Skip to main content

3. Manipulated: “Which Ice Cube Is the Best?"

User comments, ratings, and reviews are valuable because, in economic terms, they address the marketplace problem of information asymmetry.

Published onApr 08, 2019
3. Manipulated: “Which Ice Cube Is the Best?"

So I’ve got all these apps on my phone to let me rate restaurants, hotels, food. But why have all these apps when I can have just one that lets me rate everything?

Take for example this glass of water, which ice cube is the best? According to Jotly: that one, it’s melting the slowest. Helpful! …

It just makes sense. With the Jotly you can rate anything. You don’t need all of those other apps. I mean they all got pretty low ratings anyway. So just delete them.

—Promo Video for Jotly

In a 2011 video for the fictional Jotly app, an enthusiastic user descends a playground slide, and the app informs him that “5 others have slid here. 10% go faster than you. Share on Twitter!” With a frown, he comments that “this slide is not very exciting” and gives it a “D.”1 This video parodied a proliferation of rating services and applications. In fact, Jotly appeared at roughly the same time as the genuine rate-everything apps Oink and Stamped. Yet, while these were soon discontinued, an actual Jotly app was released with its sardonic edge intact: “Your life is exciting and worth sharing: everything with everyone! Everyone cares about everything you do. Now you can rate your entire life and share the experience.”2

These apps hint at the immense value of users’ online comments, including reviews and ratings. In 2013, Amazon purchased Goodreads, the book review and discussion site, for $150 million. Within four years of its founding, TripAdvisor, a travel review site, was acquired by Expedia for $200 million. This 2004 success paled in comparison to TripAdvisor’s 2011 initial public offering as a separate company valued at over $3 billion. Google purchased the restaurant rating guide Zagat in 2011 for $150 million. However, Zagat may have been Google’s second choice. It is rumored Google offered Yelp $500 million but Yelp walked away, perhaps embittered by instances of Yelp reviews appearing without attribution in Google Places. In 2012, Google continued to feed its appetite for reviews by purchasing Frommer’s Travel for $22 million. In 2013, Google sold the Frommer’s publishing brand back to its octogenarian founder Arthur Frommer: it had the data that it wanted and few plans to do anything further with the brand.3

User comments, ratings, and reviews are valuable because, in economic terms, they address the marketplace problem of information asymmetry. The classic example is a potential buyer’s fear that a car might be a “lemon.” The seller can signal quality by offering a warranty, noting resale value, or (if it is available) providing the car’s maintenance history and mileage. Because some of these things can be manipulated, such as turning back a used car’s odometer, additional sources of opinion are sought, including informal word of mouth (does your friend like the one she bought?), traditional reviews (what did Consumer Reports conclude?), or online reviews (what do users on Yelp say?).

The information asymmetry is especially problematic when we buy things sight unseen. In the late 1990s, the main concern that people had with online retail was the trustworthiness of the merchant: was the retailer just a scammer with a Webpage? In response, online merchants decorated their pages with assurance seals that vouched for their legitimacy. But this told consumers little about the quality of a particular product. One way to deal with concerns about quality is to have a friendly return policy. Because online shoe stores deal with the fact that buyers never know how well the shoe fits until worn (researchers call this an “experience product”), Zappos (and others) provide convenient and free returns. Some Zappos devotees think nothing of buying three different sizes of a shoe and returning the two that do not fit.

It would be ideal if people could avoid making poor purchases in the first place. One approach is for the merchant to select the best products, such as the stylist who curates Shoe Dazzle’s collection. The opposite approach is to sell everything, and let consumers sort it out. Zappos customers review and rate its ninety thousand styles of shoes. Amazon, which acquired Zappos in 2009, sells billions of products to its over 200 million customers. In 2000, Jeff Bezos, founder of Amazon, said, “We want to make every book available—the good, the bad, and the ugly. When you’re doing that, you actually have an obligation—for going to make the shopping environment one that’s actually conducive to shopping—to let truth loose. That’s what we try to do with the customer reviews.”4 Today Amazon sells more than books (in 2010, media sales were surpassed by sales of other products), but the philosophy of letting “truth loose” remains the same.

Amazon and others have “let truth loose” with user comment, but that truth is being overtaken by fakery and manipulation. TripAdvisor’s extraordinary growth in value has been accompanied by increasing stories of abuse: of proprietors who write their own positive reviews, slam competitors, or pay others to do so. In 2013, a hotel chain executive was writing reviews under a pseudonym, but when he did so using TripAdvisor’s Facebook app, his real name, photo, and location were exposed publicly. Was this fakery the exception or the norm? Who is writing false reviews and why? And what is being done to counteract it? In this chapter, I address these questions and argue that replacing strangers’ recommendations with friends’ endorsements will not solve the problem. Sites that take advantage of our social networks for commercial interests hasten a world in which we are all increasingly tempted to take advantage of each other.

Research on Reviews

Marketing consultants have issued dozens of reports that document consumers’ reliance on online user reviews—and these firms have a marketing service they want to sell you. Fortunately, the nonpartisan and nonprofit Pew Research Center offers some compelling findings. In a 2010 telephone survey of 3,001 adults, 58% reported that they have gone online to research products and services, and 24% said that they have posted comments or reviews about their purchases. (Among Internet users, this is 78% and 32%, respectively.) In fact, 21% of respondents report they had researched a product the day before. In a 2011 Pew survey, 55% of adults said that they seek news and information about local restaurants, bars, and clubs, and 51% of those turn to the Internet for information.5

This use of reviews by consumers affects a merchant’s bottom line. In one early study of eBay, researchers offered identical vintage postcards as a reputable merchant and as a newcomer, and the established merchant’s postcards sold for 7.6% more than the newcomer’s. More recently, a study of digital cameras listed on the price-comparison site NexTag found that a camera’s poor rating was associated with a greater discount from the manufacturer’s suggested retail price. Another early study compared how book reviews at Amazon and Barnes & Noble affected sales rankings. Because it can be unclear whether better reviews lead to better sales or if both are simply the result of a better product, using the same product on two different sites helps researchers to focus on the effects of the reviews. In this study, the conclusion was that better reviews were associated with better sales.6 When it comes to dining, clever researchers noted that Yelp rounds off the overall star rating to the nearest half star. So a restaurant with an average of 3.24 is rounded to 3.0 stars, and one with an average of 3.26 is rounded to 3.5 stars. One researcher used discontinuities around these thresholds to estimate that a one-star increase leads to 5% to 9% more revenue. Other researchers looked at how often a restaurant’s tables sold out during the busiest dining hours (using a restaurant reservation site’s database) and found that crossing the 3.5-star threshold increased a restaurant’s chance of filling all of its tables by another nineteen percentage points. Good ratings and reviews have also been shown to increase the sales of hotel rooms and video games. And even when the valence of a review (that is, whether it is positive or negative) does not drive sales, such as in one study of movie ticket sales, the number of comments does: increased awareness seemingly drives greater ticket sales.7

