Amazon deletes 20,000 product reviews written by seven of its top ten UK reviewers after a Financial Times investigation found they were written for profit
The article on the FT site has been updated with info about the reviews being deleted, but if you're hitting a paywall, the original article is available un-paywalled on Ars Technica: Amazon’s top...
I'm not surprised? I've purchased a few things from the Chinese companies that either use two seemingly random English words for their name, or seem to translate their name phonetically into the...
Overwhelmingly, those products were from little-known Chinese brands, who often offer to send reviewers products for free in return for positive posts.
I'm not surprised? I've purchased a few things from the Chinese companies that either use two seemingly random English words for their name, or seem to translate their name phonetically into the alphabet with leaflets saying "Give us five stars for a free accessory!" One of them was horrible (action cam, kept corrupting the footage on my SD card), and I refunded it, left a review explaining my experience and the attempted voter manipulation, to warn the next potential buyer.
I'm actually a bit surprised they used bots, because people are susceptible to the bribery system I described, but I'm also not completely surprised. It's just the next level of the constantly running game of technological warfare between online places and people who wish to exploit them.
Yeah, I've gotten this, too. I looked it up and 20,000 products is something like 0.01% of Amazon products. This literally did nothing for users. Having a way to show only verified reviews (or...
with leaflets saying "Give us five stars for a free accessory!"
Yeah, I've gotten this, too. I looked it up and 20,000 products is something like 0.01% of Amazon products. This literally did nothing for users. Having a way to show only verified reviews (or whatever they call reviews where they know you've actually bought the product) would be a start. Though obviously, as you point out, even that can be gamed. It seems like there are a number of things Amazon could do to make their site more useful to users and help avoid some of these types of problems, but they seem to be actively choosing not to. Like others here, I'm starting to avoid Amazon whenever possible now because of it.
It's interesting that this actually worked. There's a couple of weird stories of people just casually emailing billionaires and getting actual replies/results, I guess this one can be added to the...
But Amazon has known about the activity on Mr. Fryer’s account since at least early August, when one user of the site emailed chief executive Jeff Bezos directly after his complaints had been ignored.
"Jeff Bezos received your email,” an Amazon employee later replied, pledging to investigate Mr. Fryer and the other high-profile accounts. A number of reviews highlighted were subsequently removed—but no broader action appears to have been taken.
It's interesting that this actually worked. There's a couple of weird stories of people just casually emailing billionaires and getting actual replies/results, I guess this one can be added to the list.
I somewhat shamefully use Amazon to buy a lot of my things simply for the ease of the next day delivery with Prime. Sifting through the reviews is always a pain in the ass, you can never tell...
I somewhat shamefully use Amazon to buy a lot of my things simply for the ease of the next day delivery with Prime. Sifting through the reviews is always a pain in the ass, you can never tell what's sincere or not. I normally tend to look at 1-4 star reviews, taking the 5 stars with a spoon of salt. I also use tools like review meta to try and sift out the fake reviews, but it's not always accurate. Fake reviewing is a plague.
I have recently thought about a possible remedy. I'm not sure I can regurgitate a proper recollection of my idea, or even whether the idea was sound to begin with, so errors are to be expected....
I have recently thought about a possible remedy. I'm not sure I can regurgitate a proper recollection of my idea, or even whether the idea was sound to begin with, so errors are to be expected. This originated when I was thinking about how utterly broken peer review can be both because it is anonymous and non-anonymous in the right places. Reviewers have little incentive to give a shit; a lemon paper they accept does not look bad on them, as they're anonymous. Combine with potential conflicts of interest and peer pressure if they're deanonymized, and the only thing keeping science straight is that the people doing it are underpaid people who do it for the good of everyone.
Enter trustable anonymity. A centralized entity (or decentralized if you can pull it off) keeps a ledger of who did what. You can complain to the DB if your reviewer was shit; You don't know who your reviewer was, but the DB does. Next conference, the chair asks the DB for good reviewers, and this reviewer has a mark on their record. We don't know from whom, but we know that someone complained about him. Abuse protection needs to be done by crowd agreement: Everyone complains about the same guy (even though they don't know his name), he's probably shit. Someone complains about everyone and their mother, we can discount his opinions. Someone's complaints line up with yours, we can trust them more.
