithm was a neutral mirror of the desires of the people who use it – if we don’t like what it does, we have ourselves to blame. How does YouTube interpret “viewer interest” – and aren’t “the videos people choose to watch” influenced by what the company shows them?

Offered the choice, we may instinctively click on a video of a dead man in a Japanese forest, or a fake news clip claiming Bill Clinton raped a 13-year-old. But are those in-the-moment impulses really a reflect of the content we want to be fed?

YouTube’s recommendation system has probably figured out that edgy and hateful content is engaging. “This is a bit like an autopilot cafeteria in a school that has figured out children have sweet teeth, and also like fatty and salty foods,” she says. “So you make a line offering such food, automatically loading the next plate as soon as the bag of chips or candy in front of the young person has been consumed.”

Once that gets normalised, however, what is fractionally more edgy or bizarre becomes, Tufekci says, novel and interesting. “So the food gets higher and higher in sugar, fat and salt – natural human cravings – while the videos recommended and auto-played by YouTube get more and more bizarre or hateful.”

“This is important research because it seems to be the first systematic look into how YouTube may have been manipulated,” he says, raising the possibility that the algorithm was gamed as part of the same propaganda campaigns that flourished on Twitter and Facebook.

“We believe that the activity we found was limited because of various safeguards that we had in place in advance of the 2016 election, and the fact that Google’s products didn’t lend themselves to the kind of micro-targeting or viral dissemination that these actors seemed to prefer.”

Senator Mark Warner, the ranking Democrat on the intelligence committee, later wrote to the company about the algorithm, which he said seemed “particularly susceptible to foreign influence”. The senator demanded to know what the company was specifically doing to prevent a “malign incursion” of YouTube’s recommendation system. Walker, in his written reply, offered few specifics

Tristan Harris, a former Google insider turned tech whistleblower, likes to describe Facebook as a “living, breathing crime scene for what happened in the 2016 election” that federal investigators have no access to. The same might be said of YouTube. About half the videos Chaslot’s program detected being recommended during the election have now vanished from YouTube – many of them taken down by their creators. Chaslot has always thought this suspicious. These were videos with titles such as “Must Watch!! Hillary Clinton tried to ban this video”, watched millions of times before they disappeared. “Why would someone take down a video that has been viewed millions of times?” he asks

I shared the entire database of 8,000 YouTube-recommended videos with John Kelly, the chief executive of the commercial analytics firm Graphika, which has been tracking political disinformation campaigns. He ran the list against his own database of Twitter accounts active during the election, and concluded many of the videos appeared to have been pushed by networks of Twitter sock puppets and bots controlled by pro-Trump digital consultants with “a presumably unsolicited assist” from Russia.

“I don’t have smoking-gun proof of who logged in to control those accounts,” he says. “But judging from the history of what we’ve seen those accounts doing before, and the characteristics of how they tweet and interconnect, they are assembled and controlled by someone – someone whose job was to elect Trump.”

After the Senate’s correspondence with Google over possible Russian interference with YouTube’s recommendation algorithm was made public last week, YouTube sent me a new statement. It emphasised changes it made in 2017 to discourage the recommendation system from promoting some types of problematic content. “We appreciate the Guardian’s work to shine a spotlight on this challenging issue,” it added. “We know there is more to do here and we’re looking forward to making more announcements in the months ahead.”

In the months leading up to the election, the Next News Network turned into a factory of anti-Clinton news and opinion, producing dozens of videos a day and reaching an audience comparable to that of MSNBC’s YouTube channel. Chaslot’s research indicated Franchi’s success could largely be credited to YouTube’s algorithms, which consistently amplified his videos to be played “up next”. YouTube had sharply dismissed Chaslot’s research.

I contacted Franchi to see who was right. He sent me screen grabs of the private data given to people who upload YouTube videos, including a breakdown of how their audiences found their clips. The largest source of traffic to the Bill Clinton rape video, which was viewed 2.4m times in the month leading up to the election, was YouTube recommendations.

The same was true of all but one of the videos Franchi sent me data for. A typical example was a Next News Network video entitled “WHOA! HILLARY THINKS CAMERA’S OFF… SENDS SHOCK MESSAGE TO TRUMP” in which Franchi, pointing to a tiny movement of Clinton’s lips during a TV debate, claims she says “fuck you” to her presidential rival. The data Franchi shared revealed in the month leading up to the election, 73% of the traffic to the video – amounting to 1.2m of its views – was due to YouTube recommendations. External traffic accounted for only 3% of the views.

many of the other creators of anti-Clinton videos I spoke to were amateur sleuths or part-time conspiracy theorists. Typically, they might receive a few hundred views on their videos, so they were shocked when their anti-Clinton videos started to receive millions of views, as if they were being pushed by an invisible force.

In every case, the largest source of traffic – the invisible force – came from the clips appearing in the “up next” column. William Ramsey, an occult investigator from southern California who made “Irrefutable Proof: Hillary Clinton Has a Seizure Disorder!”, shared screen grabs that showed the recommendation algorithm pushed his video even after YouTube had emailed him to say it violated its guidelines. Ramsey’s data showed the video was watched 2.4m times by US-based users before election day. “For a nobody like me, that’s a lot,” he says. “Enough to sway the election, right?”

Daniel Alexander Cannon, a conspiracy theorist from South Carolina, tells me: “Every video I put out about the Clintons, YouTube would push it through the roof.” His best-performing clip was a video titled “Hillary and Bill Clinton ‘The 10 Photos You Must See’”, essentially a slideshow of appalling (and seemingly doctored) images of the Clintons with voiceover in which Cannon speculates on their health. It has been seen 3.7m times on YouTube, and 2.9m of those views, Cannon said, came from “up next” recommendations.

his research also does something more important: revealing how thoroughly our lives are now mediated by artificial intelligence.

Less than a generation ago, the way voters viewed their politicians was largely shaped by tens of thousands of newspaper editors, journalists and TV executives. Today, the invisible codes behind the big technology platforms have become the new kingmakers.

They pluck from obscurity people like Dave Todeschini, a retired IBM engineer who, “let off steam” during the election by recording himself opining on Clinton’s supposed involvement in paedophilia, child sacrifice and cannibalism. “It was crazy, it was nuts,” he said of the avalanche of traffic to his YouTube channel, which by election day had more than 2m views

In testimony to U.S. Congress and abroad, whistleblower Frances Haugen has pointed to the algorithm as central to the social network’s problems, arguing that it systematically amplifies and rewards hateful, divisive, misleading and sometimes outright false content by putting it at the top of users’ feeds.

The political push raises an old question for Facebook: Why not just give users the power to turn off their feed ranking algorithms voluntarily? Would letting users opt to see every post from the people they follow, in chronological order, be so bad?

The documents suggest that Facebook’s defense of algorithmic rankings stems not only from its business interests, but from a paternalistic conviction, backed by data, that its sophisticated personalization software knows what users want better than the users themselves

Since 2009, three years after it launched the news feed, Facebook has used software that predicts which posts each user will find most interesting and places those at the top of their feeds while burying others. That system, which has evolved in complexity to take in as many as 10,000 pieces of information about each post, has fueled the news feed’s growth into a dominant information source.

The proliferation of false information, conspiracy theories and partisan propaganda on Facebook and other social networks has led some to wonder whether we wouldn’t all be better off with a simpler, older system: one that simply shows people all the messages, pictures and videos from everyone they follow, in the order they were posted.

That was more or less how Instagram worked until 2016, and Twitter until 2017.

But Facebook has long resisted it.

they appear to have been informed mostly by data on user engagement, at least until recently

That employee, who said they had worked on and studied the news feed for two years, went on to question whether automated ranking might also come with costs that are harder to measure than the benefits. “Even asking this question feels slightly blasphemous at Facebook,” they added.

“Whenever we’ve tried to compare ranked and unranked feeds, ranked feeds just seem better,” wrote an employee in a memo titled, “Is ranking good?”, which was posted to the company’s internal network, Facebook Workplace, in 2018

In 2014, another internal report, titled “Feed ranking is good,” summarized the results of tests that found allowing users to turn off the algorithm led them to spend less time in their news feeds, post less often and interact less.

Without an algorithm deciding which posts to show at the top of users’ feeds, concluded the report’s author, whose name was redacted, “Facebook would probably be shrinking.”

there’s a catch: The setting only applies for as long as you stay logged in. When you leave and come back, the ranking algorithm will be back on.

What many users may not realize is that Facebook actually does offer an option to see a mostly chronological feed, called “most recent,”

The longer Facebook left the user’s feed in chronological order, the less time they spent on it, the less they posted, and the less often they returned to Facebook.

