Slashdot Mirror
Amazon: Authors Can't Review Books
In an effort to step up its fight against astroturfers, Amazon has barred authors from reviewing books. It's not simply that authors can't review their own books — they can't review any book in a similar genre to something they've published. "This means that thriller writers are prevented from commenting on works by other authors who write similar books. Critics suggest this system is flawed because many authors are impartial and are experts on novels." British author Joanne Harris had a simpler solution in mind: "To be honest I would just rather Amazon delete all their reviews as it... has caused so much trouble. It is a pity. Originally it was a good idea but it is has become such an issue now. The star rating has become how people view if a book is a success and it has become inherently corrupt." How would you improve the online review system?
Provide Data on the Poster rating based on the Star System and give an average and/or a list of all the Posters star reviews to provide a balance. This way you can see if a particular poster is always picking 1 star for anything or not.
There is absolutely no value in having random people review things. Criticism isn't a democratic principle.
Reviews are only valid from people that maintain that as their profession. There is a level of experience that comes with reviewing and editing that can't be achieved casually. Even many professional critics don't have this skills.
In each field, there are only a few peoples opinions that matter. The rest can be determined by demographic sampling.
I think online reviews are only worth anything when you have dozens or, better yet, hundreds. A few reviews are usually worthless.
You have a system that reviews the reviewers, allowing for weighted values of reviews. Not that slashdot users would have heard of mod points or metamoderating.
One blindingly obvious way to cut down on fake and artificial reviews: only allow reviews from people who have actually purchased the product.
Amazon already highlights reviews by people who have purchased the product, so the functionality already exists. Why not take the next step and only allow those people to write reviews in the first place?
Alternately, Amazon could allow anyone to write a review, but would only calculate the star rating based on purchasers' reviews.
Slashdot commenters not being able to moderate other's comments in stories they commented in?
Oh wait....
Knowledge is how to play a game, intelligence is how to win, wisdom is knowing what game to play.
http://xkcd.com/937/
This has often reflected my experience an any online store (and for anything, not just books). People don't seem to employ much perspective when assigning an overall rank. I recently saw a one star rank given to an app where the review stated the app did exactly what it was supposed to do... but he wasn't happy a particular feature wasn't present.
#DeleteChrome
Perhaps they should mark the author and publisher accounts differently than the average population (similar to /. subscribers)? The viewer of the reviews could then see the bias (if there is one). Seems simple enough, as I do like having the Amazon review system in place.
I'm reminded somewhat of two pertinent XKCD comics:
TornadoGuard/937
Star Ratings/1098
Interesting how they're kind of at odds with each other, but both true.
Video meliora proboque deteriora sequor - Ovidius
Like ban them from Amazon reviews forever. So the next time someone posts a 1 star review because the editor didn't catch a comma or the kindle version is not formatted perfectly for the very first kindle device ever made.. I think it would clean it up some.
As an (newbie-ish) author, I resisted the urge to review my own book, but I had spent a bunch of time reviewing other books thinking that it would be a nice way to find people of like mind and thereby interest them in my own writing. All my work deleted, so it seems.
I have to stop making the mistake of using websites owned by big businesses.
Stross's blog said it all a couple days ago. For those living under a rock, he's a pretty good modern sci-fi / horror type author. Disclaimer, probably biased toward him for having similar religious beliefs.
http://www.antipope.org/charlie/blog-static/2012/12/understanding-reviewers.html
TLDR poorly done summary interpretation:
Dumb people don't like feeling dumb, so most 1 star reviews are illiterate trailer trash... and the writing quality clearly reflects it. Ignore.
Hard core fans will rate everything you do as 5 stars. Meaningless. Ignore. So he doesn't like my reviews. Whatever.
A U shaped curve indicates nothing about quality and everything about high impact, also the opposite n shaped curve indicates apathy and low impact.
So.... applied to the article, first, analyze the shape of the "star" curve. Next, toss out any reviews that appear to be written in crayon by illiterates. Toss out any review where everything the author has ever written gets 5 stars. Analyze the remaining reviews by content... "apathy words" in the 3-star column of the histogram are bad news, etc.
"Science flies us to the moon. Religion flies us into buildings." - Victor Stenger
Simple: Reputation of the reviewer.
First, don't let anyone review until they've had an Amazon account for at least six months and made at least three purchases (on different days) in that time.
Second, post the reviewer's name (their real name, not a handle). Don't like that? Don't review anything.
Third, don't allow people to review products they haven't bought through Amazon.
Fourth, if someone has more than ten percent of their reviews deleted as spam or abusive, block that account from any more reviews.
Fourth-and-a-half, if a product has a large percent of its reviews deleted, "lock" it to only allow reviews by much more reputable users.
I would relax those a little for simply giving a star rating rather than writing a review, but not by much. I would also use a weighted rating system, based on the user's average rating. Not only would this get around the "No-star Nancy"s, it would work to avoid the useless inverse-exponential ratings we see on 99% of products, thus moving the "real" average rating to a three - So a five-star product would really mean a five-star product.