We can also apply the same to product reviews. Naturally, the DB has to be properly independent. We can institute (similar to a web of trust architecture) trust anchors whose reviews we implicitly trust. The German foundation Stiftung Warentest could be one such. They made a science and a business out of objectively testing products. Otherwise, people you know personally could function as trust anchors. If someone correlates with your trust anchors, you can trust them a bit. If someone gives out suspicious reviews, they are less deserving of trust, for example because they 1* or 5* everything, or because they vote against what the trust anchors say.
The article on the FT site has been updated with info about the reviews being deleted, but if you're hitting a paywall, the original article is available un-paywalled on Ars Technica: Amazon’s top UK reviewers appear to profit from fake 5-star posts
I'm not surprised? I've purchased a few things from the Chinese companies that either use two seemingly random English words for their name, or seem to translate their name phonetically into the alphabet with leaflets saying "Give us five stars for a free accessory!" One of them was horrible (action cam, kept corrupting the footage on my SD card), and I refunded it, left a review explaining my experience and the attempted voter manipulation, to warn the next potential buyer.
I'm actually a bit surprised they used bots, because people are susceptible to the bribery system I described, but I'm also not completely surprised. It's just the next level of the constantly running game of technological warfare between online places and people who wish to exploit them.
Yeah, I've gotten this, too. I looked it up and 20,000 products is something like 0.01% of Amazon products. This literally did nothing for users. Having a way to show only verified reviews (or whatever they call reviews where they know you've actually bought the product) would be a start. Though obviously, as you point out, even that can be gamed. It seems like there are a number of things Amazon could do to make their site more useful to users and help avoid some of these types of problems, but they seem to be actively choosing not to. Like others here, I'm starting to avoid Amazon whenever possible now because of it.
It's interesting that this actually worked. There's a couple of weird stories of people just casually emailing billionaires and getting actual replies/results, I guess this one can be added to the list.
I somewhat shamefully use Amazon to buy a lot of my things simply for the ease of the next day delivery with Prime. Sifting through the reviews is always a pain in the ass, you can never tell what's sincere or not. I normally tend to look at 1-4 star reviews, taking the 5 stars with a spoon of salt. I also use tools like review meta to try and sift out the fake reviews, but it's not always accurate. Fake reviewing is a plague.
I have recently thought about a possible remedy. I'm not sure I can regurgitate a proper recollection of my idea, or even whether the idea was sound to begin with, so errors are to be expected. This originated when I was thinking about how utterly broken peer review can be both because it is anonymous and non-anonymous in the right places. Reviewers have little incentive to give a shit; a lemon paper they accept does not look bad on them, as they're anonymous. Combine with potential conflicts of interest and peer pressure if they're deanonymized, and the only thing keeping science straight is that the people doing it are underpaid people who do it for the good of everyone.
Enter trustable anonymity. A centralized entity (or decentralized if you can pull it off) keeps a ledger of who did what. You can complain to the DB if your reviewer was shit; You don't know who your reviewer was, but the DB does. Next conference, the chair asks the DB for good reviewers, and this reviewer has a mark on their record. We don't know from whom, but we know that someone complained about him. Abuse protection needs to be done by crowd agreement: Everyone complains about the same guy (even though they don't know his name), he's probably shit. Someone complains about everyone and their mother, we can discount his opinions. Someone's complaints line up with yours, we can trust them more.
We can also apply the same to product reviews. Naturally, the DB has to be properly independent. We can institute (similar to a web of trust architecture) trust anchors whose reviews we implicitly trust. The German foundation Stiftung Warentest could be one such. They made a science and a business out of objectively testing products. Otherwise, people you know personally could function as trust anchors. If someone correlates with your trust anchors, you can trust them a bit. If someone gives out suspicious reviews, they are less deserving of trust, for example because they 1* or 5* everything, or because they vote against what the trust anchors say.