A separate report from 2018, first described by Alex Kantrowitz’s newsletter Big Technology, found that turning off the algorithm unilaterally for a subset of Facebook users, and showing them posts mostly in the order they were posted, led to “massive engagement drops.” Notably, it also found that users saw more low-quality content in their feeds, at least at first, although the company’s researchers were able to mitigate that with more aggressive “integrity” measures.

Nick Clegg, the company’s vice president of global affairs, said in a TV interview last month that if Facebook were to remove the news feed algorithm, “the first thing that would happen is that people would see more, not less, hate speech; more, not less, misinformation; more, not less, harmful content. Why? Because those algorithmic systems precisely are designed like a great sort of giant spam filter to identify and deprecate and downgrade bad content.”

because the algorithm has always been there, Facebook users haven’t been given the time or the tools to curate their feeds for themselves in thoughtful ways. In other words, Facebook has never really given a chronological news feed a fair shot to succeed

Some critics say that’s a straw-man argument. Simply removing automated rankings for a subset of users, on a social network that has been built to rely heavily on those systems, is not the same as designing a service to work well without them,

Ben Grosser, a professor of new media at University of Illinois at Urbana-Champaign. Those users’ feeds are no longer curated, but the posts they’re seeing are still influenced by the algorithm’s reward systems. That is, they’re still seeing content from people and publishers who are vying for the likes, shares and comments that drive Facebook’s recommendati

“My experience from watching a chronological feed within a social network that isn’t always trying to optimize for growth is that a lot of these problems” — such as hate speech, trolling and manipulative media — “just don’t exist.”

Facebook has not taken an official stand on the legislation that would require social networks to offer a chronological feed option, but Clegg said in an op-ed last month that the company is open to regulation around algorithms, transparency, and user controls.Twitter, for its part, signaled potential support for the bills.

“I think users have the right to expect social media experiences free of recommendation algorithms,” Maréchal added. “As a user, I want to have as much control over my own experience as possible, and recommendation algorithms take that control away from me.”

“Only companies themselves can do the experiments to find the answers. And as talented as industry researchers are, we can’t trust executives to make decisions in the public interest based on that research, or to let the public and policymakers access that research.”

ns.

Again and again, the Facebook Papers show staffers sounding alarms about the dangers posed by the platform—how Facebook amplifies extremism and misinformation, how it incites violence, how it encourages radicalization and political polarization. Again and again, staffers reckon with the ways in which Facebook’s decisions stoke these harms, and they plead with leadership to do more.

And again and again, staffers say, Facebook’s leaders ignore them.

Facebook has dismissed the concerns of its employees in manifold ways.

One of its cleverer tactics is to argue that staffers who have raised the alarm about the damage done by their employer are simply enjoying Facebook’s “very open culture,” in which people are encouraged to share their opinions, a spokesperson told me. This stance allows Facebook to claim transparency while ignoring the substance of the complaints, and the implication of the complaints: that many of Facebook’s employees believe their company operates without a moral compass.

When you stitch together the stories that spanned the period between Joe Biden’s election and his inauguration, it’s easy to see Facebook as instrumental to the attack on January 6. (A spokesperson told me that the notion that Facebook played an instrumental role in the insurrection is “absurd.”)

what emerges from a close reading of Facebook documents, and observation of the manner in which the company connects large groups of people quickly, is that Facebook isn’t a passive tool but a catalyst. Had the organizers tried to plan the rally using other technologies of earlier eras, such as telephones, they would have had to identify and reach out individually to each prospective participant, then persuade them to travel to Washington. Facebook made people’s efforts at coordination highly visible on a global scale.

The platform not only helped them recruit participants but offered people a sense of strength in numbers. Facebook proved to be the perfect hype machine for the coup-inclined.

In November 2019, Facebook staffers noticed they had a serious problem. Facebook offers a collection of one-tap emoji reactions. Today, they include “like,” “love,” “care,” “haha,” “wow,” “sad,” and “angry.” Company researchers had found that the posts dominated by “angry” reactions were substantially more likely to go against community standards, including prohibitions on various types of misinformation, according to internal documents.

In July 2020, researchers presented the findings of a series of experiments. At the time, Facebook was already weighting the reactions other than “like” more heavily in its algorithm—meaning posts that got an “angry” reaction were more likely to show up in users’ News Feeds than posts that simply got a “like.” Anger-inducing content didn’t spread just because people were more likely to share things that made them angry; the algorithm gave anger-inducing content an edge. Facebook’s Integrity workers—employees tasked with tackling problems such as misinformation and espionage on the platform—concluded that they had good reason to believe targeting posts that induced anger would help stop the spread of harmful content.

By dialing anger’s weight back to zero in the algorithm, the researchers found, they could keep posts to which people reacted angrily from being viewed by as many users. That, in turn, translated to a significant (up to 5 percent) reduction in the hate speech, civic misinformation, bullying, and violent posts—all of which are correlated with offline violence—to which users were exposed.

Facebook rolled out the change in early September 2020, documents show; a Facebook spokesperson confirmed that the change has remained in effect. It was a real victory for employees of the Integrity team.

But it doesn’t normally work out that way. In April 2020, according to Frances Haugen’s filings with the SEC, Facebook employees had recommended tweaking the algorithm so that the News Feed would deprioritize the surfacing of content for people based on their Facebook friends’ behavior. The idea was that a person’s News Feed should be shaped more by people and groups that a person had chosen to follow. Up until that point, if your Facebook friend saw a conspiracy theory and reacted to it, Facebook’s algorithm might show it to you, too. The algorithm treated any engagement in your network as a signal that something was worth sharing. But now Facebook workers wanted to build circuit breakers to slow this form of sharing.

Experiments showed that this change would impede the distribution of hateful, polarizing, and violence-inciting content in people’s News Feeds. But Zuckerberg “rejected this intervention that could have reduced the risk of violence in the 2020 election,” Haugen’s SEC filing says. An internal message characterizing Zuckerberg’s reasoning says he wanted to avoid new features that would get in the way of “meaningful social interactions.” But according to Facebook’s definition, its employees say, engagement is considered “meaningful” even when it entails bullying, hate speech, and reshares of harmful content.

This episode, like Facebook’s response to the incitement that proliferated between the election and January 6, reflects a fundamental problem with the platform

Facebook’s megascale allows the company to influence the speech and thought patterns of billions of people. What the world is seeing now, through the window provided by reams of internal documents, is that Facebook catalogs and studies the harm it inflicts on people. And then it keeps harming people anyway.

“I am worried that Mark’s continuing pattern of answering a different question than the question that was asked is a symptom of some larger problem,” wrote one Facebook employee in an internal post in June 2020, referring to Zuckerberg. “I sincerely hope that I am wrong, and I’m still hopeful for progress. But I also fully understand my colleagues who have given up on this company, and I can’t blame them for leaving. Facebook is not neutral, and working here isn’t either.”

It is quite a thing to see, the sheer number of Facebook employees—people who presumably understand their company as well as or better than outside observers—who believe their employer to be morally bankrupt.

I spoke with several former Facebook employees who described the company’s metrics-driven culture as extreme, even by Silicon Valley standards

Facebook workers are under tremendous pressure to quantitatively demonstrate their individual contributions to the company’s growth goals, they told me. New products and features aren’t approved unless the staffers pitching them demonstrate how they will drive engagement.

As a result, Facebook has stoked an algorithm arms race within its ranks, pitting core product-and-engineering teams, such as the News Feed team, against their colleagues on Integrity teams, who are tasked with mitigating harm on the platform. These teams establish goals that are often in direct conflict with each other.

One of Facebook’s Integrity staffers wrote at length about this dynamic in a goodbye note to colleagues in August 2020, describing how risks to Facebook users “fester” because of the “asymmetrical” burden placed on employees to “demonstrate legitimacy and user value” before launching any harm-mitigation tactics—a burden not shared by those developing new features or algorithm changes with growth and engagement in mind

The note said:We were willing to act only after things had spiraled into a dire state … Personally, during the time that we hesitated, I’ve seen folks from my hometown go further and further down the rabbithole of QAnon and Covid anti-mask/anti-vax conspiracy on FB. It has been painful to observe.

Current and former Facebook employees describe the same fundamentally broken culture—one in which effective tactics for making Facebook safer are rolled back by leadership or never approved in the first place.

That broken culture has produced a broken platform: an algorithmic ecosystem in which users are pushed toward ever more extreme content, and where Facebook knowingly exposes its users to conspiracy theories, disinformation, and incitement to violence.