There are hundreds of similar research reports on user ratings and reviews. Scholars study the effects of reviews relative to their valence, variance, volume, and helpfulness. They also analyze reviewer characteristics (familiarity and expertise), reader demographics, and the categories and features of the products. From all of this, some high-level conclusions can be drawn. First, ratings and reviews tend to be more significant for niche or relatively unknown products. Second, reviews tend to follow a “J-shape” distribution because people are most likely to participate when they have a strong positive or negative experience. Imagine a very wide J with some negative ratings on the left, a dip in moderate ratings in the middle, followed by a peak of positive ratings on the right. Third, while negative reviews do seem to have a greater impact on purchasing decisions, a few negative ratings are not necessarily detrimental: they might prompt consumers to more carefully consider the reviews or have greater confidence when they can see the balance of opinion. Finally, the sentiments expressed in the text of a review are often more persuasive than its corresponding rating.8

This last finding is likely related to the fact that people are more suspect of a numerical rating than a textual review. But people are fooled by text as well; some realized the extent of this problem only when they noticed duplicated reviews. Such was the case for Trevor Pinch, a sociology professor at Cornell University, who noted that reviews of a book that he coauthored about the history of electronic music were similar to those of another book. Intrigued by the incident, Pinch and his colleague Shay David analyzed fifty thousand user reviews of over ten thousand books and CDs and identified three likely motivations. First, different reviewers copied positive reviews from other items to their own to promote sales. Second, the same author might post the same review on multiple items to “promote a specific product, agenda, or opinion.” For instance, a creationist might post the same critical review across multiple books on evolution. She might do this from the same account or through multiple accounts, known as sockpuppets. Finally, reviewers might plagiarize their own reviews across products to increase their review count and credibility (although, as I discuss below, some top reviewers’ numbers lead to incredulity). This analysis allowed David and Pinch to conclude “that about one percent of all review data is duplicated, verbatim or with variations.”9

The use of duplicate reviews also has given researchers an interesting tool for understanding review manipulation. Computer scientists Nitin Jindal and Bing Liu performed an analysis of 5.8 million reviews from 2.14 million reviewers. Because they knew that duplicates were fake, they used these reviews to train a computer model on what fake reviews look like. They asked a computer to use duplicates to learn how to characterize fakes across thirty-five different factors, including the number of comments on a review, a review’s rated helpfulness, the length of a review’s title and text, and the ratio of positive to negative words (great versus horrible). They also included whether “a bad review was written just after the first good review of the product and vice versa” (manipulators often follow an authentic review with its opposite to maintain a product’s rating). They found that fakery was widespread and that helpfulness ratings were of little help in discerning it. In a 2012 New York Times story about “The Best Book Reviews Money Can Buy,” Liu was quoted as saying that “about one-third of all consumer reviews on the Internet are fake.”10 Nan Hu, a business school professor, has also focused on how a positive review often follows a negative one. Hu and his colleagues studied 610,713 reviews of 4,490 books (that each had at least thirty reviews) and found that the probability of seeing a positive review after a negative one was 72%, almost 2.6 times the probability of a negative review following a negative review.11 Clearly, this is suspect, leading them to conclude that 10.3% of the products in their study were subject to manipulation. Researchers who analyzed Boston restaurant reviews found that 16% of the reviews were filtered as possibly fraudulent by Yelp, and these reviews tended to hold more extreme views, are written by reviewers with less established reputations, and focused on restaurants that were facing increased competition.12 Fake reviews are common, and many expect that they will become more so.

Of Spam, Fakes, and Sockpuppets

Online comments often exhibit what linguist Donna Gibbs calls cyberlanguage, “with its own brand of quirky logic, which evolves with unprecedented speed and variety and is heavily dependent on ingenuity and humor.”13 An example of this is the word spam (much to the displeasure of the Hormel Foods Corporation, maker of the canned-meat product of the same name). The origin of the online usage is a Monty Python television sketch in which Spam was inescapably (and frustratingly) included in nearly every dish on a restaurant menu. So spam became a word for unwanted but inescapable messages. It has also been (inappropriately) used to describe fake ratings and reviews. Both unwanted email and fake reviews exploit vulnerable computers in similar ways and are contracted for on illicit markets, but they are not the same. Spam might be considered a fake ham, but few people would confuse the two. In the Monty Python sketch, it is clear that restaurant customers will be getting “Spam bacon sausage and Spam.” In South Korea, in fact, Spam is not poor man’s ham but an expensive delicacy. Accordingly, I prefer to speak of unwanted (but possibly truthful) messages as spam and manipulated reviews as fakes.

Like the word spam, the term sockpuppet entered Internet vernacular in the mid-1990s to describe people who create accounts so that they can masquerade as others. These identities can praise or agree with their creator, much as an old sock on my hand can easily be made to agree with me. Strawman puppets also can be used to take contrary but poorly argued positions. Taking an example from history, this latter tactic is the likely source of Galileo’s troubles with Pope Urban VIII. The seventeenth-century scientist and the Roman Catholic pontiff had begun their relationship amiably, meeting a handful of times during which they spoke of the “two chief world systems”—the Ptolemaic (geocentric) and the Copernican (heliocentric). Urban expressed interest in Galileo’s work and welcomed a presentation of the relative merits of both the earth- and sun-centered theories. Galileo did this in his Dialogues, a genre that has been used since the ancient Greeks in which characters discuss opposing theories. However, Galileo went beyond an impartial consideration and presented a forceful argument for heliocentrism and included the Pope’s words in the mouth of “Simplicio,” a geocentric simpleton. In the online vernacular, Simplicio was Galileo’s strawman sockpuppet, which naturally upset the Pope.

Galileo was not the first or last to create a persona to further his own aims. Yet Galileo, like many authors who have written pseudonymously, was not trying to deceive. The practice of writing fake reviews for one’s own benefit seems to have bloomed with the rise of the genre of the review itself. As mentioned in an earlier chapter, eighteenth-century literary and scientific authors were making their living and reputations by competing for the limited money and attention of their readers, both in original works and via reviews of others’ original works. Botanists Johan Andreas Murray and Carl Linnaeus wrote fake reviews, with the latter concluding that his Systema Naturae was “a masterpiece that can never be read and admired enough.”14 Scientists continue the practice today, creating sockpuppet personas whom they recommend to journal editors as qualified reviewers. On the literary side, Walter Scott was known to review his own work and in 1808 wrote an (anonymous) review of Great Britain’s great living poets, including himself as “the minion of modern popularity; for the works of no living, and of few dead authors, have been so widely and so rapidly diffused.… The effect of this extensive popularity has been almost ludicrous.”15 Walt Whitman is known to have falsely boasted about the success of Leaves of Grass (1855) and in an anonymous review predicted that “he is to prove either the most lamentable of failures or the most glorious of triumphs, in the known history of literature.”16 Anthony Burgess, author of A Clockwork Orange, confessed to his own manipulations, noting that “There is something to be said for allowing a novelist to notice his own novel: he knows its faults better than any casual reader, and he has at least read the book.”17

Putting a Face to Fakers

Researchers estimate that between 10 to 30 percent of online reviews are fake. The cast of manipulators includes fakers (those who deceptively praise their own works or pillory others’), makers (those who will do so for a fee), and the takers (those who avail themselves of such services).

