technology:recommendation algorithm

  • YouTube Algorithms: How To Avoid the Rabbit Hole - YouTube
    https://www.youtube.com/watch?v=CuFKYSSZtpo&t=4s

    Une vidéo de la série “Above the noise”, de vidéos pédagogiques subventionnée par Stat & Society (think-tank fondé par danah boyd).

    We all know how easy it is to spend hours watching videos on YouTube. Why do we go down that rabbit hole? Mostly because of a combination of computer programming and marketing know-how called ALGORITHMS. *What are algorithms? Algorithms are sets of instructions (created by people) that decide what content you see, and the order it’s listed, when you search online. **How do recommendation algorithms work on YouTube? YouTube’s algorithm captures data about videos you watch, including how long you watch. They recommend other videos based on that viewing history. They optimize advertising by selling this data to companies so they can better target you for their products. * How do these algorithms play a role in spreading misinformation? Digital platforms like YouTube are the gatekeepers of information, whether they intended to be or not. And media-savvy con artists and extremists groups—like conspiracy theorists and white supremacists—can take advantage of YouTube’s algorithms to push their agendas. *What are data voids? Propagandists often take advantage of a lack of content on a certain topic—a “data void.” This happens a lot with breaking news. It takes time for legit media outlets to fact-check and verify events and create content. Until then, Youtube will often only have false or misleading information. * What can I do to avoid falling for misinformation? The more precise you are with search keywords, the more likely you are to get relevant information. If your results look suspicious or click-baity, you might be in a data void. YouTube can be a good place to start your research, but you should use a wider range of sources.

    #YouTube #Algorithmes #Data_Society #Culture_numérique

  • Body politics: The old and new public health risks of networked health misinformation
    https://points.datasociety.net/body-politics-the-old-and-new-public-health-risks-of-networked-h

    There are clear parallels between the tactics used to spread health disinformation and political content. For instance, in 2018, researchers found that large networks of bots and trolls were spreading anti-vaccination rhetoric to sow confusion online and amplify the appearance of an anti-vaccination community. The anti-vaccination tweets often referenced conspiracy theories, and some accounts almost singularly focused on the U.S. government. As a result, real-life users and orchestrated networks of bots are engaged in a feedback loop. Recently, political public figures have used their platform to amplify vaccination misinformation, such as tweeting that measles can help fight cancer. There is a long history of people using influence to sway public opinion about vaccines—particularly among celebrities.

    These are symptoms of a larger societal crisis: disinformation campaigns aimed to undermine social institutions.

    The search and recommendation algorithms that underpin our information retrieval systems are other modern tools mediating access to health information. When a user enters an inquiry into a search engine, they receive curated results. As so many people rely on search engines for health information, they are another important mechanism that is susceptible to manipulation. For instance, the websites of some crisis pregnancy centers—which are designed to look and sound like those of clinics that provide abortion care, but instead give misleading information about the negative effects of abortion to visitors—are optimized results for Google searches often made by women seeking abortion information.

    Similarly, recommendation systems on popular social media platforms, particularly Facebook and YouTube, create easy entry points for problematic content. For example, a mother joining a generic parenting group on Facebook may subsequently receive recommendations for anti-vaxx groups. Bots, search engine optimization, and gaming of recommendation systems are foundational tools used by various actors to influence public health discourse and skew public debates — often blurring the line between medical mistrust and larger political ideologies and agendas.

    #Information_médicale #Santé_publique #Vaccination #Complotisme #Médias_sociaux #Algorithmes

  • YouTube Is Still Struggling To Rein In Its Recommendation Algorithm
    https://www.buzzfeednews.com/article/carolineodonovan/down-youtubes-recommendation-rabbithole

    Despite year-old promises to fix its “Up Next” content recommendation system, YouTube is still suggesting conspiracy videos, hyperpartisan and misogynist videos, pirated videos, and content from hate groups following common news-related searches.