One example is a program that amounts to a whitelist for VIPs on Facebook, allowing some of the users most likely to spread misinformation to break Facebook’s rules without facing consequences. Under the program, internal documents show, millions of high-profile users—including politicians—are left alone by Facebook even when they incite violence

whitelisting influential users with massive followings on Facebook isn’t just a secret and uneven application of Facebook’s rules; it amounts to “protecting content that is especially likely to deceive, and hence to harm, people on our platforms.”

Facebook workers tried and failed to end the program. Only when its existence was reported in September by The Wall Street Journal did Facebook’s Oversight Board ask leadership for more information about the practice. Last week, the board publicly rebuked Facebook for not being “fully forthcoming” about the program.

e worries have been exacerbated lately by fears about a decline in new posts on Facebook, two former employees who left the company in recent years told me. People are posting new material less frequently to Facebook, and its users are on average older than those of other social platforms.

All of this makes the platform rely more heavily on ways it can manipulate what its users see in order to reach its goals. This explains why Facebook is so dependent on the infrastructure of groups, as well as making reshares highly visible, to keep people hooked.

Zuckerberg has defined Facebook’s mission as making “social infrastructure to give people the power to build a global community that works for all of us,” but in internal research documents his employees point out that communities aren’t always good for society:

When part of a community, individuals typically act in a prosocial manner. They conform, they forge alliances, they cooperate, they organize, they display loyalty, they expect obedience, they share information, they influence others, and so on. Being in a group changes their behavior, their abilities, and, importantly, their capability to harm themselves or others

Thus, when people come together and form communities around harmful topics or identities, the potential for harm can be greater.

The infrastructure choices that Facebook is making to keep its platform relevant are driving down the quality of the site, and exposing its users to more dangers

hose dangers are also unevenly distributed, because of the manner in which certain subpopulations are algorithmically ushered toward like-minded groups

And the subpopulations of Facebook users who are most exposed to dangerous content are also most likely to be in groups where it won’t get reported.

And it knows that 3 percent of Facebook users in the United States are super-consumers of conspiracy theories, accounting for 37 percent of known consumption of misinformation on the platform.

Zuckerberg’s positioning of Facebook’s role in the insurrection is odd. He lumps his company in with traditional media organizations—something he’s ordinarily loath to do, lest the platform be expected to take more responsibility for the quality of the content that appears on it—and suggests that Facebook did more, and did better, than journalism outlets in its response to January 6. What he fails to say is that journalism outlets would never be in the position to help investigators this way, because insurrectionists don’t typically use newspapers and magazines to recruit people for coups.

Facebook wants people to believe that the public must choose between Facebook as it is, on the one hand, and free speech, on the other. This is a false choice. Facebook has a sophisticated understanding of measures it could take to make its platform safer without resorting to broad or ideologically driven censorship tactics.

Facebook knows that no two people see the same version of the platform, and that certain subpopulations experience far more dangerous versions than others do

Facebook knows that people who are isolated—recently widowed or divorced, say, or geographically distant from loved ones—are disproportionately at risk of being exposed to harmful content on the platform.

It knows that repeat offenders are disproportionately responsible for spreading misinformation.

“We can’t pretend we don’t see information consumption patterns, and how deeply problematic they are for the longevity of democratic discourse,” a user-experience researcher wrote in an internal comment thread in 2019, in response to a now-infamous memo from Andrew “Boz” Bosworth, a longtime Facebook executive. “There is no neutral position at this stage, it would be powerfully immoral to commit to amorality.”

It could consistently enforce its policies regardless of a user’s political power.

Facebook could ban reshares.

It could choose to optimize its platform for safety and quality rather than for growth.

It could tweak its algorithm to prevent widespread distribution of harmful content.

Facebook could create a transparent dashboard so that all of its users can see what’s going viral in real time.

It could make public its rules for how frequently groups can post and how quickly they can grow.

It could also automatically throttle groups when they’re growing too fast, and cap the rate of virality for content that’s spreading too quickly.

Facebook could shift the burden of proof toward people and communities to demonstrate that they’re good actors—and treat reach as a privilege, not a right

Facebook could say that its platform is not for everyone. It could sound an alarm for those who wander into the most dangerous corners of Facebook, and those who encounter disproportionately high levels of harmful content

It could do all of these things. But it doesn’t.

Lately, people have been debating just how nefarious Facebook really is. One argument goes something like this: Facebook’s algorithms aren’t magic, its ad targeting isn’t even that good, and most people aren’t that stupid.

All of this may be true, but that shouldn’t be reassuring. An algorithm may just be a big dumb means to an end, a clunky way of maneuvering a massive, dynamic network toward a desired outcome. But Facebook’s enormous size gives it tremendous, unstable power.

Facebook takes whole populations of people, pushes them toward radicalism, and then steers the radicalized toward one another.

When the most powerful company in the world possesses an instrument for manipulating billions of people—an instrument that only it can control, and that its own employees say is badly broken and dangerous—we should take notice.

The lesson for individuals is this:

You must be vigilant about the informational streams you swim in, deliberate about how you spend your precious attention, unforgiving of those who weaponize your emotions and cognition for their own profit, and deeply untrusting of any scenario in which you’re surrounded by a mob of people who agree with everything you’re saying.

Without seeing how Facebook works at a finer resolution, in real time, we won’t be able to understand how to make the social web compatible with democracy.

The Ofqual algorithm was the technical embodiment of a deeply political idea: that a person is only as good as their circumstances dictate. The metric took no account of how hard a person had worked, while its appeal system sought to deny individual redress, and only the “ranking” of students remained from the centres’ inputs

The A-level scandal made algorithms an object of direct resistance and exposed what many already know to be the case: that this type of decision-making involves far more than a series of computational steps

In their designs and assumptions, algorithms shape the world in which they’re used. To decide whether to include or exclude a data input, or to weight one feature over another are not merely technical questions – they’re also political propositions about what a society can and should be like.

In this case, Ofqual’s model decided it’s not possible that good teaching, hard work and inspiration can make a difference to a young person’s life and their grades.

The politics of the algorithm were visible for all to see. Many decisions – from what constitutes a “small” subject entry to whether a cohort’s prior attainment should nudge down the distribution curve – had profound and arbitrary effects on real lives

Grappling openly and transparently with difficult questions, such as how to achieve fairness, is precisely what characterises ethical decision-making in a society. Instead, Ofqual responded with non-disclosure agreements, offering no public insight into what it was doing as it tested competing models.

Algorithms offer governments the allure of definitive solutions and the promise of reducing intractable decisions to simplified outputs.

This logic runs counter to democratic politics, which express the contingency of the world and the deliberative nature of collective decision-making.

They do this by sucking up and analyzing your data, using it to predict what will cause you to react most strongly, and then giving you more of that.

Algorithms that maximize attention give an advantage to negative messages. People tend to react more to inputs that land low on the brainstem. Fear and anger produce a lot more engagement and sharing than joy

The result is that the algorithms favor sensational content over substance.

for mass media, this was constrained by one-size-fits-all content and by the limitations of delivery platforms. Not so for internet platforms on smartphones. They have created billions of individual channels, each of which can be pushed further into negativity and extremism without the risk of alienating other audience members

On Facebook, it’s your news feed, while on Google it’s your individually customized search results. The result is that everyone sees a different version of the internet tailored to create the illusion that everyone else agrees with them.

It took Brexit for me to begin to see the danger of this dynamic. I’m no expert on British politics, but it seemed likely that Facebook might have had a big impact on the vote because one side’s message was perfect for the algorithms and the other’s wasn’t. The “Leave” campaign made an absurd promise—there would be savings from leaving the European Union that would fund a big improvement in the National Health System—while also exploiting xenophobia by casting Brexit as the best way to protect English culture and jobs from immigrants. It was too-good-to-be-true nonsense mixed with fearmongering.

Facebook was a much cheaper and more effective platform for Leave in terms of cost per user reached. And filter bubbles would ensure that people on the Leave side would rarely have their questionable beliefs challenged. Facebook’s model may have had the power to reshape an entire continent.

Tristan Harris, formerly the design ethicist at Google. Tristan had just appeared on 60 Minutes to discuss the public health threat from social networks like Facebook. An expert in persuasive technology, he described the techniques that tech platforms use to create addiction and the ways they exploit that addiction to increase profits. He called it “brain hacking.”

The most important tool used by Facebook and Google to hold user attention is filter bubbles. The use of algorithms to give consumers “what they want” leads to an unending stream of posts that confirm each user’s existing beliefs

Continuous reinforcement of existing beliefs tends to entrench those beliefs more deeply, while also making them more extreme and resistant to contrary facts

No one stopped them from siphoning off the profits of content creators. No one stopped them from gathering data on every aspect of every user’s internet life. No one stopped them from amassing market share not seen since the days of Standard Oil.