There are plenty of fakers among authors, especially among amateur and aspiring writers. But even well regarded and successful authors are tempted to buttress their standing and bulldoze others’. In 2004, John Rechy, a prolific novelist with a fair amount of recognition, was found to have granted himself (anonymously) a five-star review on Amazon. The New York Times reported that his identity was exposed when a glitch at Amazon revealed its reviewers’ identities. Although Amazon discontinued anonymous reviews, the use of pseudonyms continues. Rechy justified his actions as payback for unfavorable reviews, which is consistent with research that shows that a positive review often follows a negative one: “That anybody is allowed to come in and anonymously trash a book to me is absurd. How to strike back? Just go in and rebut every single one of them.”18 British writers of thrillers seem especially prone to talking up their work. Stephen Leather confessed that he used sockpuppets to create a “buzz,” and R. J. Ellory was found to be writing rave reviews of his own works and attacking others via a pseudonym.

Even academics, who trade in reputation rather than royalties, have been exposed for fakery. When the popular and award-winning British historian Orlando Figes was exposed for trashing his rivals’ books, he first claimed that he had been set up, then blamed his wife, and finally apologized.19 In the same year, a review of historian Simon Winder’s Germania was removed from Amazon as an unfair personal attack. When fellow historian Diane Purkiss was revealed to be the reviewer, she posted a new review under her name and explained the circumstances of her first review:

I posted a somewhat critical review of this book about two weeks ago, and the author, Mr Simon Winder, has had it taken down on the grounds that I am an academic using a pseudonym. That’s true. I am, but not out of any malfeasance—solely because I would not want to mislead readers into thinking that I as an Oxford don have any high-level expertise in the subject matter of books I read as a general reader.… This was my response not as an expert, but as a reader. As a reader I was bored and irritated.20

Winder responded to the review in a comment, claiming her original review “impugned me professionally”:

Of course I have no problem with bad reviews. I asked for the review to be removed because I didn’t fancy an anonymous accusation like that hanging around online. Her review in its current form is fine and I am grateful to her for making it clear that she was slurring me in a private capacity under a false name, rather than as an Oxford academic.21

Besides the financial considerations, what seems to set many on the path of fakerhood is the lack of control over their public identity and the power resulting from manipulating others’. Scott Adams, creator of the popular Dilbert comic series, is known for his provocative blog posts. In one, he wrote that men’s best strategy for dealing with women is to treat them like children. Although he said that he was not equating women with children, he also said that “You don’t argue with a four-year old about why he shouldn’t eat candy for dinner. You don’t punch a mentally handicapped guy even if he punches you first. And you don’t argue when a women tells you she’s only making 80 cents to your dollar. It’s the path of least resistance. You save your energy for more important battles.” Much discussion and criticism followed, though he was not without his defenders, including the user “plannedchaos.” On Reddit and MetaFilter, this user attacked those who were critical of Adams and repeatedly noted that Adams “has a certified genius I.Q.” When the account was exposed as Adams’s puppet, he wrote on MetaFilter that “I’m sorry I peed in your cesspool. For what it’s worth, the smart people were on to me after the first post. That made it funnier.”22 Adams later justified his use of the pseudonym as permitting his speech to be judged on its merits and not harm his business. I do not condone either Purkiss’s or Adams’s use of pseudonyms, but their examples arise from the blurred boundaries of identity and place, an issue that is examined in subsequent chapters.

Some are contrite when they are exposed and follow the incident with a confession. Many fakers begin by noting that they started by wanting to set something right and then turned to self-puffery and attacks on others. In his last piece for The Independent, journalist Johann Hari wrote “a personal apology” for taking liberties with the quotations of his sources and for inappropriate edits to Wikipedia. He wrote that “several years ago I started to notice some things I didn’t like in the Wikipedia entry about me, so I took them out.” This was followed by removing “nasty passages about people I admire” and factually correcting other biographies: “But in a few instances, I edited the entries of people I had clashed with in ways that were juvenile or malicious: I called one of them anti-Semitic and homophobic, and the other a drunk.”23 Such attacks on biographies are a significant source for the dozens (sometimes hundreds) of sockpuppet investigation that are requested every week.

Fakery is not limited to commercial motives or authors. Fake reviews can be used for ideological purposes, such as to censor a viewpoint or laud a politician. Fans of Michael Jackson organized on Twitter and Facebook to suppress a Jackson biography: they objected to positive reviews, posted negative reviews, and advocated that the book be removed from sale—which it was for a short time. Ideological reviews of a different sort were made during the 2012 U.S. presidential campaigns when Barack Obama visited a Florida pizza parlor. When the owner, a registered Republican, made the most of the visit by endorsing and embracing the president in a bear hug, the owner’s Yelp page exploded with hundreds of partisan, nonpizza-related comments. Oddly, one study of a clothing store’s Website found that many deceptive reviews were written by loyal customers who posted reviews for products that they did not buy and did not want to see continued. The study’s authors suggested that people did this to enhance their standing on the store’s site or act as “self-appointed brand managers,” noting that the most loyal customers are often the harshest.24

Companies and governments take part in fakery as well. Astroturfing is the practice of simulating authentic grassroots support for a company or government. The electronics firm Samsung has reportedly paid people to comment in its favor, criticize its competitors, and participate in and hype its smartphone application challenges. Governments use the technique as a form of propaganda by extolling the virtues of those in power, removing negative comments, and attacking critics. The U.S. military uses sophisticated “persona management software” (that is, sockpuppets) to support “classified social media activities outside the U.S., intended to counter violent extremist ideology and enemy propaganda.” Other nations have been documented using these techniques on their own populations. China and Russia reportedly hire civilians for the job, and in Thailand, commenting in support of the monarchy is the duty of a unit of uniformed soldiers. In South Korea, the head of the National Intelligence Service and other high-ranking officials have been indicted for their covert Twitter campaign during the 2012 South Korean presidential elections.25 If something or someone can be applauded or pilloried in a comment—whether a hotel, gadget case, plumber, doctor, singer, or politician—there will be fakes.

The Illicit Markets of Makers and Takers

Fakers today no longer need to do it themselves or go it alone: takers can pay for the services of makers. In 2010, when Jason Rutherford launched GettingBookReviews, his basic $99 service involved posting a review to the major book sites within seven to ten days. Today, takers can pay much more or much less for this type of service, although today makers tend to be a little more discrete than Rutherford.

