Awsum Shoes: Is it ethical to fix grammatical and spelling errors in reviews?

Innovation, the Internet, gadgets, and more.
May 10 2011 5:01 PM

Awsum Shoes!

Is it ethical to fix grammatical and spelling errors in Internet reviews?

Illustration by Robert Neubecker. Click image to expand.

You're about to go on vacation and you're looking for a hotel. You log on to TripAdvisor and read the following user review: "The Wi-Fi was atrocious and I had to pay $20 to watch ESPN, plus the shower lacked the heavy massage setting that I prefer. Room service was late and the minibar had only substandard vodka." A different commenter has the following opinion about a place across town: "Trenty, classy, understated, quiet and clean. bath towels big and thirsty, robe and slippers were cozee. felt wonderful after a hot showr." So, which hotel seems more enticing?

According to Panos Ipeirotis, a professor at NYU's Stern School of Business who studies consumer reviews on the Internet, the first review will lure more travelers. In a recent blog post, Ipeirotis discussed his research showing that well-written reviews help sell products, even when the write-ups are negative. Atrocious wireless connectivity? Who cares, so long as Wi-Fi is properly capitalized.

This is nice information for entrepreneurs, but it's worthless from a business perspective if you can't do anything with it; Ipeirotis compares it to "knowing that during the cold months the demand for summer resorts drops." But certain companies, Ipeirotis learned, are starting to understand the benefits of cleaning up their comments. Zappos.com now uses crowdsourcing to fix the grammar and spelling in the site's reviews. The specific tool Zappos uses is Amazon's Mechanical Turk, where anyone can assign HITs—human intelligence tasks—to people who want to make a buck while sitting at their computers. In order to Taylorize the task, Zappos broke copy-editing into discrete steps, using a crowdsourced method of word processing (PDF) that's based on bug-fixing routines in computer programming.

Ipeirotis estimates that Zappos spent hundreds of thousands of dollars fixing more than a million reviews. We can assume that this cash outlay translated into extra cash for the Internet's most popular destination for quality footwear. While the profit motive is clear, the ethics here are murkier. Is it appropriate for Zappos to fix its users' grammar? If a bunch of illiterate people like to stay at a certain hotel, well, that's information that would be useful to have. (No one likes to walk into a Jersey Shore situation without an adequate supply of tequila.) And, if someone can't spell the word awsum, we may be inclined to devalue his opinion of trailrunning shoes.

While the ethics of comment refinement are still ill-defined, the importance of consumer reviews is unassailable. Jared Spool, the CEO of User Interface Engineering, writes how Amazon earned billions on account of what was at first a subtle change to its comments system. In the early days of Amazon, reviews were posted in chronological order. With a book like Harry Potter and the Deathly Hallows, which has 3,620 reviews as of today, that system fails. The reviews that are interesting and useful (either positive or negative) will be buried by whatever's at the top of the page. And, with the zillions of products on Amazon, it would also be impossible to hire editors to choose the "best" reviews.

The introduction of a single question—"Was this review helpful to you?"—magically solved those problems. Now, we vote on what reviews are worthy and Amazon highlights the most helpful positive reviews. How does this turn into billions? "As we've watched Amazon customers make purchases on the site, we can clearly see that promoting the most helpful reviews has increased sales in these categories by 20 percent," Spool writes. "From this, we can project it has contributed to Amazon's top line by $2.7 billion."

Another detail from Spool's analysis confirms Ipeirotis' counterintuitive finding that well-written negative reviews can help sales. Spool again: "In our studies of Amazon shoppers, we found many start by looking at only the negative reviews, using them to try to 'talk them out' of buying the product." It's better if a product has more positive than negative reviews, but it's also worthwhile to have high-quality information around your product. One man's negative—"After taking delivery, I learned that the screen on this television was way brighter than I expected!"—is another man's positive.

Amazon's reviews have always been one of the site's most popular features—a testament to how human-centered we are even in the virtual realm. (While I often struggle to finish a terrific-though-lengthy article on longreads, I can somehow wade through 10,000 words of people discussing the pros and cons of a coffee grinder.) The reviews section of Amazon is also the playground of stealth marketers, people who seem just a little bit too informed about a grill they bought yesterday: "All 16 cook settings worked to perfection!"

Product review scandals abound: Companies have been embarrassed when their own employees compose raves for their products and have used the aforementioned Mechanical Turk to generate fake positive reviews. David Friedman, the photographer/blogger, recently noted a new strategy in this great game: the fake negative review. When he was looking for home theater accessories, he noticed that an "astonishing number of people who review speaker mounts happened to mention that they bought the mounts for their Onkyo HT-6100 speakers." And here is that negative review: "Not good for medium-large size speakers, not fully compatible with Onkyo speakers. I would suggest investing more and get better mounts." The not-so-subliminal message here: Onkyo, Onkyo, Onkyo.

When we read a review on Amazon, we have to administer our own version of the Turing test—was this really written by an innocent, human consumer like me? With the amount of money at stake, and the amount of PR energy brought to bear, it will become increasingly difficult to sort the genuine from the fake. One can imagine a future (perhaps it's already arrived) when companies deliberately insert bad grammar or regional slang to give reviews the appearance of authenticity—sort of like the distressed khakis of reviews.

For now, the trend seems to be going in the opposite direction. By cleaning up its reviews, Zappos is hurting shoppers as it helps its bottom line. The lowercase reviews, the all-caps reviews, the Internet speak, the subject-verb-agreement manglings, the sentence fragments, the pathetic attempts to spell chic—all of these are factors to weigh when considering someone's opinion of low-top Chuck Taylors. Or, to be more earnest about it, our mistakes are what make us human. On the Internet, it's important that other people can tell if you're an idiot.

Michael Agger is an editor at The New Yorker. Follow him on Twitter.