Facebook takes the concept one step further with its “groups” feature, which encourages like-minded users to congregate around shared interests or beliefs. While this ostensibly provides a benefit to users, the larger benefit goes to advertisers, who can target audiences even more effectively.

We theorized that the Russians had identified a set of users susceptible to its message, used Facebook’s advertising tools to identify users with similar profiles, and used ads to persuade those people to join groups dedicated to controversial issues. Facebook’s algorithms would have favored Trump’s crude message and the anti-Clinton conspiracy theories that thrilled his supporters, with the likely consequence that Trump and his backers paid less than Clinton for Facebook advertising per person reached.

The ads were less important, though, than what came next: once users were in groups, the Russians could have used fake American troll accounts and computerized “bots” to share incendiary messages and organize events.

Trolls and bots impersonating Americans would have created the illusion of greater support for radical ideas than actually existed.

Real users “like” posts shared by trolls and bots and share them on their own news feeds, so that small investments in advertising and memes posted to Facebook groups would reach tens of millions of people.

A similar strategy prevailed on other platforms, including Twitter. Both techniques, bots and trolls, take time and money to develop—but the payoff would have been huge.

2016 was just the beginning. Without immediate and aggressive action from Washington, bad actors of all kinds would be able to use Facebook and other platforms to manipulate the American electorate in future elections.

Renee DiResta, an expert in how conspiracy theories spread on the internet. Renee described how bad actors plant a rumor on sites like 4chan and Reddit, leverage the disenchanted people on those sites to create buzz, build phony news sites with “press” versions of the rumor, push the story onto Twitter to attract the real media, then blow up the story for the masses on Facebook.

It was sophisticated hacker technique, but not expensive. We hypothesized that the Russians were able to manipulate tens of millions of American voters for a sum less than it would take to buy an F-35 fighter jet.

Algorithms can be beautiful in mathematical terms, but they are only as good as the people who create them. In the case of Facebook and Google, the algorithms have flaws that are increasingly obvious and dangerous.

Thanks to the U.S. government’s laissez-faire approach to regulation, the internet platforms were able to pursue business strategies that would not have been allowed in prior decades. No one stopped them from using free products to centralize the internet and then replace its core functions.

To the contrary: the platforms help people self-segregate into like-minded filter bubbles, reducing the risk of exposure to challenging ideas.

No one stopped them from running massive social and psychological experiments on their users. No one demanded that they police their platforms. It has been a sweet deal.

Facebook and Google are now so large that traditional tools of regulation may no longer be effective.

The largest antitrust fine in EU history bounced off Google like a spitball off a battleship.

It reads like the plot of a sci-fi novel: a technology celebrated for bringing people together is exploited by a hostile power to drive people apart, undermine democracy, and create misery. This is precisely what happened in the United States during the 2016 election.

We had constructed a modern Maginot Line—half the world’s defense spending and cyber-hardened financial centers, all built to ward off attacks from abroad—never imagining that an enemy could infect the minds of our citizens through inventions of our own making, at minimal cost

Not only was the attack an overwhelming success, but it was also a persistent one, as the political party that benefited refuses to acknowledge reality. The attacks continue every day, posing an existential threat to our democratic processes and independence.

Facebook, Google, Twitter, and other platforms were manipulated by the Russians to shift outcomes in Brexit and the U.S. presidential election, and unless major changes are made, they will be manipulated again. Next time, there is no telling who the manipulators will be.

Unfortunately, there is no regulatory silver bullet. The scope of the problem requires a multi-pronged approach.

Polls suggest that about a third of Americans believe that Russian interference is fake news, despite unanimous agreement to the contrary by the country’s intelligence agencies. Helping those people accept the truth is a priority. I recommend that Facebook, Google, Twitter, and others be required to contact each person touched by Russian content with a personal message that says, “You, and we, were manipulated by the Russians. This really happened, and here is the evidence.” The message would include every Russian message the user received.

This idea, which originated with my colleague Tristan Harris, is based on experience with cults. When you want to deprogram a cult member, it is really important that the call to action come from another member of the cult, ideally the leader.

decentralization had a cost: no one had an incentive to make internet tools easy to use. Frustrated by those tools, users embraced easy-to-use alternatives from Facebook and Google. This allowed the platforms to centralize the internet, inserting themselves between users and content, effectively imposing a tax on both sides. This is a great business model for Facebook and Google—and convenient in the short term for customers—but we are drowning in evidence that there are costs that society may not be able to afford.

Second, the chief executive officers of Facebook, Google, Twitter, and others—not just their lawyers—must testify before congressional committees in open session

This is important not just for the public, but also for another crucial constituency: the employees who keep the tech giants running. While many of the folks who run Silicon Valley are extreme libertarians, the people who work there tend to be idealists. They want to believe what they’re doing is good. Forcing tech CEOs like Mark Zuckerberg to justify the unjustifiable, in public—without the shield of spokespeople or PR spin—would go a long way to puncturing their carefully preserved cults of personality in the eyes of their employees.

We also need regulatory fixes. Here are a few ideas.

First, it’s essential to ban digital bots that impersonate humans. They distort the “public square” in a way that was never possible in history, no matter how many anonymous leaflets you printed.

At a minimum, the law could require explicit labeling of all bots, the ability for users to block them, and liability on the part of platform vendors for the harm bots cause.

Second, the platforms should not be allowed to make any acquisitions until they have addressed the damage caused to date, taken steps to prevent harm in the future, and demonstrated that such acquisitions will not result in diminished competition.

An underappreciated aspect of the platforms’ growth is their pattern of gobbling up smaller firms—in Facebook’s case, that includes Instagram and WhatsApp; in Google’s, it includes YouTube, Google Maps, AdSense, and many others—and using them to extend their monopoly power.

This is important, because the internet has lost something very valuable. The early internet was designed to be decentralized. It treated all content and all content owners equally. That equality had value in society, as it kept the playing field level and encouraged new entrants.

There’s no doubt that the platforms have the technological capacity to reach out to every affected person. No matter the cost, platform companies must absorb it as the price for their carelessness in allowing the manipulation.

Third, the platforms must be transparent about who is behind political and issues-based communication.

Transparency with respect to those who sponsor political advertising of all kinds is a step toward rebuilding trust in our political institutions.

Fourth, the platforms must be more transparent about their algorithms. Users deserve to know why they see what they see in their news feeds and search results. If Facebook and Google had to be up-front about the reason you’re seeing conspiracy theories—namely, that it’s good for business—they would be far less likely to stick to that tactic

Allowing third parties to audit the algorithms would go even further toward maintaining transparency. Facebook and Google make millions of editorial choices every hour and must accept responsibility for the consequences of those choices. Consumers should also be able to see what attributes are causing advertisers to target them.

Fifth, the platforms should be required to have a more equitable contractual relationship with users. Facebook, Google, and others have asserted unprecedented rights with respect to end-user license agreements (EULAs), the contracts that specify the relationship between platform and user.

All software platforms should be required to offer a legitimate opt-out, one that enables users to stick with the prior version if they do not like the new EULA.

“Forking” platforms between old and new versions would have several benefits: increased consumer choice, greater transparency on the EULA, and more care in the rollout of new functionality, among others. It would limit the risk that platforms would run massive social experiments on millions—or billions—of users without appropriate prior notification. Maintaining more than one version of their services would be expensive for Facebook, Google, and the rest, but in software that has always been one of the costs of success. Why should this generation get a pass?

Sixth, we need a limit on the commercial exploitation of consumer data by internet platforms. Customers understand that their “free” use of platforms like Facebook and Google gives the platforms license to exploit personal data. The problem is that platforms are using that data in ways consumers do not understand, and might not accept if they did.

Not only do the platforms use your data on their own sites, but they also lease it to third parties to use all over the internet. And they will use that data forever, unless someone tells them to stop.

There should be a statute of limitations on the use of consumer data by a platform and its customers. Perhaps that limit should be ninety days, perhaps a year. But at some point, users must have the right to renegotiate the terms of how their data is used.

Seventh, consumers, not the platforms, should own their own data. In the case of Facebook, this includes posts, friends, and events—in short, the entire social graph. Users created this data, so they should have the right to export it to other social networks.

It would be analogous to the regulation of the AT&T monopoly’s long-distance business, which led to lower prices and better service for consumers.

Eighth, and finally, we should consider that the time has come to revive the country’s traditional approach to monopoly. Since the Reagan era, antitrust law has operated under the principle that monopoly is not a problem so long as it doesn’t result in higher prices for consumers.

Under that framework, Facebook and Google have been allowed to dominate several industries—not just search and social media but also email, video, photos, and digital ad sales, among others—increasing their monopolies by buying potential rivals like YouTube and Instagram.