At the high end, fakery can be an expensive service that an agency provides as part of a reputation-scrubbing or promotional campaign. Journalist Graeme Wood’s introduction to the “world of black-ops reputation management” began with some schadenfreude. Wood’s former Harvard classmate was in trouble with the Internal Revenue Service for helping his mother hide the family’s fortune (by moving bags of cash around the world). Wood wrote that “as a taxpayer and a jealous prole, I watched his downfall with special interest, and even set up a Google Alert to keep abreast of developments.” But the “developments” were press releases and Websites that extolled his classmate’s professional and philanthropic successes. These were being orchestrated by a “boutique shop for the online reputations of very wealthy people” that charged as much as $10,000 a month to drown out negative press about its clients.26

Fakery can be had on the cheap as well, and most of these ad hoc services are found on sites like Freelancer, Fiverr, and Craigslist. On Craigslist, for instance, those who are willing to write piecemeal reviews are asked: “Are you a savvy writer? Do you like to write book reviews? Are you willing to write a glowing review for a cookbook? Are you an customer or member?” If so, “You will receive a Los Angeles Metro Discount Coupon Card for your participation. Save on pizza, car repairs, etc. $25 Value.” Some ads are written by people who are willing to review, such as the “ACTIVE YELP WRITERS” who “WANT TO WRITE REVIEWS 4 YOUR BIZ.” They explained that “we aid you obtain a better reputation by acquiring you favorable reviews,” which they claimed will lead to more customers. But this ad does not sound as though it was written by a native English speaker. In fact, it probably was created by an overseas “review sweatshop” where workers will never experience the things that they review. Even tech-savvy homeless people can get in on the game with an old laptop and access to wifi and power outlets. In an August 2013 story, unemployed computer technician Jesse Angle estimated that he had earned about $500 via BitCoinGet, a site that pays “every time you watch a video or complete a task”—likely inflating viewership numbers.27

Among “made in America” review makers, the going rate seems to be about $5 to $20 dollars per review. Some review takers can require reviewers to visit the establishments and have an existing account on sites like Yelp, Citysearch, and Facebook. One such sophisticated ad even required that the reviewers write a range of reviews (to avoid appearing overly positive) over a period of time (because a rush of positive reviews is suspect) from unique accounts: “Reviews should be very different from each other—ie, one might say ‘Item was shipped quickly’ and another might say ‘A+ great service!!’ while another (3-star) might say ‘I was satisfied with their customer service’, etc.”28

For some of the most prominent Amazon reviewers, it’s not immediately clear what they gain, but there are suspicions. In a 2008 essay in Slate, Garth Hallberg noted how pleased he was to receive a five-star review from Grady Harp, Amazon’s seventh top reviewer with over 3,500 reviews to his name. Hallberg wrote “Sure, he’d spelled my name wrong, but hadn’t he also judged me ‘a sensitive observer of human foibles’?” And over one hundred readers clicked that they found the review helpful. However, “after a brief e-mail exchange, my publicist confirmed that she’d solicited Grady Harp’s review.”29 Comments on reviews and in discussion boards are full of questions about the legitimacy of top reviewers. Hallberg noted that Amazon reviewers are suspected of receiving free promotional copies or other benefits (including payments). Also, Harp’s one hundred votes of helpfulness might indicate the existence of “back-scratching” wherein cliques of reviewers rate one another’s reviews as helpful. Finally, many wonder how top reviewers can review so many books. Could Harriet Klausner, a woman who dominated Amazon’s reviewer rankings for many years, read and review, on average, forty-five books per week for over five years (even if she is a speed-reading insomniac, as she says)? A commenter on the (sarcastically titled) Harriet Klausner Appreciation Society blog noted that if she kept the books that she has read, she could fill over 240 bookcases, although it appears she sells some of the books online at Critics also question the usefulness of top reviews given that the vast majority of those reviews are positive: more than 99 percent of Klausner’s reviews are four or five stars relative to the (still high) 80 percent of all reviews.30

For fake-review takers, the results can be impressive. Author John Locke was one of the first to sell a million $0.99 self-published ebooks on Amazon. He followed this success with How I Sold One Million eBooks in Five Months (the Kindle edition sells for $2.99). However, as “the most helpful critical review” of the book on Amazon noted, there is “a secret he left out”: “In an interview with Locke in today’s New York Times, he admitted that he paid for 300 reviewers to heap praise on his books, a sleazy promotional technique that seems to have worked for him.”31 In the Times article, journalist David Streitfeld noted that Locke was confident in his writing but frustrated with his efforts to reach readers via labor-intensive blogging, tweeting, and personalized email. So he made use of Rutherford’s GettingBookReviews, starting with purchasing fifty reviews for $1,000 and eventually purchasing three hundred. Locke specified that “If someone doesn’t like my book they should feel free to say so” but that reviewers should purchase books from Amazon so they would show as “Amazon verified purchases,” lending credibility to reviews and improving the book’s Amazon sales rank. Locke has sold millions of inexpensive digital books, and in 2011 he scored a deal with Simon & Schuster to manage the sales and distribution of the print editions of one of his spy series. Locke might argue this tactic was little different than that of authors who use the more traditional Kirkus reviews. The Kirkus Indie review service charges self-publishing authors $425 for a review: “If the review is negative, you will have the option of keeping it private and simply using the assessment as feedback to improve your craft. If it’s positive, you will be able to use the review to market your book to consumers or to catch the attention of a literary agent or publisher.”32 However, this service follows a recent transfer of ownership as Kirkus struggles to remain viable in the new millennium. The New York Times reported on “The End of Kirkus Reviews” in 2009, and it was subsequently purchased by a new owner. Presumably, the self-publishing and review market is one way that the company hopes to stay afloat.

Sometimes review takers get taken themselves. Complaints of incomplete, shoddy, or tardy service can be found about Rutherford’s GettingBookReviews and Kirkus. On Craigslist, I found a disappointed reviewer warning others of a scammer who promised $60 per review of cool musical gear. But “when you want to get paid,” you receive only $20 per review: “Stay away from this one, unless, you really enjoy writing boring product reviews for little compensation.” Similarly, researchers who bought fake Twitter accounts found that out of their twenty-seven black-market merchants, eight attempted to sell them accounts already sold; one merchant tried to resell them the same thousand accounts three times.33

Although product reviews and ratings are obvious targets of manipulation, they are not the only things that can be gamed. In a study of reviews and ranking on a photography Website, I concluded that although it can be hard to quantify the qualitative, such as artistic merit, we attempt it nonetheless, and these quantitative mechanisms beget their own manipulation. At, people “mate” rated friends, “revenge” rated enemies, and inflated their own standing. Fixes to such manipulations often take the form of more elaborate quantification, such as labeling comments as “helpful,” which are then manipulated as well. This insight is popularly referred to as Goodhart’s or Campbell’s law, concisely expressed by anthropologist Marilyn Strathern: “When a measure becomes a target it ceases to be a good measure.”34 For instance, online influence can be measured by Twitter followers and retweets, yet these (as well as YouTube and Facebook likes) are easily purchased by the thousands. In May 2012, the network and data security firm Barracuda set up three new Twitter accounts and was able to purchase between twenty and seventy thousand Twitter followers for each account, finding some twenty eBay sellers and fifty-eight Websites that sell fake followers. For a handful of dollars, hundreds or thousands of people will look at a Webpage for half a minute, click on the “share on Facebook” button, or retweet. Even a credit card stealing virus was modified to instead create false likes for the Instagram photo-sharing site.35 All of the big sites have either purged suspect reviews or demoted them by changing their filtering algorithms to stay abreast of the fakes. For instance, in 2008, Amazon replaced its “classic” reviewer ranking system with a “new” one that demoted many of the top book reviewers. Amazon’s intent was for the new system to favor more recent reviews that are rated as helpful while ignoring attempts to stuff “the ballot box.”36 This change caused Harriet Klausner to drop from her top seat to 2,057th place by April 2014. Conversely, Grady Harp, who has maintained a higher helpfulness rating, has fallen only to 47th place.

Clearly, this is a bit of a cat-and-mouse game, as demonstrated by those annoying puzzles that we sometimes must solve to post a comment.

