You May Also Like (39 page)

“The sweet taste of sucrose”
: John Prescott,
Taste Matters
(London: Reaktion Books, 2012), 31.

“That was twenty years ago”
: As Small notes, “When an odor is experienced with a taste, the odor later comes to smell more like the taste with which it was experienced.”
The suntan lotion virtually
becomes
Malibu. See Dana Small and Barry Green, “A Proposed Model of a Flavor Modality,” in Murray and Wallace,
Neural Bases of Multisensory Processes
.

“a distributed circuit”
: See ibid.

In one experiment
: Ivan E. de Araújo et al., “Metabolic Regulation of Brain Response to Food Cues,”
Current Biology
23, no. 10 (May 2013): 878–83.

Kent Berridge, a neuroscientist
: Mike J. F. Robinson and Kent Berridge, “Instant Transformation of Learned Repulsion,”
Current Biology
23, no. 4 (2013): 282–89. The authors note that in previous work, it was “not clear whether an instant transformation is powerful enough to reverse intense learned repulsion (such as to a CS for Dead Sea concentrations of 9% NaCl) into instant strong desire. Our results show that both do happen: a CS instantly powerful enough to reverse cue value from strongly negative to strongly positive.”

“mesolimbic circuits of ‘wanting' ”
: Kent Berridge, “Wanting and Liking: Observations from the Neuroscience and Psychology Laboratory,”
Inquiry
(Oslo) 52, no. 4 (2009): 378.

The idea of finding
: As one neuroscientist has noted, “Not a single area implicated in pleasure in the human literature has failed to be implicated in aversive processing as well.” The quotation is by Siri Leknes, in Morten L. Kringelbach and Kent C. Berridge,
Pleasures of the Brain
(New York: Oxford University Press, 2010), 15.

And yet their own creators
: For a good discussion of the work involved in the Netflix optimization prize, and Netflix's evolution of rating systems, see Clive Thompson, “If You Liked This, You're Sure to Love This,”
New York Times Magazine
, Nov. 23, 2008.

the “cheap talk” problem
: See Raymond Fisman and Edward Miguel,
Economic Gangsters: Corruption, Violence, and the Poverty of Nations
(Princeton, N.J.: Princeton University Press, 2008).

So where Netflix once relied
: For instance, John Riedl, who headed GroupLens at the University of Minnesota and created an early system to help people filter through the increasing torrent of Usenet articles, based on their ratings, told
The New Yorker:
“What you tell us about what you like is more predictive of what you like in the future than anything else we've tried…It seems almost dumb to say it, but you tell that to marketers sometimes and they look at you puzzled.” Riedl himself sensed some of the limitations of ratings-based systems, including how to get people to actually rate things. “Some researchers have proposed compensation systems that reward users for entering ratings. While the economic consequences of this solution are interesting, we wonder whether compensation would be necessary if ratings could be captured without any effort on the part of the user. We believe an ideal solution is to improve the user interface to acquire
implicit
ratings by watching user behaviors. Implicit ratings include measures of interest such as whether the user read an article and, if so, how much time the user spent reading it. Our initial studies show that we can obtain substantially more ratings by using implicit ratings and that predictions based on time spent
reading are nearly as accurate as predictions based on explicit numerical ratings.” See Joseph A. Konstan et al., “Grouplens: Applying Collaborative Filtering to Usenet News,”
Communications of the ACM
40 (1997): 77–87.

“We find that upward mobility”
: Erving Goffman,
The Presentation of Self in Everyday Life
(New York: Anchor, 1959), 37.

“I'm actually a quite different person”
: I found the Horvath quotation in Hartmut Rosa,
Social Acceleration: A New Theory of Modernity
(New York: Columbia University Press, 2013), 24.

“an offensive strategy”
: Robert Trivers and William von Hippel, “The Evolution and Psychology of Self-Deception,”
Behavioral and Brain Sciences
34, no. 1 (2011): 1–56.

It wants to recommend things
: Which may or may not be the things you like. In a paper written before he joined Netflix, Amatriain noted that “modeling user preferences on the basis of implicit feedback has a major limitation: the underlying assumption is that the amount of time that users spend accessing a given content is directly proportional to how much they like it.” Xavier Amatriain et al., “I Like It…I Like It Not: Evaluating User Noise in Recommender Systems,”
UMAP: Proceedings of the 17th International Conference on User Modeling, Adaptation, and Personalization
(2009), 247–58.

“contraction bias”
: See E. C. Poulton,
Bias in Quantifying Judgments
(London: Taylor and Francis, 1989), 172.

“integer bias”
: See, for example, “A Better Way to Rate Films,”
Bad Films Are Bad
(blog),
http://​goodfil.​ms/​blog/​posts/​2011/​10/​07/​a-​better-​way-​to-​rate-​films/
.

“A great many people”
: Francis Newman, “Short Stories of 1925,”
New York Times
, Feb. 7, 1926.