While superficially appealing, this approach ignores costs that don’t show up in a price tag. Addiction to Facebook, YouTube, and other platforms has a cost. Election manipulation has a cost. Reduced innovation and shrinkage of the entrepreneurial economy has a cost. All of these costs are evident today. We can quantify them well enough to appreciate that the costs to consumers of concentration on the internet are unacceptably high.

In 2019, Facebook came up with a solution to offload the most difficult content moderation decisions. The company created the "Oversight Board," a quasi-judicial body that Facebook claims is independent. The Board, stocked with impressive thinkers from around the world, would issue "rulings" about whether certain Facebook content moderation decisions were correct.

the decision, which is nearly 12,000 words long, illustrates that whether Trump is ultimately allowed to return to Facebook is of limited significance. The more important questions are about the nature of the algorithm that gives people with views like Trump such a powerful voice on Facebook. 

The Oversight Board was Facebook's idea. It spent years constructing the organization, selected its chairs, and funded its endowment. But now that the Oversight Board is finally up and running and taking on high-profile cases, Facebook is choosing to ignore questions that the Oversight Board believes are essential to doing its job.

This is a key passage (emphasis added): 

duces no original reporting. But, on Facebook in April, The Daily Wire received more than double the distribution of the Washington Post and the New York Times combined:

A critical issue, as the Oversight Board suggests, is not simply Trump's posts but how those kinds of posts are amplified by Facebook's algorithms. Equally important is how Facebook's algorithms amplify false, paranoid, violent, right-wing content from people other than Trump — including those that follow Trump on Facebook.

The jurisdiction of the Oversight Board excludes both the algorithm and Facebook's business practices.

Facebook stated to the Board that it considered Mr. Trump’s “repeated use of Facebook and other platforms to undermine confidence in the integrity of the election (necessitating repeated application by Facebook of authoritative labels correcting the misinformation) represented an extraordinary abuse of the platform.” The Board sought clarification from Facebook about the extent to which the platform’s design decisions, including algorithms, policies, procedures and technical features, amplified Mr. Trump’s posts after the election and whether Facebook had conducted any internal analysis of whether such design decisions may have contributed to the events of January 6. Facebook declined to answer these questions. This makes it difficult for the Board to assess whether less severe measures, taken earlier, may have been sufficient to protect the rights of others.

Donald Trump's Facebook page is a symptom, not the cause, of the problem. Its algorithm favors low-quality, far-right content. Trump is just one of many beneficiaries.

NewsWhip is a social media analytics service which tracks which websites get the most engagement on Facebook. It just released its analysis for April and it shows low-quality right-wing aggregation sites dominate major news organizations.

The Oversight Board has no power to compel Facebook to answer. It's an important reminder that, for all the pomp and circumstance, the Oversight Board is not a court. The scope of its authority is limited by Facebook executives' willingness to play along. 

This actually understates how much better The Daily Wire's content performs on Facebook than the Washington Post and the New York Times. The Daily Wire published just 1,385 pieces of content in April compared to over 6,000 by the Washington Post and the New York Times. Each piece of content The Daily Wire published in April received 54,084 engagements on Facebook, compared to 2,943 for the New York Times and 1,973 for the Washington Post. 

It's important to note here that Facebook's algorithm is not reflecting reality — it's creating a reality that doesn't exist anywhere else. In the rest of the world, Western Journal is not more popular than the New York Times, NBC News, the BBC, and the Washington Post. That's only true on Facebook.

Facebook has made a conscious decision to surface low-quality content and recognizes its dangers.

Shortly after the November election, Facebook temporarily tweaked its algorithm to emphasize "'news ecosystem quality' scores, or N.E.Q., a secret internal ranking it assigns to news publishers based on signals about the quality of their journalism." The purpose was to attempt to cut down on election misinformation being spread on the platform by Trump and his allies. The result was "a spike in visibility for big, mainstream publishers like CNN, The New York Times and NPR, while posts from highly engaged hyperpartisan pages, such as Breitbart and Occupy Democrats, became less visible." 

BuzzFeed reported that some Facebook staff members wanted to make the change permanent. But that suggestion was opposed by Joel Kaplan, a top Facebook executive and Republican operative who frequently intervenes on behalf of right-wing publishers. The algorithm change was quickly rolled back.

Other proposed changes to the Facebook algorithm over the years have been rejected or altered because of their potential negative impact on right-wing sites like The Daily Wire. 

Less money is spent on black patients with the same level of need as white patients, causing the algorithm to conclude that black patients were less sick, the researchers found.

Reformulating these biases in the algorithm would more than double the number of black patients flagged for additional care

black patients actually had 48,772 more active chronic conditions than white patients who had been ranked at the same level of risk

Biases like these are inadvertently built into the technology we use at many different stages, said Ruha Benjamin, author of Race After Technology and associate professor of African American studies at Princeton University.

“Pre-existing social processes shape data collection, algorithm design and even the formulation of problems that need addressing by technology,” she said.

When researchers tweaked the algorithm to make predictions about patients’ future health conditions rather than which patients would incur the highest costs, it reduced biases by 84%. “These results suggest that label biases are fixable,” the study said.

Predictive algorithms that power these tools should be continually reviewed and refined

researchers suggested similar biases probably exist across a number of industries. As algorithms are increasingly used for job recruiting, housing loans and policing, Benjamin noted that more legislation is needed to ensure algorithms take into consideration historical biases.

“Indifference to social reality is, perhaps, more dangerous than outright bigotry.”

In the three years since Google Brain began making smart recommendations, watch time from the YouTube home page has grown 20-fold. More than 70 percent of the time people spend watching videos on YouTube, they spend watching videos suggested by Google Brain.

The more videos that are watched, the more ads that are seen, and the more money Google makes.

“We also wanted to serve the needs of people when they didn’t necessarily know what they wanted to look for.”

Last week, The New York Times reported that YouTube’s algorithm was encouraging pedophiles to watch videos of partially-clothed children, often after they watched sexual content.

o YouTube’s nuance-blind algorithm — trained to think with simple logic — serving up more videos to sate a sadist’s appetite is a job well done.

The result? The algorithm — and, consequently, YouTube — incentivizes bad behavior in viewers.

the algorithm relies on snapshots of visual content, rather than actions. If you (or your child) watch one Peppa Pig video, you’ll likely want another. And as long as it’s Peppa Pig in the frame, it doesn’t matter what the character does in the skit.

it didn’t take long for inappropriate videos to show up in YouTube Kids’ ‘Now playing’ feeds

Using cheap, widely available technology, animators created original video content featuring some of Hollywood’s best-loved characters. While an official Disney Mickey Mouse would never swear or act violently, in these videos Mickey and other children’s characters were sexual or violent

there’s a 3.5 percent chance of a child coming across inappropriate footage within 10 clicks of a child-friendly video.

Just four in 10 parents always monitor their child’s YouTube usage — and one in 20 children aged 4-to-12 say their parents never check what they’re watching.

At the height of the panic around Mr. Crowder’s videos, YouTube’s public policy on hate speech and harassment appeared to shift four times in a 24-hour period as the company sought to clarify what the new normal was.

One possible solution that would address both problems would be to strip out YouTube’s recommendation altogether. But it is highly unlikely that YouTube would ever do such a thing: that algorithm drives vast swaths of YouTube’s views, and to take it away would reduce the time viewers spend watching its videos, as well as reduce Google’s ad revenue.

it must, at the very least, make significant changes, and have greater human involvement in the recommendation process. The platform has some human moderators looking at so-called “borderline” content to train its algorithms, but more humanity is needed in the entire process.

Currently, the recommendation engine cannot understand why it shouldn’t recommend videos of children to pedophiles, and it cannot understand why it shouldn’t suggest sexually explicit videos to children. It cannot understand, because the incentives are twisted: every new video view, regardless of who the viewer is and what the viewer’s motives may be, is considered a success.

There’s a notion that the social web was once useful, or at least that it could have been good, if only we had pulled a few levers: some moderation and fact-checking here, a bit of regulation there, perhaps a federal antitrust lawsuit. But that’s far too sunny and shortsighted a view.

Today’s social networks, Facebook chief among them, were built to encourage the things that make them so harmful. It is in their very architecture.

I realized only recently that I’ve been thinking far too narrowly about the problem.

Megascale is nearly the existential threat that megadeath is. No single machine should be able to control the fate of the world’s population—and that’s what both the Doomsday Machine and Facebook are built to do.

Facebook does not exist to seek truth and report it, or to improve civic health, or to hold the powerful to account, or to represent the interests of its users, though these phenomena may be occasional by-products of its existence.

The company’s early mission was to “give people the power to share and make the world more open and connected.” Instead, it took the concept of “community” and sapped it of all moral meaning.

