Is it possible to get a refund from a publisher if I bought an article that contains a major flaw (i.e.,one that invalidates the main results or the main conclusions)? Assume the article was bought through one of the main academic publication's paywall.
Of course you can request one - just send an email. You'll brighten up the publisher's office for the afternoon, and they'll be chuckling all the way home.
So just as long as you don't seriously expect to get your money back, you'll be fine. When you paid your fee, you were paying for access to read the paper. As long as that access was provided, then the publisher has met their side of the bargain. Peer review is not a guarantee of correctness. It's a first-level junk filter, nothing more.
And on top of you getting what you've paid for, you've got the additional bonus that you've now got a new paper that you can write, that's practically already written itself; the paper where you take apart your bought paper's argument. Magnanimously, the publisher won't require you to pay extra for this bonus - you got it thrown in for free.
You can ask for anything you want, but you will not get a refund. The terms and condition of the site where you purchased it most likely clearly stated that it wasn't an option.
One example from Taylor & Francis:
The content in this site is provided "as is" and without warranties of any kind either express or implied.
Taylor & Francis Group Ltd do not warrant or make any representations regarding the use or the results of the use of the content in this site in terms of their correctness, accuracy, reliability, or otherwise.
And another piece of sound advice:
If you do not agree to these terms, please do not use this site.
More generally, the money you pay for the article is strictly for its publishing, the actual research that resulted in a "major flaw" was paid via other channels, on which it's very unlikely that you'll have any direct influence. The people making the original claim about the content of an article, and taking responsibility for it, are the authors not the publisher to which you paid the fee.
If you witness that a given journal has consistently low acceptance standards, make sure to notify your institution's library, they might consider resigning subscription if they gather enough similar complaints.
Sadly, for the occasional 20 bucks you consider wasted for that article (I'm sure this is purely fictional, there are many ways of getting subscription journal's content without paying for it) there are 10 unscrupulous scholars, somewhere who use 3000$ of your tax money to publish complete rubbish in an open access journal. And here the perspective of a refund is nonexistent.
The publisher does not guarantee that the article has no major flaw. Indeed, for all they know, the reason why you requested access to the article in the first place was to refute it. Or to check other people's criticism of it.
Also, consider how many articles there are that are wrong yet of great historical interest. One example that pops into mind is the papers Einstein published before November 1915, containing his "work in progress" on general relativity, including some blatantly flawed thoughts.
So no, I do not believe you are entitled to a refund. You certainly don't have a legal basis, but I don't believe you have a moral basis either.
I can, however, imagine a scenario in which you might be legally entitled to a refund: if the article in question was based on fraudulent research, and the publisher was complicit in the fraud (e.g., the paper remains available for a fee even after it has been demonstrated unambiguously that it represents fraudulent research). Respectable publishers retract such papers. Other publishers... well, good luck with them.
There is a case where I think you might have a very strong moral claim even if not a legal one: if the paper is retracted, there is no question that you ought to be able to obtain a refund.
Addendum One argument contrariwise: The last science needs is any further disincentive for publishers to retract papers. Maybe better for the enterprise as a whole to have a few readers get cheated than to have fraudulent and otherwise untrustworthy papers remain in the literature.
There's only a few narrow reasons you might demand a refund in this case: first, if the download/manuscript was somehow damaged or malformed, e.g. pages missing, or somehow "broken"; secondly, if the precis you used to make the purchase decision described something fundamentally different than what the paper delivered; third, if there was something fraudulent about the paper.
Being "incorrect" or "false" is different than being "fraudulent."
To be successful in a request for a refund, either the publisher would have to be remarkably generous, or they would need to conclude that it is in their best interest. I'm skeptical that appeal to threatened "public hue and cry" would persuade them, since academic publishers are generally immune to negative publicity. A potentially more powerful motivating force would be legal pressure, for which I see two bases. One would be fraud, where the publisher knowingly represented falsehoods as truth. It would be challenging to establish fraud. The second is via warranty of merchantability (in the US enshrined in law via the Uniform Commercial Code article 2). In some cases, e.g. Taylor & Francis (following the lead of every known software producer), publications are offered as-is, in which case you have no option. Any product without an as-is disclaimer carries a warranty, to the effect that if the seller knows (or should know) that the item is used for a particular purpose, then the item is fit for that purpose. This is why one can sue a company for negligence, if they sell something as fit for a purpose when it is not.
In order to make any traction with a threat for violating the implied warranty of merchantability, you would have almost certainly have to enlist the aid of lawyers willing to help you pursue the matter. The two main things that you would have to objectively establish is that the article has a known purpose, and that the article actually is not fit for that purpose. I don't believe that you would have to establish that the publisher was aware of the defect (I'd like to be more assertive about that, but that's what lawyers do).
If you just email them asking for a refund, they will probably say that they don't make any guarantees. However, if your attorney uses suitable language, they would probably reply that they have a stated as-is policy (you're out of luck), or, that there is no way they could know of the mentioned purpose, or that the article is in fact fit for that purpose. Since "merchantability" isn't defined under the law, courts would generally compare the item in question against comparable products. The reason is that no product can be absolutely flawless, and a claimed based on merchantability would have to show that the product was egregiously defective, not just less than satisfactory.
That all said, you did not actually purchase "the article", you purchased a license to copy the article in a particular manner, for which reason it is not clear that UCC article 2 is applicable.
If you are just looking to get your money back, then I am afraid it is not worth an effort; you probably will spend more on the telephone calls than they charged you for the article.
However, it could be nice to set a precedent like this: to request a refund and eventually to sue a lazy publisher, who do not bother to introduce a proper level of scrutiny in their peer review process, and instead charge the authors to publish some bogus papers and then charge the readers to access them. I have no idea whether or not such a case can win (and I guess it depends on jurisdictions). The question just adds to the bigger question of very complicated relations between academic publishers and academia.