CAPTCHA Cat and Mouse

The morning programming on my local NPR station is often “brought to you by” Angie’s List: “Reviews you can trust.” It also is supported by “Working to protect individuals and businesses by ensuring accurate search results and authentic online reviews.” This is an intriguing juxtaposition: at Angie’s List, members pay a fee to be able to read and write reviews of local service providers, and at, clients pay to limit the damage from, among other things, negative reviews. These two companies (and consequently NPR) benefit from the comments at the bottom half of the Web. But what about the actual customers and merchants? I have discussed the value of reviews, the extent of manipulation, and some actual manipulators. But manipulation is not only the actions of mistaken, desperate, or immoral people. Comment manipulation is best understood in terms of an arms race in which people scramble to make their way on the Web.

This arms race is exemplified by CAPTCHAs, which became widespread around 2002 in response to the proliferation of free online services. (CAPTCHA is an acronym for “Completely Automated Public Turing test to tell Computers and Humans Apart.”) Although the ability to comment from a Web-based email site or social network is a boon, such services are also abused to create sockpuppets and to post spam. Automated scripts (robots or bots) can easily create such accounts by the thousands. A CAPTCHA regulates the creation of accounts by requiring the completion of a task that is easy for most humans but difficult for computers. One prominent implementation, reCAPTCHA, uses garbled text from book-scanning projects. By solving a CAPTCHA, users prove that they are human and also contribute to the public good by helping digitize books.

Yet, “necessity is the mother of invention,” and fakers respond with ever more sophisticated techniques to automatically solve CAPTCHAs. This then prompts the invention and deployment of new CAPTCHAs schemes. In 2010, researchers interested in “how good humans were at solving CAPTCHAs” had twenty-one popular schemes to test against. It turns out that they often are annoyingly difficult: when presented with the same challenge, three human subjects agreed with one another only 71 percent of the time.37 Real humans are being left behind in the CAPTCHA arms race and many have simply stopped trying. When the video service Animoto replaced CAPTCHA with a different robot-frustration technique, completion of its sign-up form by humans increased by 33 percent.38 Also, the efficacy of CAPTCHA has been undercut by the arrival of cheap human labor online.

In the Future Sophisticated CAPTCHAs Will Lock Out Any Bot

© 2008 CC-BY-3.0

Amazon’s Mechanical Turk, which was launched in 2005, is named after the famous eighteenth-century chess-playing automaton. The chess-playing contraption appeared to be a machine but had a human chess master hidden inside. Similarly, “MTurk” appears to be an automated Web service, but its tasks are actually completed by people. These human intelligence tasks (HITs) are things that people are good at, like categorizing images, translating text, and verifying information. Using MTurk, some people can request that other people perform HITs for them in a fast and automated way. Amazon noted that “Turk Workers” (or “Turkers”) can work from home, on their own schedule, and “get paid for doing good work.” On the other hand, “Requesters” have access to a “global, on-demand, 24x7 workforce”; they can “get thousands of HITs completed in minutes” and “pay only when you’re satisfied with the results.”39 This system could also be used to solicit fake reviews and ratings, send spam, fraudulently click on ads, fill out forms, and create sockpuppets. In 2010, computer scientists at New York University measured how many new Requesters were submitting such HITs. Cleverly using Turkers to do the actual categorizations, they found that 41 percent of the 5,841 HITs were illicit.40 Such fakery is now against MTurk’s terms of service, and Amazon will seize the accounts of violators, though some complain of unfairness. One Turker faulted Amazon for not making these restrictions clear. He lost over $300, mostly gained from image categorization, when he responded to a HIT “like the one that asked me to type in a captcha image for 1 penny. That’s right. One penny cost me more than $300 dollars.”41

Although MTurk is no longer as palatable to manipulators, some sites traffic specifically in CAPTCHAs. Death by CAPTCHA (the “cheapest CAPTCHA bypass service”) has plans that start “from an incredible low price of $1.39 for 1000 solved CAPTCHAs.” Solutions must be offered before the challenge expires; this service claims that the average response time is fifteen seconds with an accuracy rate of 90 percent but warns that “our services should be used only for research projects and any illegal use of our services is strictly prohibited.” Such illicit uses would include creating email accounts from which to post fake reviews and spam. Yet Death by CAPTCHA’s page used to say “Don’t let CAPTCHAs get in the way of your marketing goals! With Death by Captcha, you can bypass any CAPTCHA from any website. All you need to do is implement our API, pass us your CAPTCHAs and we’ll return the text. It’s that easy!” The ironies and hypocrisies of these services seem lost on their proprietors, such as the folks behind Bypass Captcha: they can be contacted only by a Web form that is protected by a CAPTCHA “anti-fraud code.”42 However, not all CAPTCHAs and accounts are valued equally. The authors of a 2013 study of “the underground market in Twitter spam and abuse” were able to purchase fake Twitter accounts from twenty-seven different merchants with a median cost of four cents per account. This is a relatively high cost compared to the fractions of a penny that Hotmail and Yahoo accounts go for because Twitter uses CAPTCHAs and phone-verified accounts. That is, a user is verified by receiving and entering a code that is sent to them on their phone. Sadly, on Craigslist one can find schemes that trick people into receiving such codes and handing them over to scammers. Other people who are duped by scammers are the unwitting downloaders of free pornography, movies, and software. For example, a scammer in need of CAPTCHA solutions can place the challenges on a pirated music site. The solutions provided by those downloading music are then passed back to the scammer to use. In response, more sophisticated CAPTCHAs are being deployed, which themselves have commercial overtones. One CAPTCHA service requires users to watch a commercial and type in a product slogan to “prove one is human.” Supposedly this delivers “1200% greater message recall than banner ads.”43

This arms race, a competitive mix of exploitation and frustration, is characteristic of the larger dynamic of manipulation online. Increasingly, this arms race takes a commercial turn. Although we once solved CAPTCHAs to help book-scanning projects, we are now forced to prove ourselves by interacting with advertisements.

“Pay to Play” and Extortion

Although I rely on online reviews, I know that they can be frustrating and duplicitous; the emergence of review sites and reputation services does not allay my concerns. Consider the following scenario. You wish to find a local service provider (such as a plumber or doctor) and wonder whether you should join Angie’s List, which you have heard about on the radio, so that you can see what its members recommend. Before you pay the membership fee, you investigate what people say about Angie’s List itself. After a simple search, you find warnings that it is difficult to terminate a membership, people are automatically resubscribed at a higher cost, and the fee would not necessarily keep a provider’s family and friends from joining and posting inflated reviews. It also appears that service providers—including those not associated with the site—are pressured by aggressive sales people. One small business owner wrote that when Angie’s List solicited him by noting he had a negative review, he was advised to “have all my good customers join and add their reviews. The customers would have to pay to join then pay to leave reviews. Pretty much like blackmail.”44 Is the desperation to have positive reviews the reason that some businesses give coupons to review writers?