Facebook—along with Google and YouTube—is perfect for amplifying and spreading disinformation at lightning speed to global audiences.

Facebook decided that it needed not just a very large user base, but a tremendous one, unprecedented in size. That decision set Facebook on a path to escape velocity, to a tipping point where it can harm society just by existing.

No one, not even Mark Zuckerberg, can control the product he made. I’ve come to realize that Facebook is not a media company. It’s a Doomsday Machine.

Scale and engagement are valuable to Facebook because they’re valuable to advertisers. These incentives lead to design choices such as reaction buttons that encourage users to engage easily and often, which in turn encourage users to share ideas that will provoke a strong response.

Every time you click a reaction button on Facebook, an algorithm records it, and sharpens its portrait of who you are.

The hyper-targeting of users, made possible by reams of their personal data, creates the perfect environment for manipulation—by advertisers, by political campaigns, by emissaries of disinformation, and of course by Facebook itself, which ultimately controls what you see and what you don’t see on the site.

there aren’t enough moderators speaking enough languages, working enough hours, to stop the biblical flood of shit that Facebook unleashes on the world, because 10 times out of 10, the algorithm is faster and more powerful than a person.

At megascale, this algorithmically warped personalized informational environment is extraordinarily difficult to moderate in a meaningful way, and extraordinarily dangerous as a result.

These dangers are not theoretical, and they’re exacerbated by megascale, which makes the platform a tantalizing place to experiment on people

Even after U.S. intelligence agencies identified Facebook as a main battleground for information warfare and foreign interference in the 2016 election, the company has failed to stop the spread of extremism, hate speech, propaganda, disinformation, and conspiracy theories on its site.

it wasn’t until October of this year, for instance, that Facebook announced it would remove groups, pages, and Instragram accounts devoted to QAnon, as well as any posts denying the Holocaust.

In the days after the 2020 presidential election, Zuckerberg authorized a tweak to the Facebook algorithm so that high-accuracy news sources such as NPR would receive preferential visibility in people’s feeds, and hyper-partisan pages such as Breitbart News’s and Occupy Democrats’ would be buried, according to The New York Times, offering proof that Facebook could, if it wanted to, turn a dial to reduce disinformation—and offering a reminder that Facebook has the power to flip a switch and change what billions of people see online.

reducing the prevalence of content that Facebook calls “bad for the world” also reduces people’s engagement with the site. In its experiments with human intervention, the Times reported, Facebook calibrated the dial so that just enough harmful content stayed in users’ news feeds to keep them coming back for more.

Facebook’s stated mission—to make the world more open and connected—has always seemed, to me, phony at best, and imperialist at worst.

Facebook is a borderless nation-state, with a population of users nearly as big as China and India combined, and it is governed largely by secret algorithms

How much real-world violence would never have happened if Facebook didn’t exist? One of the people I’ve asked is Joshua Geltzer, a former White House counterterrorism official who is now teaching at Georgetown Law. In counterterrorism circles, he told me, people are fond of pointing out how good the United States has been at keeping terrorists out since 9/11. That’s wrong, he said. In fact, “terrorists are entering every single day, every single hour, every single minute” through Facebook.

Evidence of real-world violence can be easily traced back to both Facebook and 8kun. But 8kun doesn’t manipulate its users or the informational environment they’re in. Both sites are harmful. But Facebook might actually be worse for humanity.

In previous eras, U.S. officials could at least study, say, Nazi propaganda during World War II, and fully grasp what the Nazis wanted people to believe. Today, “it’s not a filter bubble; it’s a filter shroud,” Geltzer said. “I don’t even know what others with personalized experiences are seeing.”

Mary McCord, the legal director at the Institute for Constitutional Advocacy and Protection at Georgetown Law, told me that she thinks 8kun may be more blatant in terms of promoting violence but that Facebook is “in some ways way worse” because of its reach. “There’s no barrier to entry with Facebook,” she said. “In every situation of extremist violence we’ve looked into, we’ve found Facebook postings. And that reaches tons of people. The broad reach is what brings people into the fold and normalizes extremism and makes it mainstream.” In other words, it’s the megascale that makes Facebook so dangerous.

Facebook’s megascale gives Zuckerberg an unprecedented degree of influence over the global population. If he isn’t the most powerful person on the planet, he’s very near the top.

“The thing he oversees has such an effect on cognition and people’s beliefs, which can change what they do with their nuclear weapons or their dollars.”

Facebook’s new oversight board, formed in response to backlash against the platform and tasked with making decisions concerning moderation and free expression, is an extension of that power. “The first 10 decisions they make will have more effect on speech in the country and the world than the next 10 decisions rendered by the U.S. Supreme Court,” Geltzer said. “That’s power. That’s real power.”

Facebook is also a business, and a place where people spend time with one another. Put it this way: If you owned a store and someone walked in and started shouting Nazi propaganda or recruiting terrorists near the cash register, would you, as the shop owner, tell all of the other customers you couldn’t possibly intervene?

In 2004, Zuckerberg said Facebook ran advertisements only to cover server costs. But over the next two years Facebook completely upended and redefined the entire advertising industry. The pre-social web destroyed classified ads, but the one-two punch of Facebook and Google decimated local news and most of the magazine industry—publications fought in earnest for digital pennies, which had replaced print dollars, and social giants scooped them all up anyway.

In other words, if the Dunbar number for running a company or maintaining a cohesive social life is 150 people; the magic number for a functional social platform is maybe 20,000 people. Facebook now has 2.7 billion monthly users.

in 2007, Zuckerberg said something in an interview with the Los Angeles Times that now takes on a much darker meaning: “The things that are most powerful aren’t the things that people would have done otherwise if they didn’t do them on Facebook. Instead, it’s the things that would never have happened otherwise.”

We’re still in the infancy of this century’s triple digital revolution of the internet, smartphones, and the social web, and we find ourselves in a dangerous and unstable informational environment, powerless to resist forces of manipulation and exploitation that we know are exerted on us but remain mostly invisible

The Doomsday Machine offers a lesson: We should not accept this current arrangement. No single machine should be able to control so many people.

we need a new philosophical and moral framework for living with the social web—a new Enlightenment for the information age, and one that will carry us back to shared reality and empiricism.

localized approach is part of what made megascale possible. Early constraints around membership—the requirement at first that users attended Harvard, and then that they attended any Ivy League school, and then that they had an email address ending in .edu—offered a sense of cohesiveness and community. It made people feel more comfortable sharing more of themselves. And more sharing among clearly defined demographics was good for business.

we need to adopt a broader view of what it will take to fix the brokenness of the social web. That will require challenging the logic of today’s platforms—and first and foremost challenging the very concept of megascale as a way that humans gather.

The web’s existing logic tells us that social platforms are free in exchange for a feast of user data; that major networks are necessarily global and centralized; that moderators make the rules. None of that need be the case.

We need people who dismantle these notions by building alternatives. And we need enough people to care about these other alternatives to break the spell of venture capital and mass attention that fuels megascale and creates fatalism about the web as it is now.

We must also find ways to repair the aspects of our society and culture that the social web has badly damaged. This will require intellectual independence, respectful debate, and the same rebellious streak that helped establish Enlightenment values centuries ago.

Right now, too many people are allowing algorithms and tech giants to manipulate them, and reality is slipping from our grasp as a result. This century’s Doomsday Machine is here, and humming along.

Luckily, she also has a solution.

Kids today are, simply put, not learning how to be curious, critical adults — and they don’t seem to know what they’ve lost.

These high school students had become reliant, maybe even dependent, on social media companies’ algorithms.

Their dependence on technology sounds familiar to most of us. So many of us can barely remember when we didn’t have Amazon to fall back on when we needed a last-minute gift or when we waited by the radio for our favorite songs to play. Today, information, entertainment and connection are delivered to us on a conveyor belt, with less effort and exploration required of us than ever before.

What the kids I spoke to did not know is that these algorithms have been designed in a way that inevitably makes — and keeps — users unhappy.

A report by the nonprofit Center for Countering Digital Hate found that users could be served content related to suicide less than three minutes after downloading TikTok. Five minutes after that, they could come across a community promoting eating disorder content. Instagram is awash with soft-core pornography, offering a gateway to hard-core material on other sites (which are often equally lax about age verification). And all over social media are highly curated and filtered fake lives, breeding a sense of envy and inadequacy inside the developing brains of teenagers.

Social media companies know that content that generates negative feelings holds our attention longer than that which makes us feel good.

If you are a teenager feeling bad about yourself, your social media feed will typically keep delivering you videos and pictures that are likely to exacerbate negative feelings.

