Hacker News Reader

Hacker News

Comment by thenerdhead | Hacker News Reader

Parent Full thread

thenerdhead

I mean some straight-forward ones might be...

Accounts need to verify themselves somehow.

Accounts should be old enough before contributing reviews.

Accounts should only be able to review X amount of titles an hour/day/etc.

Moderators should have tools to identify influx of reviews and decide mitigation strategies.

Moderators should be able to put up a banner on a title that may be review bombed so the community can report suspicious reviews.

Etc.

These aren't honestly policing much. Just general common sense for these platforms. Also heavily needed on many other platforms that get spam in similar ways.

View on HN

The question is not whether they can detect an influx of fake positive reviews on a product, it’s what to do about it. How do you tell the difference between someone review bombing their own products with fake positive reviews and their competitors review bombing them with fake positive reviews to try to get them banned?

Maybe you decide that sounds too difficult, so they should stop banning listings for that and focus on policing the reviews themselves. But that actually turns out to be a very difficult problem in itself! Sellers of crappy products will often include notes that induce/bribe the customers into leaving fake reviews. A seller of cheap products can rack up thousands of verified purchaser fake reviews this way.

They can also do things like hire a botnet to purchase the product in bulk using thousands of hacked PCs all over the world and get those PCs to review the product, generating more fake verified purchaser reviews.

In the end, if you don’t ban sellers for fake positive reviews then you run into these tactics and it’s very hard to police individual reviews. If you do decide to ban sellers then you have the opposite problem of dealing with sabotage.

As for detecting “influxes”, that problem is far from trivial. If a review bomber is using a botnet of thousands of PCs then it’s quite trivial to stagger the timing of the fake reviews over a period of weeks. It’s not like they’re forced to push a button and have them all hit the review page at the same time.

Sounds like there is a need for identity verified reviews. Ie the account doing the review has provided some government issued id during their registration process. Then add in another criteria that they can’t post reviews unless their account is say 3 months old (or some timeframe).

You could still record unverified reviews, but hide them and their ratings by default until the user explicitly wants to see them.

Who is going to scan in their government ID and send it to a book review site except for the people on the pay roll of a book publishing company that requires their employees to register an account on their first day of work so they can "review" books in three months.

If modern computer games have shown us anything, people will go out of their way (grind, pay) for shiny badges by their names, fancy skins, etc.

I’m sure many reviewers will enjoy having a special badge against their name. Not all, sure. We don’t need everyone’s reviews in this situation.

Sorry but your idea is terrible. You're only incentiving the exact people nobody wants a review from with a bad analogy.

I'm a legit user of Amazon. My account is probably over 20 years old. I've ordered tons and tons of products. I've never reviewed anything, however I have received products containing "gift certificates" promising me a $10 reward for leaving a fake review (of the product I just purchased). How does Amazon stop me from accepting this bribe?

I'm unsure if I'm playing devil's advocate or not, but should that not be allowed? That seems more about getting over the bias that people who are content with something are generally less inclined to leave a review than somebody who is unhappy with something. If you get something and it's complete crap, I doubt a $10 gift certificate is going to convince you to go write a positive review of it. If anything, that'll probably end up motivating you to write a negative review, with further mention of it.

If you get something and it's complete crap, I doubt a $10 gift certificate is going to convince you to go write a positive review of it. If anything, that'll probably end up motivating you to write a negative review, with further mention of it.

I think it is if the product itself is cheap enough. A $10 product where you get a free $10 gift certificate is essentially free. I think lots of people would be willing to leave a quick 5 star review in exchange for getting the thing they just bought for free, even if the product is kinda crappy. As long as it’s not totally broken and useless!

False flag operations? Amazon slips their own false review gift certificate in. If you use it, they flag you as an ignorable reviewer (shadow ban or whatever).

What incentive does Amazon have to do this? Best case scenario: they ban and alienate a bunch of customers who bought stuff on the site. Worst case scenario: they spend tons of money on these false flag operations and end up removing a lot of bestselling products from the site, reducing sales overall.

It would be a review shadow ban only. Wouldn’t stop them buying anything or leaving reviews.

I doubt it would cost a lot of money from Amazons perspective.

But yeah I agree Amazons incentives are not the same as their customers. They obviously have a lot of insight into what drives and lowers sales.

Maybe, although I seriously doubt it, at some point they will see enough reduction in sales to take action against poor quality products and their mispresentative reviews.

You could report the bribe to Amazon, now and then. (If you believed they would do something about it.)

The overwhelming majority of review bombs are from real people, of which ID would do nothing. Steam has shown a really simple solution the problem - a mixture of proof of purchase, and then algorithmically walling off 'irregular activity.' Review bombs on Steam remain relatively regular, but now have basically 0 impact - even extremely large scale ones like when somebody gains the ire of Chinese gamers.

Proof of purchase would require some modest effort on the part of publishers to create a workable and shared system, but it should otherwise be trivial to get going. Publishers have a strong motivation to opt-in since fake reviews are probably more detrimental than not to their bottom line.

Once someone is identified as a serial review bomber, then you’d just ban them and all their reviews. Or at least flag their reviews so they aren’t shown by default (similar to a shadow ban).

That person could no longer write verified reviews. Maybe it’s a permanent ban, maybe it’s a 1-5 year ban. They wouldn’t be able to verify another account with their identity until the ban is listed.

Yeah Steam is good, but mostly because it is a very closed loop system as they own the entire transaction-review-lifecycle. Not soo easy with other things like books.

You can’t identify the review bombers. These are 3rd parties hiring the reviewers to leave reviews. That hiring process takes place offline (from the perspective of your site), so you have no way to monitor and detect these directly.

As for the individual reviewers, there are tens of thousands or millions of these, and review bombers can trivially cycle these in and out, so no individual reviewer is a “serial fake reviewer.”

It’s trivial under current systems. But once you start placing higher value on verified identities, those millions of bombers are going to turn out to be a lot smaller.

It doesn't take millions to boost a specific product's sales. As few as a thousand fake reviews can get the job done. And bombers will find ways to corrupt the verified reviewers, since their reviews will go up in value. And the cost for Amazon to verify the reviewers is really important to take into consideration. If verified reviewers end up being untrustworthy, they've put in that effort (time and money) for nothing!

This is really just a special case of the wider issue of fraud. And I'm fairly convinced of the argument that the optimal amount is nonzero [1]. A further wrinkle to this problem is that fraud tends to only get more sophisticated over time, so the costs of fighting it always go up. At some point we may reach a scenario where fraud is so rampant that entire business models become non-viable.

[1] https://news.ycombinator.com/item?id=32701913

Naively verified reviews is a way to enforce "mostly positive reviews" - at least in legal-threat-happy countries like the US or the UK.

That Steam idea of giving the viewer more options on how to filter the reviews sounds great actually: Let reviews come in, but allow the viewer to filter against accounts that burst reviews, or by country, etc. Pretty cool idea (because in particular it allows privacy.)

Although really the current concept of "review" is terrible. Few reviews are detailed and specific enough and multi-dimensional enough. They cannot be searched well enough (Although we might be nearly there with LLMs - asking them to critique a book from our point of view and preferences.) It's hard to follow reviewers that you have learned to trust (hard enough to notice them in the first place). Star ratings are not useful in the intended way and only a few sites ask for multi-dimensional star ratings (generally asking to rate irrelevant things :-) Bleh!

Reply on news.ycombinator.com