How far is too far in encouraging positive reviews? At Amazon, a merchant of an inexpensive case for the Kindle Fire offered a refund to consumers who would “write a product review for the Amazon community.” Although the merchant did not require five-star reviews, it hinted that they were appreciated: “We strive to earn 100 percent perfect ‘FIVE-STAR’ scores from you!”45 If such reviews do not mention the refund, they are likely in violation of U.S. Federal Trade Commission (FTC) guidelines from 2009. The rules require the disclosure of nonobvious “material connections” (such as payments or free products) for consumer endorsements, including those “disseminated via a form of consumer-generated media.” These guidelines were widely reported on but little noticed, understood, or enforced. In 2013, the FTC again attempted to remind the public of the requirements, even in “space-constrained ads” such as tweets.46

In addition to claims of “blackmail,” one can also find accusations of “pay to play” and “extortion.” Sites like Pissed Consumer and Ripoff Report are filled with such complaints about Angie’s List, Yelp, and others. Yet many “consumer advocacy” sites are the subject of similar accusations: they are said to encourage false reports or refuse to remove reports unless they are paid or receive kickbacks from a reputation-management service. The worst pay-to-play schemes post “revenge porn” (embarrassing photographs or videos), mugshots, and photos of supposed sexual predators that can be removed for a fee. The now defunct site STD Carriers published claims about people’s alleged sexually transmitted diseases. It refused to remove reports but would place a no-index tag on the reports of those who paid, rendering the report invisible to most search engines.47

From a legal perspective, Pissed Consumer and Ripoff Report are smart to refuse to remove reports. Section 230 of the U.S. Communication Decency Act (CDA) declares that “No provider or user of an interactive computer service shall be treated as the publisher or speaker of any information provided by another information content provider.”48 That is, a site that hosts information posted by users is not liable for that content. (Sites are obliged to remove illegal content, however, such as child pornography.) Some legal scholars feel that the immunity should not be extended to sites that knowingly host or solicit content that is defamatory or invasive of privacy. A judge felt similarly in a 2013 ruling against the gossip site The Dirty, although the decision is likely to be appealed. In any case, people are understandably alarmed to find negative comments about themselves online. Writer James Lasdun wrote of being harassed online, including in online reviews of his early work. Even if most people discounted the slander, he felt that “its real harm was the notice it gave that I was a person to whom such a thing could be done: that I had attracted an enemy who wanted the world to categorize me as an object of scorn.”49 While those who pay to have such content removed probably feel it is necessary, they also likely feel doubly abused.

The Racket of Reputation Management?

The main tactic of reputation services is to ensure that enough positive content is on the Web to drown out the bad. While customers hope to displace the negative content from at least the first page of search results, nothing is guaranteed. Sites that host negative comments also can be paid to lessen their complaints’ impact. Pissed Consumer offers a “service” that allows clients to write a response or preemptively “get contact information of the consumer before [the] complaint is posted on the site.” Ripoff Report offers a free rebuttal and a similar $90 per month program “that affords a business the opportunity to fix a problem with a complaining customer before the world has the opportunity to read the Report.”50 Other fee-based services can “make your search engine listings change from a negative to a positive” by changing the title of a report which is seen in search results. In 2010, Ripoff Report introduced a “VIP Arbitration Program,” something that external “reputation management services can’t deliver”:

Is your company the subject of a false Ripoff Report? Do you feel a free rebuttal is just not enough? Would you like the opportunity to prove that the Ripoff Report about you or your business is false? Would you like false statements of fact redacted from the report? Rip-off Report has contracted with private arbitrators who have extensive experience, including judges, to decide your case. Clear your name once and for all the right way.51

The “filing fee” for this service is $2,000 to “pay for the arbitrator’s time and for administration of the program.”

Even widely known and used sites are suspect. Journalist Kathleen Richards wrote a series of “Extortion 2.0” articles that documented merchants’ complaints that Yelp pressed them to advertise on the site. If merchants joined Yelp’s sponsorship program, Yelp would feature a positive review on the merchant’s page, cause the business to show up in related searches, and have the merchant’s ads appear on competitors’ pages but not vice versa. This, combined with two other facets of the site, led to allegations of extortion. First, the opaqueness of Yelp’s filtering algorithms can be alarming: some reviews are shown by default and included in the overall star rating while others are shunted to the “filtered reviews” section. Yelp, like most all sites, keeps its algorithms proprietary, and this is a cause for concern and confusion across the Web. (Yet if Google’s ranking or Yelp’s filtering algorithms were known, they could be more easily gamed.) Second (and more troubling), Yelp employees write reviews. Merchants cry foul when they decline to advertise on Yelp and a negative review by a Yelp employee soon follows.52

Such complaints have yet to make much headway in U.S. courts. In 2011, two related legal suits against Yelp were dismissed. At issue was whether Yelp’s filtering and aggregation of user comments stripped it of its CDA Section 230 immunity: that is, does Yelp’s filtering of user-generated comment make it a publisher and consequently liable for the content? Regardless of Yelp’s possible motives in the presentation of users’ comments, the judge wrote that the plaintiffs did “not raise more than a mere possibility that Yelp has authored or manipulated content related to Plaintiffs in furtherance of an attempt to ‘extort’ advertising revenues.” Cyberlaw scholar Eric Goldman opined that “This ruling makes clear that Yelp can manage its database of user reviews however it wants.… However, it doesn’t mean that we as consumers will find Yelp trustworthy.”53 Other scholars contend that Section 230 immunity needs revision given the harassment, extortion, and defamation that occur within its shadow. Law professor Ann Bartow goes further, arguing that the “monetization of Internet harassment” creates “unsavory incentives” for the services to see the status quo (of harassment) perpetuated.54

Even the nonprofit Better Business Bureau (BBB), which has been an arbiter of reputable businesses since well before the Internet, has been tainted by monetization. In 2009, Bill Mitchell, the Los Angeles chapter’s chief executive officer, was forced to resign because of allegations of a pay-to-play scheme. Mitchell, head of the BBB’s largest national chapter, also originated the new letter-grade accreditation system, replacing the classic satisfactory or unsatisfactory labels with grades such as A plus and B minus. The scheme was so successful that it was adopted across the country. Yet after merchants were solicited by sales representatives (who earned 45 percent commissions on first-year memberships), they claimed that they were assigned unsatisfactory grades as a strong-arm sales tactic. The television news show 20/20 reported that unsatisfactory grades were elevated overnight after payment was received. To test this claim, producers were able to easily register a number of fictitious businesses, including an A plus “bogus firm named after Stormfront, a white supremacist group.”55

Consumers are getting in on the racket as well. One reviewer claimed food poisoning, refused a $60 gift card at a restaurant of his or her choosing, and demanded $100 to refrain from posting a negative review. At Google’s app store (that is, Android’s Google Play), users often promise to raise a rating if a shortcoming is addressed: “Fix that and I’ll give it 5 stars.” For many years, app developers could not respond to reviews and would beg users to send questions, bug reports, and feature requests to them before posting a negative review. In 2013, when Google enabled developers to reply to reviews, the Android blog Phandroid quipped: “No longer will developers have to break down and cry when a user leaves a 1-star review for a feature they don’t know how to use.”56 For those willing to pay $100 for a plastic card, the ReviewerCard is the “first-of-its-kind membership card and community for reviewers”: members who display the ReviewerCard “enjoy premium service.”57 It is a fancy way of telling the restaurateur that cardholders might write a review. However, no restaurant is obliged to do anything because of the card, and a similar card is easily made for less than $100, so it is not clear who is exploiting whom.

