A published author expects to be paid: when a prospective reader buys a book, some money ends up in the author's pockets. Often, though, there's no money:
starving starting authors may be paid in exposure for their short stories, i.e., a magazine publishes them under the expectation of a win-win, where the magazine fulfils its duty to its readers (who pay), and the author's name and prose enters public knowledge . With such exposure, an author then hopes to attract commissioned (and paid) pieces in the future, and to become a household name, whose books are bought on brand recognition alone.
But there are other kinds of content. A company may wish to advertise its products or services to the broader public. One mechanism to do so was (hardly is nowadays) to buy space for an ad in a periodical. A more modern version is product placement: pay for the product to be present in scenes that a broad range of viewers will encounter, be that social media ("influencers") or in a movie or live show.
In academic publishing we encounter an odd situation. An academic pays a journal for publishing her paper, and at the same time, the journal sells subscriptions to readers. Following from above, this places the academic square into the advertisement business: an academic is buying space in a periodical to reach a broad public. But is this true? What sense does this make? What does an academic get out of this?
It is, in some sense, true. The logic could be as follows. Publishing in some journals is considered more prestigious than publishing in others. Scientists gladly pay what often are exorbitant fees  for the privilege of publishing in premier, or glamour, journals, which results in their work being received more favorably, directly impacting their chances at promotions and securing grants from funding agencies. Yet this logic quickly breaks down: strangely, academics also pay similar fees for publishing in just about any journal, famous or not. Turns out that merely publishing, in any venue, is considered good; publishing lots of papers is considered in some academic circles the hallmark of excellence, of success, with the number of publications being far more important than their content, in a blatant example of Goodhart's law, emerging from a simplistic approach to quantifying an academic's worth.
Who reads academic papers? Primarily, academics. Not academic administrators, who instead reduce academic "outputs" to numbers, in the hope of making wildly divergent and unique academic outputs comparable to each other. (Never mind that this is, in fact, impossible, as is predicting the impact of basic research.)
There are even more surprises. Unlike ads, academics can't just pay a journal to publish their manuscripts. Instead, these are submitted to peer review: the journal's editor is meant to be an expert, who happens to know other experts in the field and who commissions unpaid reviews of the newly submitted work. On the basis of these reviews, the editor may solicit additional work from the authors, or edits, or both; or reject the manuscript altogether. This modus operandi results in work being done for free by academic peers, which benefits above all the journal (i.e. academic institutions subsidize for-profit journals; ), but also, if the reviews are constructive and expert, the manuscript authors. When a manuscript is rejected, all that work has been done without any payment. The authors then may edit (or not) the manuscript and resubmit it to another journal, where it will be sent out for review again. The output of the review process is mostly random, and therefore submitting elsewhere in the hope of finding more sympathetic reviewers and editors is entirely legit .
Usually, only upon acceptance would publishers require payment for publishing the manuscript. Which means that accepted papers carry the burden of financing the whole operation of managing peer reviews, the submission systems, editors, etc. I've argued before that academics ought to pay when submitting a manuscript, not when the latter is accepted, therefore linking together costs with labor, keeping authors honest (don't submit a manuscript if it isn't ready) and keeping journals honest because accepting or rejecting papers would be unrelated to payment .
Under such a pay-for-manuscript-submission honest system you'd expect the cost per submission to be low. After all, there are far more submissions than accepted papers. A major publisher recently posited that such cost would be over $2000 (!!!), which is more than most journals charge for publishing an accepted paper in the first place.
In all this, remember, academics aren't paid a dime for publishing their papers. Nor are reviewers (also academics) for proofreading the manuscripts, evaluating their relevance, timeliness and significance, and for making constructive suggestions to improve the work.
But that's not all. Not content enough with authors paying to publish, reviewers working for free, and editors often also working for free (as all three roles are fulfilled by academics; rarely are editors dedicated professionals), journals charge readers via subscriptions . How much? Often, we don't know: the deals with universities and other academic institutions are secret. Careful investigation can reveal some of the costs, which often are in the millions of dollars per year. Wonder no more, scientific publishing is one of the most lucrative industries, with large double-digit percent profit margins.