    #youtube #algorithme #recommendation #merde

  • Cheap Words | The New Yorker
    https://www.newyorker.com/magazine/2014/02/17/cheap-words

    Amazon is a global superstore, like Walmart. It’s also a hardware manufacturer, like Apple, and a utility, like Con Edison, and a video distributor, like Netflix, and a book publisher, like Random House, and a production studio, like Paramount, and a literary magazine, like The Paris Review, and a grocery deliverer, like FreshDirect, and someday it might be a package service, like U.P.S. Its founder and chief executive, Jeff Bezos, also owns a major newspaper, the Washington Post. All these streams and tributaries make Amazon something radically new in the history of American business.

    Recently, Amazon even started creating its own “content”—publishing books. The results have been decidedly mixed. A monopoly is dangerous because it concentrates so much economic power, but in the book business the prospect of a single owner of both the means of production and the modes of distribution is especially worrisome: it would give Amazon more control over the exchange of ideas than any company in U.S. history. Even in the iPhone age, books remain central to American intellectual life, and perhaps to democracy. And so the big question is not just whether Amazon is bad for the book industry; it’s whether Amazon is bad for books.

    According to Marcus, Amazon executives considered publishing people “antediluvian losers with rotary phones and inventory systems designed in 1968 and warehouses full of crap.” Publishers kept no data on customers, making their bets on books a matter of instinct rather than metrics. They were full of inefficiences, starting with overpriced Manhattan offices. There was “a general feeling that the New York publishing business was just this cloistered, Gilded Age antique just barely getting by in a sort of Colonial Williamsburg of commerce, but when Amazon waded into this they would show publishing how it was done.”

    During the 1999 holiday season, Amazon tried publishing books, leasing the rights to a defunct imprint called Weathervane and putting out a few titles. “These were not incipient best-sellers,” Marcus writes. “They were creatures from the black lagoon of the remainder table”—Christmas recipes and the like, selected with no apparent thought. Employees with publishing experience, like Fried, were not consulted. Weathervane fell into an oblivion so complete that there’s no trace of it on the Internet. (Representatives at the company today claim never to have heard of it.) Nobody at Amazon seemed to absorb any lessons from the failure. A decade later, the company would try again.

    Around this time, a group called the “personalization team,” or P13N, started to replace editorial suggestions for readers with algorithms that used customers’ history to make recommendations for future purchases. At Amazon, “personalization” meant data analytics and statistical probability. Author interviews became less frequent, and in-house essays were subsumed by customer reviews, which cost the company nothing. Tim Appelo, the entertainment editor at the time, said, “You could be the Platonic ideal of the reviewer, and you would not beat even those rather crude early algorithms.” Amazon’s departments competed with one another almost as fiercely as they did with other companies. According to Brad Stone, a trash-talking sign was hung on a wall in the P13N office: “people forget that john henry died in the end.” Machines defeated human beings.

    In December, 1999, at the height of the dot-com mania, Time named Bezos its Person of the Year. “Amazon isn’t about technology or even commerce,” the breathless cover article announced. “Amazon is, like every other site on the Web, a content play.” Yet this was the moment, Marcus said, when “content” people were “on the way out.” Although the writers and the editors made the site more interesting, and easier to navigate, they didn’t bring more customers.

    The fact that Amazon once devoted significant space on its site to editorial judgments—to thinking and writing—would be an obscure footnote if not for certain turns in the company’s more recent history. According to one insider, around 2008—when the company was selling far more than books, and was making twenty billion dollars a year in revenue, more than the combined sales of all other American bookstores—Amazon began thinking of content as central to its business. Authors started to be considered among the company’s most important customers. By then, Amazon had lost much of the market in selling music and videos to Apple and Netflix, and its relations with publishers were deteriorating. These difficulties offended Bezos’s ideal of “seamless” commerce. “The company despises friction in the marketplace,” the Amazon insider said. “It’s easier for us to sell books and make books happen if we do it our way and not deal with others. It’s a tech-industry thing: ‘We think we can do it better.’ ” If you could control the content, you controlled everything.