All of this leads to conflict and headlines. Sometimes review sites are sued, such as Yelp for pay-to-play abuses and TripAdvisor for negligence in allowing defamatory reviews. More often, organizations try to make it impossible to be sued, or they sue their reviewers. In 2014, the food conglomerate General Mills updated its terms of service so that anyone interacting with it via social media waived the right to sue. Aggrieved parties that had liked Cheerios or any other GM brand would have to enter “forced arbitration” instead.58 Medical practitioners are especially concerned about deleterious comments online. Reputation-management firm Medical Justice (“making healthcare safe for doctors”) is infamous for its attempts to prevent bad reviews by providing its members with contracts that forbid patients from posting negative comments about a service provider. Another tactic is to prompt the patient (in fine print) to assign any copyrights in such comments to providers, who may then remove them at their discretion. In a bizarre case, a New York dentist sued a former patient for a review in which the patient claimed that the dentist overcharged him by thousands for what should have been a couple-hundred dollar procedure. However, Medical Justice refused to defend the terms of the contract (and has since abandoned the criticism clauses), and the dentist closed her office and was unreachable by her lawyers.59 In an unrelated but similar incident, Utah consumers John and Jen Palmer claimed that their credit had been ruined by a company that wielded a nondisparagement clause. In 2008, the couple had expressed disappointment with in a negative review on Ripoff Report. Although I have reservations about Ripoff Report, KlearGear’s 2012 levy of a $3,500 fine for “nondisparagement” is also troubling. First, the clause was allegedly added to the site’s terms of service after the Palmers’ 2008 transaction. Second, when the Palmers refused to pay the fee, the nonpayment was referred to credit agencies, damaging the family’s ability to get credit for essential purchases.60

Sometimes online arguments about reviews become real, as when an owner of a bookstore showed up at a surprised Yelp reviewer’s door because he had said that her store was a “total mess.” Also, news stories of crackdowns are common. In 2012, all of the following were reported. Amazon users noted that thousands of reviews had been deleted, although Amazon would not disclose numbers or a rationale. Facebook removed suspect likes, including 200,000 (about 3 percent) from a popular poker game page. YouTube purged more than two billion suspect views from channels belonging to large music labels. And of Yelp’s thirty million reviews, roughly 20 percent were found to be filtered as suspect.61 In addition to these purges, the big sites are also hoping to solve fakery via the “social graph,” favoring the comments and activities of your acquaintances, although this will likely lead to new forms of manipulation.

Exploiting the Social Graph

On one hand, it’s good to hear about services cracking down on fake comments, reviews, likes, and views. However, I suspect that this is as much for the benefit of the sites as it is for their users: external manipulators are in competition with the sites’ own manipulation of their search results, ad placements, and “sponsored” reviews. Even if only true comments were posted at a site, the site would still seek to make money by manipulating the prominence of certain comments. There’s a reason that Consumer Reports is widely trusted: it has no financial relationship with the makers of products, it even pays for all the products it tests. This is why I tend to be more trusting of Amazon among online retailers: they are relatively impartial among items sold and the reviews posted. This doesn’t mean that Amazon is angelic: it has been accused of exploiting publishers and its employees to keep costs low, and it does sell ads on its sites for “sponsored products.”62 I also wish that Amazon did more to remove fraudulent reviews and helpfulness votes. Yet, while the particulars of its search algorithm and product placements are proprietary, Amazon makes a commission on sales, so it should be biased toward items that sell well—things that people actually want. Also, it’s easy to see which reviewers actually purchased a product (“verified”) and Amazon has formalized “kick-backs” by way of the Amazon Vine program: products given to reviewers are explicitly identified as such.

Even so, Vine reviews can be amusingly weak, such as one for the engineering book System Identification: A Frequency Domain Approach; a Vine reviewer wrote little more than “If your into math you will most likely enjoy this book.”63 And although more expensive (and complicated to purchase), illicit reviews can be had from those who actually buy the books from Amazon such that they are labeled as verified purchases. The persistence of such fakery is why the big Web services are now battling over a new paradigm of user comment that is known as the social graph. Instead of relying on strangers to inform consumption, people could use their friends’ comments instead. People would no longer have to fear manipulation from strangers.

However, users still would not be free from manipulation by the services themselves. In spring 2012, Facebook demonstrated this with the deployment of Sponsored Stories—“messages coming from friends about them engaging with a Page, app or event that a business, organization or individual has paid to highlight so there’s a better chance people see them.”64 (A Facebook “Page” is a paid-for account that is integrated into Facebook’s advertising system with an unlimited number of friends/fans.) Sponsored Stories were labeled as such and could appear on the right side of the Facebook interface, but users were annoyed to find them also appearing in their News Feed (where Facebook limited their appearance to once a day). Those who had Page accounts also found that they were having a more difficult time reaching their fans. An average of only 15 percent of a Page’s fans were “reached” and saw its messages. If those who had Page accounts wished to reach more fans, they needed to pay for Sponsored Stories. Cries of ransom, robbery, and extortion soon followed.

Worse yet, some users were surprised to find their photographs in Facebook ads. Facebook user Nick Bergus came across a beguilingly odd product on Amazon: a fifty-five-gallon drum of sexual lubricant that was listed for over a thousand dollars and had 146 customer reviews, although there was little evidence of any verified purchases. The reviews—and probably the product—were farcical. The highest-rated review was by actor and social media star George Takei (Mr. Sulu from Star Trek), who discussed how he tested the lubricant by spraying it on revelers during the San Diego Pride Parade. After Bergus posted a link on Facebook, commenting that the lube was “For Valentine’s Day. And every day. For the rest of your life,” his face began appearing in Facebook ads for the product. Similarly, Angel Fraley, an eventual plaintiff in a lawsuit against Facebook, appeared in ads for the Rosetta Stone language learning system because she had liked its online French course in hopes of getting a discount. She might not have received the discount, but she did appear in Rosetta Stone ads. Although this advertising practice was disclosed in the fine print of Facebook’s terms of service, users challenged (and sued) that this was a consentless use of their identities, especially for users under the age of eighteen.65

Yet Sponsored Stories seemed to be effective for those who were willing to pay. One Web analytics company concluded that the ads cost roughly 20 percent less per click and per Facebook fan and were clicked on 46 percent more than standard Facebook ads. Although the service was soon bringing in $1 million in revenue per day, the bad publicity and the class action suit (in which Facebook agreed to pay $20 million) led Facebook to discontinue the service in 2013 as part of a larger restructuring of its advertising. However, Facebook’s product marketing director Brian Boland stressed that “Sponsored Stories as an idea doesn’t go away. Sponsored Stories as a product goes away.”66 Indeed, others have followed Facebook’s model. Twitter has “promoted tweets” and the comment platform Disqus offers “Sponsored Comments,” which “let businesses deliver a message to the people they need to reach” by pinning their content to the top of your discussion.67 In 2013, Facebook’s main competitor in the space, Google, announced “shared endorsements.” While Google attempted to be more careful of upsetting its users, it still prompted some concern with the declaration that: “To help your friends and others find cool stuff online, your activity (such as reviews, +1s, follows, shares, etc.) may be used along with your name and photo in commercial or other promotional contexts.”68 And to help you generate sufficient comment, Google is also thinking about mechanisms that prompt users with “suggestions for personalized reactions or messages” based on what it knows about its users and their social networks. For instance, the patent application for this mechanism noted that “it may be very important to say ‘congratulations’ to a friend when that friend announces that she/he has gotten a new job,” and if the busy user forgets to do so,69 Google will suggest an appropriate comment. Facebook and Google want to help us generate grist for their mills.