While we might take this
: Chinese online movie reviews, to take one example, seem to be kinder, and more evenly distributed than their U.S. equivalents, perhaps, it has been suggested, because of a tendency toward consensus seeking in Chinese society and a subdued expression of likes and dislikes. See Nooi Sian Koh, Nan Hu, and Eric K. Clemons, “Do Online Reviews Reflect a Product's True Perceived Quality? An Investigation of Online Movie Reviews Across Cultures,”
Electronic Commerce Research and Applications
9, no. 5 (Sept.–Oct. 2010): 374–84. The authors note, “Western reviews are much more likely to be extreme over time, while Chinese reviews tend to have a much more bell-shaped distribution and newer additional posts are much more likely to be closer to the mean rather than more extreme.”

“Users are increasingly rating”
: See Yedua Koren, “Collaborative Filtering with Temporal Dynamics,”
Communications of the ACM
53, no. 4 (April 2010): 89–97.

Ask someone to re-rate a movie
: Dan Cosley et al., “Is Seeing Believing? How Recommender System Interfaces Affect Users' Opinions,”
Proceedings of the SIGCHI Conference on Human Factors in Computing Systems
(New York: ACM, 2003), 585–92.

People seem to rate things differently
: See “Statistics Can Find You a Movie, Part 2,” Web site article from AT&T Labs,
http://​www.​research.​att.​com/​articles/​featured_​stories/​2010_​05/​201005_​netflix2_​article.​html?​fbid=​IH3Z2Gar6-b
.

Perhaps as a concession
: The noise goes on and on. Netflix engineers, for example, must also wade through seemingly conflicting taste signals from the same user. The initial problem, the “household” account where everyone's taste—from
kids' movies to romantic comedies to action flicks—was mashed up, was easily solved by adding separate “profiles” for family members. But what about when a single person's account seems to oscillate all over the place? Netflix calls these “moods.” “You might be in a horror mood because your cousin is visiting and he loves horror,” Yellin said. Is that a “true” signal? How long should that signal last?

“a piece of praise”
: The discussion of Harvey Sacks comes from Camilla Vásquez's interesting study of the nature of complaints, “Complaints Online: The Case of
TripAdvisor
,”
Journal of Pragmatics
43 (2011): 1707–17. As she notes, “There is no doubt that complaints develop differently in an online forum, where people do not ‘know' one another in the same way they do in face-to-face interactions. As was illustrated, the differences in participant structure allowed online complaints to be simultaneously direct and indirect. With respect to other features of complaints, Heinemann and Traverso (2009) also claim that in face-to-face interactions, complaints require delicacy and implicitness because they make the complainant vulnerable, and that therefore, explicit ‘complaint-devices' like extreme case formulations, idiomatic expressions, and negative observations ‘only surface in extraordinary situations.' ”

“The difficulty of distinguishing”
: See George Akerlof,
An Economic Theorist's Book of Tales
(New York: Cambridge University Press, 1984), 22.

“The customers are seldom local”
: George Akerlof, “The Market for ‘Lemons': Quality Uncertainty and the Market Mechanism,”
Quarterly Journal of Economics
84, no. 5 (Dec. 1963): 941–73.

“higher relative share”
: Judith Chevalier and Austan Goolsbee, “Measuring Prices and Price Competition Online: Amazon.​com and Barnesand​Noble.​com,”
Quantitative Market and Economics
1 (2003): 203–22. While Amazon is not particularly forthcoming with its data—it refused my request for an interview—the authors were able to extrapolate data by using differences in review and ranking at Amazon's and Barnes & Noble's Web sites. “Of course, data limitations force our analysis to differ somewhat from the ideal experiment, as we discuss later. However, we observe the same books, their customer reviews, and a proxy for each book's market share at each site.”

Hotels were either responding
: Pádraig Cunningham et al., “Does TripAdvisor Make Hotels Better?,”
Technical Report UCD-CSI-2010–06
, Dec. 2010.

When all “known information”
: This description of the efficient market hypothesis is drawn from Burton G. Malkiel, “The Efficient Market Hypothesis and Its Critics” (CEPS, working paper 91, April 2003).

“The excising of the expert review”
: Suzanne Moore, “What Does the TripAdvisor Furore Teach Us About Critics?,”
Guardian
, Feb. 12, 2012.

“scattered about the world”
: José Ortega y Gasset,
The Revolt of the Masses
(New York: W. W. Norton, 1994), 13.

“Anybody who believes Yelp”
: The Reichl quotation is from Russ Parsons, “Ruth Reichl on Conde Nast, Gourmet Live, and Online Reviews,”
Los Angeles Times
, Feb. 11, 2013.
http://​articles.​latimes.​com/​2013/​feb/​11/​news/​la-​dd-​ruth-​reichl-​conde-​nast-​gourmet-​live-​online-​reviews-​20130211
.

Slippery though it may be
: See Balázs Kovács, Glenn R. Carroll, and David W. Lehman, “Authenticity and Consumer Value Ratings: Empirical Tests from the Restaurant Domain,”
Organization Science
25, no. 2 (2014): 458–78. The authors
created an index of keywords by which they attempted to quantify (on a scale from 1 to 100) the notion of “authentic”—with authentic itself generating a 95, “scam” netting a 4, and “decent,” as you might expect from such a middling word, sitting rather in the middle, at 51.

