Summary: don't ask for one book, ask for a curriculum.
Big-list questions don't work well on Stack Exchange. This is not immediately obvious; at first sight you might think that with one item per answer, the best items get upvoted and come out on top, and all is well. In other words, you have a poll. In fact, this never¹ works, for several reasons.
- The main trend in votes is that whoever gets in first gets the most votes. A lot of voters only see the thread once, so they won't see later additions.
- The secondary trend is that most popular items get the most votes. This may look good, but it isn't. Items get upvotes based on their general popularity — “hey, I've heard of this” (if we're lucky, “hey, I know this, it's good”). Whether the item is actually suitable for the question is rarely a consideration.
- The one-item-per-answer tends to attract a lot of “read this” answers that don't bother to say why. Sure, someone can come later and edit a reason into the post. But that rarely happens, so we end up with lots of items with no clue of why we should pursue them.
- Often there isn't a One Single item that answers the question. Let's say your opinion is “as a tutorial, get book A and book B (they're equally good, so pick whichever you find first); and also get book C, because it the reference on the topic (but it's hard on beginners)”. You can't express this in the one-item-per-answer format.
Insisting that answers provide a complete recommendation mostly solves these problems. That is, if you recommend A and B in an answer, it should also be an assertion that A+B is sufficient for the question, that you don't need anything else.
- Whoever gets in first still tends to get the most votes. But that's not so much of a problem, because each answer makes a whole. There is value in having each answer individually, not just as part of a poll.
- Popular items still get more votes, but this is mitigated by having answers that tend to mix popular and less-popular items. The same popular item may be mentioned by several answers, and they'll be on equal terms.
- When answerers are encouraged to make a complete answer, they tend to provide a description of each item: “get A because … and B because … and C because …”. Answers can't be judged solely on the quality or popularity of the recommended item, the description and justification given in the answer also comes into play.
- Answerers are free to make their selection however they like, with no artificial constraint.
There are guidelines on this topic in the FAQ. In particular, going for a curriculum rather than a book is a big step towards meeeting the six guidelines on good subjective questions apply here:
1. Great subjective questions inspire answers that explain “why” and “how”. See point 3 above.
2. Great subjective questions tend to have long, not short, answers. Not one item, but a complete selection.
3. Great subjective questions have a constructive, fair, and impartial tone. Because you aren't advertising for a book, you're providing a reading list.
4. Great subjective questions invite sharing experiences over opinions. Surely you've read more than one book in your life, so draw on your experience, don't just take mention one thing you liked.
5. Great subjective questions insist that opinion be backed up with facts and references. See point 3 above.
6. Great subjective questions are more than just mindless social fun. A reading list calls for expertise, it's not something you can throw up in 30 seconds.
Now how does this work in practice? The now-defunct Literature Stack Exchange site tried allowing book recommendation questions with some guidelines that included an exhortation against single-item answers. The book recommendation questions weren't a great success, but some of them did work well. The ones that worked had some characteristics in common:
- Multiple answers provided a selection of several books with a paragraph or so explaining why each book was included in the list. Often the good answers would not be structured as a straight list, but had at least an introduction showing the philosophy behind the selection.
- Answers that didn't play ball were not upvoted (even if they happened to mention a popular book), or even downvoted or even deleted.
Making book recommendations work requires that enough of the community plays along, by upvoting the complete answers and downvoting the answers with a too small selection or with no justifications. It would help to have official guidelines allowing moderators to delete bad answers regardless of their score.
¹ For values of “never” meaning “almost never”. Let's say: it fails 99.9% of the times.