Social networking sites might even benefit from external manipulators. Accusations of fraud surfaced again in 2014 when users noticed that paying Facebook to promote their content caused their engagement—and subsequent reach—to drop. Reach is the number of people who see content, and engagement is the number of people who click, like, or comment on it. At Facebook, users do not see all the posts of those they follow. Facebook parcels out users’ posts to portions of their friends and fans based on how much engagement the content is getting and whether it was “promoted.” People were concerned that although the promoted posts resulted in a lot of new fans, engagement numbers dropped even further. Online science communicator Derek Muller suspected that this was the result of like farms trying to hide their activity from detection. Manipulators often pay others (often in poor countries like Indonesia and Bangladesh) to like their content so that Facebook will share their posts with more users. (Facebook sees this as engagement, infers that it is useful, and spreads it further.) To avoid detection, these scammers hide their fraudulent activity in a storm of random likes, often landing on recently promoted posts. Although Facebook could boast about the benefits of advertising based on the number of new fans, people were falling further down a hole. Promoted content on Facebook is clicked on by the farms (to mask their fraudulent clicks), which subsequently drives down relative engagement because real fans are outnumbered by the now dormant fakers. Muller complained that Facebook was doing nothing to address the problem and that “they have since benefited from those 80,000 likes” because “I paid to boost posts out to these useless likes.” Others concluded that the hundreds of thousands of dollars that they paid to Facebook had been wasted.70

The dynamics of all of this is captured in a Geek&Poke Webcomic in which two pigs discuss the benefits of free services like Facebook and Google. One pig says, “Isn’t it great? We have to pay nothing for the barn,” and the other responds, “Yeah! And even the food is free.”71

A “Loss of Innocence”

“E.Z.” (a pseudonym) is an Amazon Top 500 Reviewer who has posted over 250 reviews (a relatively low number for top reviewers), and he has accumulated nearly two thousand “helpful” votes on those reviews. He started using the Internet in the 1980s and has shifted from being an active and public participant on early Internet forums to tending to his own online activities and Websites. He told me, “I am not a twenty-year-old who is always posting on his Facebook account and telling everyone what he ate for breakfast every day.” Although many Amazon Top Reviewers focus on items that are easy to review, E.Z. believes that quality matters more than quantity, so he is selective in what he reviews and is often exhaustive in scope.

Pigs Talking about the Free Model

© 2010 CC-BY-3.0

I first encountered E.Z. while I was looking for a digital voice recorder to use for interviews for this book. My purchase was influenced by the depth of his seven thousand–word recommendation. It was so large that it exceeded Amazon’s word limit and continued in his review’s first comment. Over a hundred other appreciative comments followed. When E.Z. began reviewing he started slowly and did it as “a cathartic way to vent about my frustrations using bad products and to also warn other people not to buy those products.” Since then, E.Z.’s reviews have been mostly positive, which he attributes to his thorough research and comparison shopping. Given his long history with online comments, I was interested in E.Z.’s sense of how things have changed over the years, especially his thoughts about Amazon’s Vine program in which top reviewers are given free products to review.

E.Z. told me that Amazon invited him to participate in Vine in the spring of 2008 and that the several thousands of dollars’ worth of free merchandise every year does motivate him to write more reviews. (To stay in the program, reviewers originally had to review 75% of the items they received; this was later moved to 80%. In 2013 some reviewers objected to a proposal that they must review 100% of received products within thirty days.)72 His time in the program “has filled every room in my house with free goodies,” including books, four digital cameras, a large-format photo printer, a $1,400 home security camera system, other electronics, and lots of software. Other Vine reviewers have received refrigerators, washing machines, and dryers. He conceded that this likely prompts “the three amigos of greed, jealousy, and envy” as reviewers “clamor for higher reviewer rankings and better products than their fellow Vine members.”

An aphorism among those who study human motivation is that extrinsic motivations crowd out intrinsic ones. This means that an external imposition (such as a payment or punishment) often lessens people’s internal motives related to feelings such as autonomy, mastery, usefulness, and self-esteem. A classic example of this is a daycare center that struggled with some parents who frequently were late in picking up their children. When researchers experimented by adding a fine for tardy parents, the unwanted behavior actually increased. The late parents saw the “fine as a price” that they were willing to pay. The external punishment displaced their internal motive to see themselves as responsible parents who abided by an implicit social contract with the teachers. Even when the fine was removed and the researchers left, the high rate of tardiness continued. At Amazon, top reviewers report that they are motivated by intrinsic values such as self-expression and enjoyment.73 However, Vine might be changing that. E.Z. noted that if he no longer received free Vine products, “my interest in writing online reviews would likely plummet, and I would not care how low my Amazon Reviewer Ranking would plummet because of that.” His motive to warn others away from a bad product might remain, but he rarely purchases a dud now and probably would not bother to write positive reviews.

Finally, E.Z. felt that there has been a “‘loss of innocence’ with online discussions, comments, and reviews of products.” Before 1996, when advertisements on the Internet were verboten, E.Z. enjoyed authentic discussions, such as those about bicycling and mountain bikes:

Back then, there was not a single advertisement posted on those newsgroups and no mountain bike manufacturers were posting comments or (as some manufacturers do now with Amazon reviews) posting fake comments that raved about a specific brand of mountain bike. So if someone was interested in buying a mountain bike during those Usenet days, they could either post questions themselves or read other postings on a newsgroup and all of the raves, criticisms, and good or bad comments about products were much more sincere and authentic than what we have to sift through on the Web these days.

So with Amazon’s Vine program, with Amazon’s reviews in general, and with reviews on other sites like Yelp, TripAdvisor, etc., I feel that there has been a “loss of innocence,” and this loss of innocence is often happening simply because money is involved, either with the free products being given by Amazon to their Vine reviewers or with manufacturers realizing the increasing importance of user reviews/comments in this globally socially networked world that we now live in.74

As seen in this chapter, many are complicit in this loss. People traffic in the illicit markets of comment, some click a like in hopes of a discount, diners loudly discuss Yelp when their server is nearby, restaurateurs give coupons in exchange for reviews, authors ask friends to write reviews (or do so themselves), sockpuppets edit Wikipedia biographies, pundits purchase fake followers, and sites profit by manipulating users’ praise and pillory. Much of this behavior is driven by the high value of comment today, an obsessive desire to rate and rank everything, the dynamic of competition, and the sense that everyone else is already doing it. The world of online comment is quite different from that of early likers, like E.Z., who reviewed for the love of it. At the bottom of the Web we are increasingly tempted to become manipulators, and as we do so we lose something in the process.

Header icon made by Gregor Cresnar from


kamal bisht:

Wonderful post keep going