    Many publishers had come to regard Amazon as a heavy in khakis and oxford shirts. In its drive for profitability, Amazon did not raise retail prices; it simply squeezed its suppliers harder, much as Walmart had done with manufacturers. Amazon demanded ever-larger co-op fees and better shipping terms; publishers knew that they would stop being favored by the site’s recommendation algorithms if they didn’t comply. Eventually, they all did. (Few customers realize that the results generated by Amazon’s search engine are partly determined by promotional fees.)

    In late 2007, at a press conference in New York, Bezos unveiled the Kindle, a simple, lightweight device that—in a crucial improvement over previous e-readers—could store as many as two hundred books, downloaded from Amazon’s 3G network. Bezos announced that the price of best-sellers and new titles would be nine-ninety-nine, regardless of length or quality—a figure that Bezos, inspired by Apple’s sale of songs on iTunes for ninety-nine cents, basically pulled out of thin air. Amazon had carefully concealed the number from publishers. “We didn’t want to let that cat out of the bag,” Steele said.

    The price was below wholesale in some cases, and so low that it represented a serious threat to the market in twenty-six-dollar hardcovers. Bookstores that depended on hardcover sales—from Barnes & Noble and Borders (which liquidated its business in 2011) to Rainy Day Books in Kansas City—glimpsed their possible doom. If reading went entirely digital, what purpose would they serve? The next year, 2008, which brought the financial crisis, was disastrous for bookstores and publishers alike, with widespread layoffs.

    By 2010, Amazon controlled ninety per cent of the market in digital books—a dominance that almost no company, in any industry, could claim. Its prohibitively low prices warded off competition.

    Publishers looked around for a competitor to Amazon, and they found one in Apple, which was getting ready to introduce the iPad, and the iBooks Store. Apple wanted a deal with each of the Big Six houses (Hachette, HarperCollins, Macmillan, Penguin, Random House, and Simon & Schuster) that would allow the publishers to set the retail price of titles on iBooks, with Apple taking a thirty-per-cent commission on each sale. This was known as the “agency model,” and, in some ways, it offered the publishers a worse deal than selling wholesale to Amazon. But it gave publishers control over pricing and a way to challenge Amazon’s grip on the market. Apple’s terms included the provision that it could match the price of any rival, which induced the publishers to impose the agency model on all digital retailers, including Amazon.

    Five of the Big Six went along with Apple. (Random House was the holdout.) Most of the executives let Amazon know of the change by phone or e-mail, but John Sargent flew out to Seattle to meet with four Amazon executives, including Russ Grandinetti, the vice-president of Kindle content. In an e-mail to a friend, Sargent wrote, “Am on my way out to Seattle to get my ass kicked by Amazon.”

    Sargent’s gesture didn’t seem to matter much to the Amazon executives, who were used to imposing their own terms. Seated at a table in a small conference room, Sargent said that Macmillan wanted to switch to the agency model for e-books, and that if Amazon refused Macmillan would withhold digital editions until seven months after print publication. The discussion was angry and brief. After twenty minutes, Grandinetti escorted Sargent out of the building. The next day, Amazon removed the buy buttons from Macmillan’s print and digital titles on its site, only to restore them a week later, under heavy criticism. Amazon unwillingly accepted the agency model, and within a couple of months e-books were selling for as much as fourteen dollars and ninety-nine cents.

    Amazon filed a complaint with the Federal Trade Commission. In April, 2012, the Justice Department sued Apple and the five publishers for conspiring to raise prices and restrain competition. Eventually, all the publishers settled with the government. (Macmillan was the last, after Sargent learned that potential damages could far exceed the equity value of the company.) Macmillan was obliged to pay twenty million dollars, and Penguin seventy-five million—enormous sums in a business that has always struggled to maintain respectable profit margins.

    Apple fought the charges, and the case went to trial last June. Grandinetti, Sargent, and others testified in the federal courthouse in lower Manhattan. As proof of collusion, the government presented evidence of e-mails, phone calls, and dinners among the Big Six publishers during their negotiations with Apple. Sargent and other executives acknowledged that they wanted higher prices for e-books, but they argued that the evidence showed them only to be competitors in an incestuous business, not conspirators. On July 10th, Judge Denise Cote ruled in the government’s favor.