“because there is little”
: See Judith Donath, “Signals, Cues, and Meaning” (unpublished paper),
http://​smg.​media.​mit.​edu/​papers/​Donath/​Signals​Truth​Design/​Signals​CuesAnd​Meaning.​pdf
.

Even as it aggregates
: After filtering out reviews for various reasons, including those suspected of being fraudulent (which seems to also exclude many honest, but hyper-enthusiastic, five-star reviews).

But have the online review sites
: For an interesting discussion of TripAdvisor and the idea of trust and authority, see Ingrid Jeacle and Chris Carter, “In TripAdvisor We Trust: Rankings, Calculative Regimes, and Abstract Systems,”
Accounting
,
Organizations, and Society
36, nos. 4–5 (2011): 293–309.

In fact, Groupon users
: See John W. Byers, Michael Mitzenmacher, and Georgios Zervas, “The Groupon Effect on Yelp Ratings: A Root Cause Analysis,” in
Proceedings of the 13th ACM Conference on Electronic Commerce
(New York: ACM, 2012), 248–65.

“Work increasingly isn't”
: Paul Myerscough, “Short Cuts,”
London Review of Books
, Jan. 3, 2013.

Nearly one-fourth of Yelp reviews
: Susan Seligson, “Yelp Reviews: Can You Trust Them?,”
BU Today
, Nov. 4, 2013.

A group of Cornell University researchers
: See “Finding Deceptive Opinion Spam by Any Stretch of the Imagination,”
Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics
(2011), 309–19.

When people lie
: As Matthew L. Newman and his colleagues point out, “When people are attempting to construct a false story, we argue that simple, concrete actions are easier to string together than false evaluations. Unpublished data from our labs have shown a negative relationship between cognitive complexity and the use of motion verbs (e.g., walk, move, go). Thus, if deceptive communications are less cognitively complex, liars should use more motion verbs and fewer exclusive words.” See Newman et al., “Lying Words: Predicting Deception from Linguistic Styles,”
Personal and Social Psychology Bulletin
29, no. 5 (May 2003): 665–75.

Curiously, the researchers noted
: Myle Ott, Claire Cardie, and Jeffrey T. Hancock, “Negative Deceptive Opinion Spam,”
Proceedings of the 2013 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies
(Atlanta, June 9–14, 2013), 497–501.

When I ran my imagined
: It did, however, flag my use of the rather vague word “place.”

At one online retailer
: See Jessica Love, “Good Customers, Bad Reviews,” KelloggInsight, Aug. 5, 2013,
http://​insight.​kellogg.​north​western.​edu/​article/​good_​customers_​bad_​reviews/
.

“brag and moan phenomenon”
: See Christopher S. Leberknight, Soumay Sen, and Mung Chiang, “On the Volatility of Online Ratings: An Empirical Study” (10th Workshop on E-business, Shanghai, 2011).

“artificially high baseline”
: Byers, Mitzenmacher, and Zervas, “Groupon Effect on Yelp Ratings.”

when a similar property
: See Georgios Zervas, Davide Proserpio, and John Byers, “A First Look at Online Reputation on Airbnb, Where Every Stay Is Above Average” (Jan. 28, 2015),
http://​ssrn.​com/​abstract=​2554500
.

Similarly, on eBay
: See Judith A. Chevalier and Dina Mayzlin, “The Effect of Word of Mouth on Online Book Sales” (NBER, working paper 10148, National Bureau of Economic Research, Cambridge, Mass., Dec. 2003).

“minimum service standard”
: Geoffrey Fowler, “On the Internet, Everyone's a Critic but They're Not Very Critical,”
Wall Street Journal
, Oct. 5, 2009.

“Seems like when it comes”
: “Five Stars Dominate Ratings,” YouTube blog,
http://​youtube-​global.​blogspot.​com/​2009/​09/​five-​stars-​dominate-​ratings.​html
.

The most helpful reviews
: See Pei-yu Chen, Samita Dhanasobhon, and Michael D. Smith, “All Reviews Are Not Created Equal: The Disaggregate Impact of Reviews and Reviewers at Amazon.​com,” May 2008,
http://​papers.​ssrn.​com/​sol3/​papers.​cfm?​abstract_​id=918083
.

On Amazon, reviews
: Ibid.

When the performance is over
: Computer scientists have tried to model the so-called standing ovation problem. One model posits a rather simple formula: “Each audience member uses a majority rule heuristic—if a majority of the people that she sees are standing, she stands, if not she sits.” There are a number of variables to consider, however: Is the audience composed mostly of groups of friends? How much of the audience can an audience member actually see? Is there a time lag in which various audience members decide to stand? The standing ovation—and its extreme reverse, the cacophony of boos—are expressions of taste en masse and in real time but are seemingly open to conformity or social learning effects. See John H. Miller and Scott E. Page, “The Standing Ovation Problem,”
Complexity
9, no. 5 (May-June 2004): 8–16.