But reviews online are certainly corrupt. I don't use the star ratings for anything, unless an item only has a few reviews and all bad, and rely almost entirely on the BAD reviews for everything I purchase. If the bad reviews follow a common theme, it's a believable problem, and if I care about that problem vs. the price of the item, then I look for another item. Honestly I put less faith in the good reviews than the bad ones, especially when they're all glurge - no book, no product is perfect.
Just my $0.55 (US inflation, 1774-2008, for $0.02)
Critics suggest this system is flawed because many authors are impartial and are experts on novels
I had problems with other authors of similar books writing bad reviews of my book on Amazon. It was pretty obvious who was doing it because every time I'd get a good review, a bad one would pop up a couple days later from someone who obviously hadn't read the book.
I'm not vain enough to think everyone who reads my book should like it but the neg reviews were sometimes disagreements about topics not even covered in the book.
It was very frustrating and I complained to Amazon. They didn't respond directly but a short time later the behavior stopped.
I pay attention to what readers like and don't like and make refinements based on their feedback, so I appreciate thoughtful feedback even if it's not positive.
That's our life, the big wheel of shit. - The Fat Man, Blue Tango Salvage
If Amazon has enough spare computing power and resources I'd suggest that they allow and even encourage everyone[0] with an account to review and rate stuff.
;) ).
;).
Then what they do is crunch through all the numbers and figure out groups that have similar preferences (how coarse or fine is up to them). These are now your "ViewGroups" (viewpoints, but "points" can be confusing and you can't trademark viewpoints as easily
Then OPTIONALLY have someone clever name the top[0] X largest of these ViewGroups appropriately and make them either explicitly or indirectly (and other unnamed ones[1]) available to people to use as they wish when they are looking for stuff. This is optional, you can still have people use viewgroups without explicitly naming them.
So if you are looking for a present for your grand aunt, you just select the appropriate ViewGroup[1] and then search for stuff or her. You might be less likely to find "Call of Duty" but that's a feature.
And if you are looking for a gift for someone who likes "50 Shades", you do something similar. Or if you're being naughty you pick something "opposite"
Lastly, ideas are easy, implementation and "polish" is difficult. Nowadays patents suck and slow down progress. I'm getting old, and lots of cool stuff just isn't happening soon enough. Too often it's because of stupid crap like "one click".
[0] You might need to tweak a bit for cleverer spammers and their sock puppets. The dumber spammers and sock puppets will just end up in their own ViewGroups that hardly anyone but researchers would use.
[1] You could have people go through a wizard/form to pick a ViewGroup (whether named or unnamed). The wizard could ask them what items are liked or disliked. e.g. "Likes 50 Shades", "Likes Hello Kitty", "Dislikes Harry Potter".
It seems to be the rage these days to knock any online review site. Restauranteurs hate yelp, authors hate amazon, etc. Guess what - nothing's perfect, but they're pretty good. Are the Amazon ratings perfect? No, especially for situations with few reviews. But who the hell doesn't already know that and take it into account?
I like amazon's ratings system a lot. You can tell a lot from the distribution of scores. You have the actual reviews you can actually read. They flag the most useful favorable/unfavorable review, and in my experience, they really are useful. They also aggregate commonly mentioned topics, so you can identify common themes with respect to a product - like a common defect.
Same thing with yelp. Sure, the scores can be skewed by hipsters, yuppies, or assholes who make it their life's mission to review things and be clever. But more often than not, I've found the reviews to be fair. And I've also found that reading a handful of positive reviews and negative reviews gives you a very good impression of a place - same with amazon.
So no, these review systems aren't perfect. But they're really good, and as a result I'm much more likely to actually like the stuff I buy than back in the bad old days when you waited until someone you know bought something you want, or bite the bullet yourself. It's much better now, and just because there's room for improvement doesn't mean we should throw it all away.
> The star rating has become how people view if a book is a success and it has become inherently corrupt.
The star rating system is riddled with rubbish like "The book arrived on time and was in good condition. Would definitely recommend this seller. 10/10" and often worse "This book was late and the damaged. 0/10".
That is, the Amazon rating system is a rating of the SELLER, and seldom the book! The uselessness of this has been pointed out to Amazon but instead of telling customers to review the book and not the seller (the easy way), they've added a stupid "Is this review useful?" button which doesn't fix the erroneous star ratings. Amazon have an awesome resource, a user-based rating of nearly every book on the planet, and they squandered it out of sheer laziness by their IT staff and management. Crazy they are cracking down on author reviews without fixing this.
You can automatically judge reviewers of large numbers of book by determining how well their reviews fall into a normal distribution.
That is - from a individual's subjective opinion, after reading 1000 books - should form a roughly normal distribution (the worst, the best etc).
So you weight a person's contribution to the overall score by their fit to the normal distribution on their own reviews and the number of reviews they have made.
A rough approximation for the purposes of a 5 star score is 1: 5%, 2: 15%, 3:60%, 4: 15%, 5: 5%. Of course these numbers could be adjusted to allow for equal numbers in each group if they wanted.
EMail: 0110001101100010010000000110001101110010 0110000101111010011011100110000101110010 0010111001100011011011110110