    Apple, facing up to eight hundred and forty million dollars in damages, has appealed. As Apple and the publishers see it, the ruling ignored the context of the case: when the key events occurred, Amazon effectively had a monopoly in digital books and was selling them so cheaply that it resembled predatory pricing—a barrier to entry for potential competitors. Since then, Amazon’s share of the e-book market has dropped, levelling off at about sixty-five per cent, with the rest going largely to Apple and to Barnes & Noble, which sells the Nook e-reader. In other words, before the feds stepped in, the agency model introduced competition to the market. But the court’s decision reflected a trend in legal thinking among liberals and conservatives alike, going back to the seventies, that looks at antitrust cases from the perspective of consumers, not producers: what matters is lowering prices, even if that goal comes at the expense of competition.

    With Amazon’s patented 1-Click shopping, which already knows your address and credit-card information, there’s just you and the buy button; transactions are as quick and thoughtless as scratching an itch. “It’s sort of a masturbatory culture,” the marketing executive said. If you pay seventy-nine dollars annually to become an Amazon Prime member, a box with the Amazon smile appears at your door two days after you click, with free shipping. Amazon’s next frontier is same-day delivery: first in certain American cities, then throughout the U.S., then the world. In December, the company patented “anticipatory shipping,” which will use your shopping data to put items that you don’t yet know you want to buy, but will soon enough, on a truck or in a warehouse near you.

    Amazon employs or subcontracts tens of thousands of warehouse workers, with seasonal variation, often building its fulfillment centers in areas with high unemployment and low wages. Accounts from inside the centers describe the work of picking, boxing, and shipping books and dog food and beard trimmers as a high-tech version of the dehumanized factory floor satirized in Chaplin’s “Modern Times.” Pickers holding computerized handsets are perpetually timed and measured as they fast-walk up to eleven miles per shift around a million-square-foot warehouse, expected to collect orders in as little as thirty-three seconds. After watching footage taken by an undercover BBC reporter, a stress expert said, “The evidence shows increased risk of mental illness and physical illness.” The company says that its warehouse jobs are “similar to jobs in many other industries.”

    When I spoke with Grandinetti, he expressed sympathy for publishers faced with upheaval. “The move to people reading digitally and buying books digitally is the single biggest change that any of us in the book business will experience in our time,” he said. “Because the change is particularly big in size, and because we happen to be a leader in making it, a lot of that fear gets projected onto us.” Bezos also argues that Amazon’s role is simply to usher in inevitable change. After giving “60 Minutes” a first glimpse of Amazon drone delivery, Bezos told Charlie Rose, “Amazon is not happening to bookselling. The future is happening to bookselling.”

    In Grandinetti’s view, the Kindle “has helped the book business make a more orderly transition to a mixed print and digital world than perhaps any other medium.” Compared with people who work in music, movies, and newspapers, he said, authors are well positioned to thrive. The old print world of scarcity—with a limited number of publishers and editors selecting which manuscripts to publish, and a limited number of bookstores selecting which titles to carry—is yielding to a world of digital abundance. Grandinetti told me that, in these new circumstances, a publisher’s job “is to build a megaphone.”

    After the Kindle came out, the company established Amazon Publishing, which is now a profitable empire of digital works: in addition to Kindle Singles, it has mystery, thriller, romance, and Christian lines; it publishes translations and reprints; it has a self-service fan-fiction platform; and it offers an extremely popular self-publishing platform. Authors become Amazon partners, earning up to seventy per cent in royalties, as opposed to the fifteen per cent that authors typically make on hardcovers. Bezos touts the biggest successes, such as Theresa Ragan, whose self-published thrillers and romances have been downloaded hundreds of thousands of times. But one survey found that half of all self-published authors make less than five hundred dollars a year.

