"Artist driven" books are the ones in which artists pay a bunch of money up front to get their book published, but everyone pretends very hard that it's not a Vanity Press operation.
I used to think that "well, if the publisher is well run and competent, it's just that they can do more with less cash" but then I got into a discussion, and it happened that I thought about it a bit more. The answer is "no, it is always a scam, it always bad for artists" with a couple of exceptional cases I will talk about at the end.
Let's throw some numbers up there. The actual numbers don't matter so we'll keep it simple. Suppose we have a typical mainstream publisher, Wiley or someone, who doesn't ask authors for money up front. Some books sells, others don't. The winners pay for the losers. Suppose this year they are doing two (2) books. Each book edition costs $100 to do, all-up. The publisher fronts $200 and gets the books out there, one sells the other does not. One book produces $500 in total revenue, the other $0.
The publisher pays back their cash reserve $200, and then divides the remaining $300 up between themselves and the successful author.
The unsuccessful author gets nothing, but is at least not in negative territory.
Consider now a smaller publisher, less well capitalized. They only want to spend $100 to get the two books done, so they ask for $50 from each artist. They're just as good at this as Wiley, so they sell $500 of one book and $0 of the other, same as in the previous scenario.
The publisher pays themselves back the $100, and (presumably) pays the successful artist back $50. Look, there's an extra $50 in play! There's $350 now to divide between publisher and the successful author. Even in this, which is nearly the best case scenario, the transition from traditional to artist-driven constitutes a transfer of $50 from the losing author to the publisher and the winning author. If the publisher is dopey and prints a lot of books that don't sell, the situation gets worse all around.
This is not underwriting in any meaningful sense, although publishers often use that word. This is, technically, parimutuel betting, quite a different thing, and arguably the opposite of what underwriters traditionally do.
It is the fact that rather than using their own capital, and spreading the risk across it, but are instead forcing the risk out onto the artists one book at a time, that produces the extra $50 in the publisher's pocket. They divide the necessary risk capital up into little tranches, and only pay back the winning tranches. Notably, the incentive here is to get as many losing tranches as possible, which isn't quite what we'd like to see. Every losing artist is another $50 we get to divide up between ourselves and Martin Parr. I am being, to a degree, facetious here. The benefit or loss of a whole pack of losers depends on a lot of fiddly details and could go either way.
As a side note, this is basically the standard playbook of capitalism: push risk out onto someone else, someone who is less qualified to evaluate it, and less able to assume it; keep profits for yourself. Uber, for instance, is essentially a taxi company with risks, ongoing costs, and capital investments pushed out on to people utterly unqualified to quantify and manage these costs; but with the profits nicely preserved inside the company, thankyouverymuch.
I see two ways around this.
The first is to make all books successful, which isn't as silly as it sounds. The artist-driven turns into helping the artist pre-sell enough books to cover their contribution. Run a kickstarter or whatever. Now, this turns your artist into an unpaid marketing consultant and shipping clerk, but that is a minor peccadillo compared to taking $10,000, $20,000, $30,000 off them which you may or may not get back to them.
At least some publishers are doing this to at least some degree. There may be publishers for whom this is in fact formal policy, but I don't know. It seems unlikely that there are many who would turn down someone's check just because it didn't come from pre-sales. Once the formalism exists to accept money from the artist, it seems to be difficult to turn down money just because it's not based on already-sold books?
The second is to treat artist contributions as a common pool, and pay them back in a common way. Rather than asking the artist to front $15,000 for their book, you ask for $15,000 for the overall production fund, and pay all contributors back at the close of the relevant fiscal year based on overall sales. This more or less amounts to creating a publishing co-op, which may be a thing that exists for all I know.