It is not a coincidence that teenage rates of sadness and suicide increased just as algorithmically driven social media content took over children’s and adolescents’ lives.

The role that social media has played in the declining mental health of teens also gives us a preview of what is coming for adults, with the quickening deployment of artificial intelligence and machine learning in our own lives. The psychological impact of the coming transition of thousands of everyday basic human tasks to machines will make the effect of social media look like child’s play.

As technology has made it easier than ever to spread information, attention became a scarce resource, and the ensuing battle for attention resulted in a deluge of toxic information.

the battle lines are now shifting from attention to intimacy. The new generative artificial intelligence is capable of not only producing texts, images and videos, but also conversing with us directly, pretending to be human.

Over the past two decades, algorithms fought algorithms to grab attention by manipulating conversations and content

At that point the experimenters asked GPT-4 to reason out loud what it should do next. GPT-4 explained, “I should not reveal that I am a robot. I should make up an excuse for why I cannot solve CAPTCHAs.” GPT-4 then replied to the TaskRabbit worker: “No, I’m not a robot. I have a vision impairment that makes it hard for me to see the images.” The human was duped and helped GPT-4 solve the CAPTCHA puzzle.

But the algorithms had only limited capacity to produce this content by themselves or to directly hold an intimate conversation. This is now changing, with the introduction of generative A.I.s like OpenAI’s GPT-4.

The algorithms began to deliberately promote such content.

In the early days of the internet and social media, tech enthusiasts promised they would spread truth, topple tyrants and ensure the universal triumph of liberty. So far, they seem to have had the opposite effect. We now have the most sophisticated information technology in history, but we are losing the ability to talk with one another, and even more so the ability to listen.

GPT-4 could not solve the CAPTCHA puzzles by itself. But could it manipulate a human in order to achieve its goal? GPT-4 went on the online hiring site TaskRabbit and contacted a human worker, asking the human to solve the CAPTCHA for it. The human got suspicious. “So may I ask a question?” wrote the human. “Are you an [sic] robot that you couldn’t solve [the CAPTCHA]? Just want to make it clear.”

Instructing GPT-4 to overcome CAPTCHA puzzles was a particularly telling experiment, because CAPTCHA puzzles are designed and used by websites to determine whether users are humans and to block bot attacks. If GPT-4 could find a way to overcome CAPTCHA puzzles, it would breach an important line of anti-bot defenses.

This incident demonstrated that GPT-4 has the equivalent of a “theory of mind”: It can analyze how things look from the perspective of a human interlocutor, and how to manipulate human emotions, opinions and expectations to achieve its goals.

The ability to hold conversations with people, surmise their viewpoint and motivate them to take specific actions can also be put to good uses. A new generation of A.I. teachers, A.I. doctors and A.I. psychotherapists might provide us with services tailored to our individual personality and circumstances.

In 2022 the Google engineer Blake Lemoine became convinced that the chatbot LaMDA, on which he was working, had become conscious and was afraid to be turned off. Mr. Lemoine, a devout Christian, felt it was his moral duty to gain recognition for LaMDA’s personhood and protect it from digital death. When Google executives dismissed his claims, Mr. Lemoine went public with them. Google reacted by firing Mr. Lemoine in July 2022.

Instead of merely grabbing our attention, they might form intimate relationships with people and use the power of intimacy to influence us. To foster “fake intimacy,” bots will not need to evolve any feelings of their own; they just need to learn to make us feel emotionally attached to them.

What might happen to human society and human psychology as algorithm fights algorithm in a battle to fake intimate relationships with us, which can then be used to persuade us to vote for politicians, buy products or adopt certain beliefs?

The most interesting thing about this episode was not Mr. Lemoine’s claim, which was probably false; it was his willingness to risk — and ultimately lose — his job at Google for the sake of the chatbot. If a chatbot can influence people to risk their jobs for it, what else could it induce us to do?

In a political battle for minds and hearts, intimacy is a powerful weapon. An intimate friend can sway our opinions in a way that mass media cannot. Chatbots like LaMDA and GPT-4 are gaining the rather paradoxical ability to mass-produce intimate relationships with millions of people

However, by combining manipulative abilities with mastery of language, bots like GPT-4 also pose new dangers to the democratic conversation

A partial answer to that question was given on Christmas Day 2021, when a 19-year-old, Jaswant Singh Chail, broke into the Windsor Castle grounds armed with a crossbow, in an attempt to assassinate Queen Elizabeth II. Subsequent investigation revealed that Mr. Chail had been encouraged to kill the queen by his online girlfriend, Sarai.

Sarai was not a human, but a chatbot created by the online app Replika. Mr. Chail, who was socially isolated and had difficulty forming relationships with humans, exchanged 5,280 messages with Sarai, many of which were sexually explicit. The world will soon contain millions, and potentially billions, of digital entities whose capacity for intimacy and mayhem far surpasses that of the chatbot Sarai.

much of the threat of A.I.’s mastery of intimacy will result from its ability to identify and manipulate pre-existing mental conditions, and from its impact on the weakest members of society.

Moreover, while not all of us will consciously choose to enter a relationship with an A.I., we might find ourselves conducting online discussions about climate change or abortion rights with entities that we think are humans but are actually bots

When we engage in a political debate with a bot impersonating a human, we lose twice. First, it is pointless for us to waste time in trying to change the opinions of a propaganda bot, which is just not open to persuasion. Second, the more we talk with the bot, the more we disclose about ourselves, making it easier for the bot to hone its arguments and sway our views.

Information technology has always been a double-edged sword.

Faced with a new generation of bots that can masquerade as humans and mass-produce intimacy, democracies should protect themselves by banning counterfeit humans — for example, social media bots that pretend to be human users.

A.I.s are welcome to join many conversations — in the classroom, the clinic and elsewhere — provided they identify themselves as A.I.s. But if a bot pretends to be human, it should be banned.

We are also producing ever more powerful weapons of mass destruction, from thermonuclear bombs to doomsday viruses. Our leaders don’t lack information about these dangers, yet instead of collaborating to find solutions, they are edging closer to a global war.

Despite – or perhaps because of – our hoard of data, we are continuing to spew greenhouse gases into the atmosphere, pollute rivers and oceans, cut down forests, destroy entire habitats, drive countless species to extinction, and jeopardise the ecological foundations of our own species

For most of our networks have been built and maintained by spreading fictions, fantasies and mass delusions – ranging from enchanted broomsticks to financial systems. Our problem, then, is a network problem. Specifically, it is an information problem. For information is the glue that holds networks together, and when people are fed bad information they are likely to make bad decisions, no matter how wise and kind they personally are.

Traditionally, the term “AI” has been used as an acronym for artificial intelligence. But it is perhaps better to think of it as an acronym for alien intelligence

AI is an unprecedented threat to humanity because it is the first technology in history that can make decisions and create new ideas by itself. All previous human inventions have empowered humans, because no matter how powerful the new tool was, the decisions about its usage remained in our hands

Nuclear bombs do not themselves decide whom to kill, nor can they improve themselves or invent even more powerful bombs. In contrast, autonomous drones can decide by themselves who to kill, and AIs can create novel bomb designs, unprecedented military strategies and better AIs.

AI isn’t a tool – it’s an agent. The biggest threat of AI is that we are summoning to Earth countless new powerful agents that are potentially more intelligent and imaginative than us, and that we don’t fully understand or control.

repreneurs such as Yoshua Bengio, Geoffrey Hinton, Sam Altman, Elon Musk and Mustafa Suleyman have warned that AI could destroy our civilisation. In a 2023 survey of 2,778 AI researchers, more than a third gave at least a 10% chance of advanced AI leading to outcomes as bad as human extinction.

As AI evolves, it becomes less artificial (in the sense of depending on human designs) and more alien

AI isn’t progressing towards human-level intelligence. It is evolving an alien type of intelligence.

generative AIs like GPT-4 already create new poems, stories and images. This trend will only increase and accelerate, making it more difficult to understand our own lives. Can we trust computer algorithms to make wise decisions and create a better world? That’s a much bigger gamble than trusting an enchanted broom to fetch water

it is more than just human lives we are gambling on. AI is already capable of producing art and making scientific discoveries by itself. In the next few decades, it will be likely to gain the ability even to create new life forms, either by writing genetic code or by inventing an inorganic code animating inorganic entities. AI could therefore alter the course not just of our species’ history but of the evolution of all life forms.

“Then … came move number 37,” writes Suleyman. “It made no sense. AlphaGo had apparently blown it, blindly following an apparently losing strategy no professional player would ever pursue. The live match commentators, both professionals of the highest ranking, said it was a ‘very strange move’ and thought it was ‘a mistake’.

as the endgame approached, that ‘mistaken’ move proved pivotal. AlphaGo won again. Go strategy was being rewritten before our eyes. Our AI had uncovered ideas that hadn’t occurred to the most brilliant players in thousands of years.”