    Every year, Fine distributes grants of twenty-five thousand dollars, on average, to dozens of hard-up literary organizations. Beneficiaries include the pen American Center, the Loft Literary Center, in Minneapolis, and the magazine Poets & Writers. “For Amazon, it’s the cost of doing business, like criminal penalties for banks,” the arts manager said, suggesting that the money keeps potential critics quiet. Like liberal Democrats taking Wall Street campaign contributions, the nonprofits don’t advertise the grants. When the Best Translated Book Award received money from Amazon, Dennis Johnson, of Melville House, which had received the prize that year, announced that his firm would no longer compete for it. “Every translator in America wrote me saying I was a son of a bitch,” Johnson said. A few nonprofit heads privately told him, “I wanted to speak out, but I might have taken four thousand dollars from them, too.” A year later, at the Associated Writing Programs conference, Fine shook Johnson’s hand, saying, “I just wanted to thank you—that was the best publicity we could have had.” (Fine denies this.)

    By producing its own original work, Amazon can sell more devices and sign up more Prime members—a major source of revenue. While the company was building the Kindle, it started a digital store for streaming music and videos, and, around the same time it launched Amazon Publishing, it created Amazon Studios.

    The division pursued an unusual way of producing television series, using its strength in data collection. Amazon invited writers to submit scripts on its Web site—“an open platform for content creators,” as Bill Carr, the vice-president for digital music and video, put it. Five thousand scripts poured in, and Amazon chose to develop fourteen into pilots. Last spring, Amazon put the pilots on its site, where customers could review them and answer a detailed questionnaire. (“Please rate the following aspects of this show: The humor, the characters . . . ”) More than a million customers watched. Engineers also developed software, called Amazon Storyteller, which scriptwriters can use to create a “storyboard animatic”—a cartoon rendition of a script’s plot—allowing pilots to be visualized without the expense of filming. The difficulty, according to Carr, is to “get the right feedback and the right data, and, of the many, many data points that I can collect from customers, which ones can tell you, ‘This is the one’?”

    Bezos applying his “take no prisoners” pragmatism to the Post: “There are conflicts of interest with Amazon’s many contracts with the government, and he’s got so many policy issues going, like sales tax.” One ex-employee who worked closely with Bezos warned, “At Amazon, drawing a distinction between content people and business people is a foreign concept.”

    Perhaps buying the Post was meant to be a good civic deed. Bezos has a family foundation, but he has hardly involved himself in philanthropy. In 2010, Charlie Rose asked him what he thought of Bill Gates’s challenge to other billionaires to give away most of their wealth. Bezos didn’t answer. Instead, he launched into a monologue on the virtue of markets in solving social problems, and somehow ended up touting the Kindle.

    Bezos bought a newspaper for much the same reason that he has invested money in a project for commercial space travel: the intellectual challenge. With the Post, the challenge is to turn around a money-losing enterprise in a damaged industry, and perhaps to show a way for newspapers to thrive again.

    Lately, digital titles have levelled off at about thirty per cent of book sales. Whatever the temporary fluctuations in publishers’ profits, the long-term outlook is discouraging. This is partly because Americans don’t read as many books as they used to—they are too busy doing other things with their devices—but also because of the relentless downward pressure on prices that Amazon enforces. The digital market is awash with millions of barely edited titles, most of it dreck, while readers are being conditioned to think that books are worth as little as a sandwich. “Amazon has successfully fostered the idea that a book is a thing of minimal value,” Johnson said. “It’s a widget.”

    There are two ways to think about this. Amazon believes that its approach encourages ever more people to tell their stories to ever more people, and turns writers into entrepreneurs; the price per unit might be cheap, but the higher number of units sold, and the accompanying royalties, will make authors wealthier. Jane Friedman, of Open Road, is unfazed by the prospect that Amazon might destroy the old model of publishing. “They are practicing the American Dream—competition is good!” she told me. Publishers, meanwhile, “have been banks for authors. Advances have been very high.” In Friedman’s view, selling digital books at low prices will democratize reading: “What do you want as an author—to sell books to as few people as possible for as much as possible, or for as little as possible to as many readers as possible?”

    The answer seems self-evident, but there is a more skeptical view. Several editors, agents, and authors told me that the money for serious fiction and nonfiction has eroded dramatically in recent years; advances on mid-list titles—books that are expected to sell modestly but whose quality gives them a strong chance of enduring—have declined by a quarter.