Change in scientific publishing is needed. First, to stop the absurd and unnecessary costs: tax-payer funded research belongs to the public to begin with. Second, to bring about increased transparency in the evaluation process of submitted manuscripts, such as publicly posted reviews on preprint servers (an approach that, finally, a major journal is embracing: eLife, not surprisingly ). Third, to stop the absurdity of a manuscript "doing the rounds": when a journal rejects a manuscript, this is often resubmitted elsewhere, until finding a welcoming venue that is, hopefully, not a predatory journal.
After all, hosting accepted academic papers in an online archive isn't expensive, as the Journal of Machine Learning Research demonstrates. 
Academics are on the march, for change. But change is slow, apathy is high, engagement is costly in various ways (including time), and journals have a way to come up with "solutions" to match the funders requests for publishing in open access mode that is, frankly, astonishing. The gold open access model was meant to address the funder's mandate for open access, but, journals have been caught engaging in double-dipping: charging a reader for accessing a paper for which authors, in the first place, paid extra to make freely accessible. And what extra: a major publisher recently posted their options for gold open access, with individual manuscripts costing over $10,000 per manuscript. For that price, you could fund 1/3rd of the salary of a postdoctoral researcher for a year in an European research center. Or buy a car.
And let's remember: such absurdly high costs are for scientists to publish their tax-funded research for everybody to read, despite not getting paid any royalties, and with the journal in any case still charging subscription fees to academic libraries and individuals for the privilege of accessing their archive of past publications. As if electronic documents, their zero-cost copies, emails, newsletters, blogs, and the whole internet era had not happened.
Frankly, public goods such as tax-funded scientific research being
held ransom locked behind paywalls, as a result of a convoluted publication process that entangles dissemination of research findings with the evaluation of the researcher's careers, and where someone was spending someone else's money on someone else, is a situation becoming ripe for consideration under eminent domain laws, to legally recover what can only be considered stolen public goods. And in the process, send for-profit academic journals to the trash heap of history.
Did anybody ever considered paying academic authors for their published work? Yes, the tax payers did. And the charities. So there.
- Exposure, though, doesn't pay any bills, and is often a mere excuse for not wanting to pay authors.
- Some journals charge thousands of dollars, even ten thousand, for publishing your peer reviewed scientific paper. Scientists don't pay from their own pockets (or at least they shouldn't, even if some do out of desperation from lack of other funding); rather, from grants, or from their institution's funds.
- Read about the NeurIPS experiment with re-reviewing manuscripts.
- Predatory journals are named as such because they will take your paper, even any paper, even papers generated at random, for the sole purpose of charging the authors to publish their paper.
- Never mind that most, almost all, scientific research is paid for by taxes or by charities (who have deals with tax agencies and could be considered in any case a public good).
- The journal eLife was established as a joint venture by three major funders (the Wellcome Trust, the Howard Hughes Medical Institute and the Max Planck Society) precisely to upend scientific publishing.
- See more examples, including PeerJ which charges a one-time fee allowing any number of papers to be published.
- Not only academic institutions subsidize for-profit journals by paying the salaries and expenses of authors, reviewers, and editors, all of whom are almost always academics, but also by paying subscriptions to the journals themselves: as if an institution's members couldn't legally read the finished manuscripts reporting their own scientific fundings. Yes it's as absurd as it reads. Why does this system persist? Only because of the entanglement of scientific publications with the evaluation system for academics which, fallen prey to Goodhart's law (optimize for what is measured: number of papers in "high-impact" or glamour journals), judges the proverbial book by its covers and puts more value into both more manuscripts and more manuscripts accepted in select journals. In my view, if a scientific evaluator cannot understand the findings of a laboratory (their correctness, their timeliness, their potential impact at various levels), then he or she cannot possibly have any say in their evaluation. Counting papers is absurd; what must be counted is significance. The latter is hard, the former isn't, but that doesn't justify in any way whatsoever using quantity as a proxy for quality.