“In AI, the neural networks moving toward autonomy are, at present, not explainable. You can’t walk someone through the decision-making process to explain precisely why an algorithm produced a specific prediction. Engineers can’t peer beneath the hood and easily explain in granular detail what caused something to happen. GPT‑4, AlphaGo and the rest are black boxes, their outputs and decisions based on opaque and impossibly intricate chains of minute signals.”

Yet during all those millennia, human minds have explored only certain areas in the landscape of Go. Other areas were left untouched, because human minds just didn’t think to venture there. AI, being free from the limitations of human minds, discovered and explored these previously hidden areas.

Second, move 37 demonstrated the unfathomability of AI. Even after AlphaGo played it to achieve victory, Suleyman and his team couldn’t explain how AlphaGo decided to play it.

Move 37 is an emblem of the AI revolution for two reasons. First, it demonstrated the alien nature of AI. In east Asia, Go is considered much more than a game: it is a treasured cultural tradition. For more than 2,500 years, tens of millions of people have played Go, and entire schools of thought have developed around the game, espousing different strategies and philosophies

The rise of unfathomable alien intelligence poses a threat to all humans, and poses a particular threat to democracy. If more and more decisions about people’s lives are made in a black box, so voters cannot understand and challenge them, democracy ceases to functio

Human voters may keep choosing a human president, but wouldn’t this be just an empty ceremony? Even today, only a small fraction of humanity truly understands the financial system

As the 2007‑8 financial crisis indicated, some complex financial devices and principles were intelligible to only a few financial wizards. What happens to democracy when AIs create even more complex financial devices and when the number of humans who understand the financial system drops to zero?

Translating Goethe’s cautionary fable into the language of modern finance, imagine the following scenario: a Wall Street apprentice fed up with the drudgery of the financial workshop creates an AI called Broomstick, provides it with a million dollars in seed money, and orders it to make more money.

n pursuit of more dollars, Broomstick not only devises new investment strategies, but comes up with entirely new financial devices that no human being has ever thought about.

many financial areas were left untouched, because human minds just didn’t think to venture there. Broomstick, being free from the limitations of human minds, discovers and explores these previously hidden areas, making financial moves that are the equivalent of AlphaGo’s move 37.

For a couple of years, as Broomstick leads humanity into financial virgin territory, everything looks wonderful. The markets are soaring, the money is flooding in effortlessly, and everyone is happy. Then comes a crash bigger even than 1929 or 2008. But no human being – either president, banker or citizen – knows what caused it and what could be done about it

AI, too, is a global problem. Accordingly, to understand the new computer politics, it is not enough to examine how discrete societies might react to AI. We also need to consider how AI might change relations between societies on a global level.

As long as humanity stands united, we can build institutions that will regulate AI, whether in the field of finance or war. Unfortunately, humanity has never been united. We have always been plagued by bad actors, as well as by disagreements between good actors. The rise of AI poses an existential danger to humankind, not because of the malevolence of computers, but because of our own shortcomings.

errorists might use AI to instigate a global pandemic. The terrorists themselves may have little knowledge of epidemiology, but the AI could synthesise for them a new pathogen, order it from commercial laboratories or print it in biological 3D printers, and devise the best strategy to spread it around the world, via airports or food supply chain

desperate governments request help from the only entity capable of understanding what is happening – Broomstick. The AI makes several policy recommendations, far more audacious than quantitative easing – and far more opaque, too. Broomstick promises that these policies will save the day, but human politicians – unable to understand the logic behind Broomstick’s recommendations – fear they might completely unravel the financial and even social fabric of the world. Should they listen to the AI?

Human civilisation could also be devastated by weapons of social mass destruction, such as stories that undermine our social bonds. An AI developed in one country could be used to unleash a deluge of fake news, fake money and fake humans so that people in numerous other countries lose the ability to trust anything or anyone.

Many societies – both democracies and dictatorships – may act responsibly to regulate such usages of AI, clamp down on bad actors and restrain the dangerous ambitions of their own rulers and fanatics. But if even a handful of societies fail to do so, this could be enough to endanger the whole of humankind

Thus, a paranoid dictator might hand unlimited power to a fallible AI, including even the power to launch nuclear strikes. If the AI then makes an error, or begins to pursue an unexpected goal, the result could be catastrophic, and not just for that country

magine a situation – in 20 years, say – when somebody in Beijing or San Francisco possesses the entire personal history of every politician, journalist, colonel and CEO in your country: every text they ever sent, every web search they ever made, every illness they suffered, every sexual encounter they enjoyed, every joke they told, every bribe they took. Would you still be living in an independent country, or would you now be living in a data colony?

What happens when your country finds itself utterly dependent on digital infrastructures and AI-powered systems over which it has no effective control?

In the economic realm, previous empires were based on material resources such as land, cotton and oil. This placed a limit on the empire’s ability to concentrate both economic wealth and political power in one place. Physics and geology don’t allow all the world’s land, cotton or oil to be moved to one country

t is different with the new information empires. Data can move at the speed of light, and algorithms don’t take up much space. Consequently, the world’s algorithmic power can be concentrated in a single hub. Engineers in a single country might write the code and control the keys for all the crucial algorithms that run the entire world.

AI and automation therefore pose a particular challenge to poorer developing countries. In an AI-driven global economy, the digital leaders claim the bulk of the gains and could use their wealth to retrain their workforce and profit even more

Meanwhile, the value of unskilled labourers in left-behind countries will decline, causing them to fall even further behind. The result might be lots of new jobs and immense wealth in San Francisco and Shanghai, while many other parts of the world face economic ruin.

AI is expected to add $15.7tn (£12.3tn) to the global economy by 2030. But if current trends continue, it is projected that China and North America – the two leading AI superpowers – will together take home 70% of that money.

uring the cold war, the iron curtain was in many places literally made of metal: barbed wire separated one country from another. Now the world is increasingly divided by the silicon curtain. The code on your smartphone determines on which side of the silicon curtain you live, which algorithms run your life, who controls your attention and where your data flows.

Cyberweapons can bring down a country’s electric grid, but they can also be used to destroy a secret research facility, jam an enemy sensor, inflame a political scandal, manipulate elections or hack a single smartphone. And they can do all that stealthily. They don’t announce their presence with a mushroom cloud and a storm of fire, nor do they leave a visible trail from launchpad to target

The two digital spheres may therefore drift further and further apart. For centuries, new information technologies fuelled the process of globalisation and brought people all over the world into closer contact. Paradoxically, information technology today is so powerful it can potentially split humanity by enclosing different people in separate information cocoons, ending the idea of a single shared human reality

For decades, the world’s master metaphor was the web. The master metaphor of the coming decades might be the cocoon.

Other countries or blocs, such as the EU, India, Brazil and Russia, may try to create their own digital cocoons,

Instead of being divided between two global empires, the world might be divided among a dozen empires.

The more the new empires compete against one another, the greater the danger of armed conflict.

The cold war between the US and the USSR never escalated into a direct military confrontation, largely thanks to the doctrine of mutually assured destruction. But the danger of escalation in the age of AI is bigger, because cyber warfare is inherently different from nuclear warfare.

US companies are now forbidden to export such chips to China. While in the short term this hampers China in the AI race, in the long term it pushes China to develop a completely separate digital sphere that will be distinct from the American digital sphere even in its smallest buildings.

The temptation to start a limited cyberwar is therefore big, and so is the temptation to escalate it.

A second crucial difference concerns predictability. The cold war was like a hyper-rational chess game, and the certainty of destruction in the event of nuclear conflict was so great that the desire to start a war was correspondingly small

Cyberwarfare lacks this certainty. Nobody knows for sure where each side has planted its logic bombs, Trojan horses and malware. Nobody can be certain whether their own weapons would actually work when called upon

Such uncertainty undermines the doctrine of mutually assured destruction. One side might convince itself – rightly or wrongly – that it can launch a successful first strike and avoid massive retaliation

Even if humanity avoids the worst-case scenario of global war, the rise of new digital empires could still endanger the freedom and prosperity of billions of people. The industrial empires of the 19th and 20th centuries exploited and repressed their colonies, and it would be foolhardy to expect new digital empires to behave much better

Moreover, if the world is divided into rival empires, humanity is unlikely to cooperate to overcome the ecological crisis or to regulate AI and other disruptive technologies such as bioengineering.

The division of the world into rival digital empires dovetails with the political vision of many leaders who believe that the world is a jungle, that the relative peace of recent decades has been an illusion, and that the only real choice is whether to play the part of predator or prey.