    #Amazon

  • Quantifying Biases in Online Information Exposure | Center for Complex Networks and Systems Research, Indiana University
    https://arxiv.org/abs/1807.06958
    https://arxiv.org/pdf/1807.06958.pdf

    Our consumption of online #information is mediated by filtering, ranking, and recommendation algorithms that introduce unintentional biases as they attempt to deliver relevant and engaging content. It has been suggested that our reliance on online technologies such as search engines and social media may limit exposure to diverse points of view and make us vulnerable to manipulation by disinformation. In this paper, we mine a massive dataset of Web traffic to quantify two kinds of bias: (i) homogeneity bias, which is the tendency to consume content from a narrow set of information sources, and (ii) popularity bias, which is the selective exposure to content from top sites. Our analysis reveals different bias levels across several widely used Web platforms. Search exposes users to a diverse set of sources, while social media traffic tends to exhibit high popularity and homogeneity #bias. When we focus our analysis on traffic to news sites, we find higher levels of popularity bias, with smaller differences across applications. Overall, our results quantify the extent to which our choices of online systems confine us inside “social bubbles.”

    #personnalisation #médias_sociaux #algorithme via @pomeranian99

  • YouTube, the Great Radicalizer - The New York Times
    https://www.nytimes.com/2018/03/10/opinion/sunday/youtube-politics-radical.html

    Par Zeynep Tufekci

    It seems as if you are never “hard core” enough for YouTube’s recommendation algorithm. It promotes, recommends and disseminates videos in a manner that appears to constantly up the stakes. Given its billion or so users, YouTube may be one of the most powerful radicalizing instruments of the 21st century.

    This is not because a cabal of YouTube engineers is plotting to drive the world off a cliff. A more likely explanation has to do with the nexus of artificial intelligence and Google’s business model. (YouTube is owned by Google.) For all its lofty rhetoric, Google is an advertising broker, selling our attention to companies that will pay for it. The longer people stay on YouTube, the more money Google makes.

    What keeps people glued to YouTube? Its algorithm seems to have concluded that people are drawn to content that is more extreme than what they started with — or to incendiary content in general.

    Is this suspicion correct? Good data is hard to come by; Google is loath to share information with independent researchers. But we now have the first inklings of confirmation, thanks in part to a former Google engineer named Guillaume Chaslot.

    It is also possible that YouTube’s recommender algorithm has a bias toward inflammatory content. In the run-up to the 2016 election, Mr. Chaslot created a program to keep track of YouTube’s most recommended videos as well as its patterns of recommendations. He discovered that whether you started with a pro-Clinton or pro-Trump video on YouTube, you were many times more likely to end up with a pro-Trump video recommended.

    Combine this finding with other research showing that during the 2016 campaign, fake news, which tends toward the outrageous, included much more pro-Trump than pro-Clinton content, and YouTube’s tendency toward the incendiary seems evident.

    YouTube has recently come under fire for recommending videos promoting the conspiracy theory that the outspoken survivors of the school shooting in Parkland, Fla., are “crisis actors” masquerading as victims. Jonathan Albright, a researcher at Columbia, recently “seeded” a YouTube account with a search for “crisis actor” and found that following the “up next” recommendations led to a network of some 9,000 videos promoting that and related conspiracy theories, including the claim that the 2012 school shooting in Newtown, Conn., was a hoax.

    What we are witnessing is the computational exploitation of a natural human desire: to look “behind the curtain,” to dig deeper into something that engages us. As we click and click, we are carried along by the exciting sensation of uncovering more secrets and deeper truths. YouTube leads viewers down a rabbit hole of extremism, while Google racks up the ad sales.

    #Zeynep_Tufekci #Google #YouTube #Radicalisation #Pouvoir_algorithmes #Politique_algorithmes

  • ’Fiction is outperforming reality’: how YouTube’s algorithm distorts truth | The Guardian
    https://www.theguardian.com/technology/2018/feb/02/how-youtubes-algorithm-distorts-truth

    Fascinant, fascisant.

    During the three years he worked at Google, he was placed for several months with a team of YouTube engineers working on the recommendation system. The experience led him to conclude that the priorities YouTube gives its algorithms are dangerously skewed.

    “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.”

    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.

    #GAFA #YouTube

  • ’Fiction is outperforming reality’: how YouTube’s algorithm distorts truth | Technology | The Guardian
    https://www.theguardian.com/technology/2018/feb/02/how-youtubes-algorithm-distorts-truth

    There are 1.5 billion YouTube users in the world, which is more than the number of households that own televisions. What they watch is shaped by this algorithm, which skims and ranks billions of videos to identify 20 “up next” clips that are both relevant to a previous video and most likely, statistically speaking, to keep a person hooked on their screen.

    Company insiders tell me the algorithm is the single most important engine of YouTube’s growth. In one of the few public explanations of how the formula works – an academic paper that sketches the algorithm’s deep neural networks, crunching a vast pool of data about videos and the people who watch them – YouTube engineers describe it as one of the “largest scale and most sophisticated industrial recommendation systems in existence”.

    Lewd and violent videos have been algorithmically served up to toddlers watching YouTube Kids, a dedicated app for children. 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.”

    During the three years he worked at Google, he was placed for several months with a team of YouTube engineers working on the recommendation system. The experience led him to conclude that the priorities YouTube gives its algorithms are dangerously skewed.

    “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.”

    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.”

    The software Chaslot wrote was designed to provide the world’s first window into YouTube’s opaque recommendation engine. The program simulates the behaviour of a user who starts on one video and then follows the chain of recommended videos – much as I did after watching the Logan Paul video – tracking data along the way.

    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.

    Over the last 18 months, Chaslot has used the program to explore bias in YouTube content promoted during the French, British and German elections, global warming and mass shootings, and published his findings on his website, Algotransparency.com. Each study finds something different, but the research suggests YouTube systematically amplifies videos that are divisive, sensational and conspiratorial.

    It was not a comprehensive set of videos and it may not have been a perfectly representative sample. But it was, Chaslot said, a previously unseen dataset of what YouTube was recommending to people interested in content about the candidates – one snapshot, in other words, of the algorithm’s preferences.

    Jonathan Albright, research director at the Tow Center for Digital Journalism, who reviewed the code used by Chaslot, says it is a relatively straightforward piece of software and a reputable methodology. “This research captured the apparent direction of YouTube’s political ecosystem,” he says. “That has not been done before.”

    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.

    Some of the videos YouTube was recommending were the sort we had expected to see: broadcasts of presidential debates, TV news clips, Saturday Night Live sketches. There were also videos of speeches by the two candidates – although, we found, the database contained far more YouTube-recommended speeches by Trump than Clinton.

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

    Tufekci, the sociologist who several months ago warned about the impact YouTube may have had on the election, tells me 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.”

    But why would a bias toward ever more weird or divisive videos benefit one candidate over another? That depends on the candidates. Trump’s campaign was nothing if not weird and divisive. Tufekci points to studies showing that “field of misinformation” largely tilted anti-Clinton before the election. “Fake news providers,” she says, “found that fake anti-Clinton material played much better with the pro-Trump base than did fake anti-Trump material with the pro-Clinton base.”

    She adds: “The question before us is the ethics of leading people down hateful rabbit holes full of misinformation and lies at scale just because it works to increase the time people spend on the site – and it does work.”

    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 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.

    Franchi is a professional who makes a living from his channel, but 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.

    #YouTube #Algorithme_recommendation #Politique_USA #Elections #Fake_news

  • ’Fiction is outperforming reality’ : how YouTube’s algorithm distorts truth
    https://www.theguardian.com/technology/2018/feb/02/how-youtubes-algorithm-distorts-truth

    An ex-YouTube insider reveals how its recommendation algorithm promotes divisive clips and conspiracy videos. Did they harm Hillary Clinton’s bid for the presidency ? It was one of January’s most viral videos. Logan Paul, a YouTube celebrity, stumbles across a dead man hanging from a tree. The 22-year-old, who is in a Japanese forest famous as a suicide spot, is visibly shocked, then amused. “Dude, his hands are purple,” he says, before turning to his friends and giggling. “You never stand (...)

    #YouTube #algorithme #manipulation