Ask Slashdot: How Could You Statistically Identify The Best Sci-Fi Books?

jimharris writes: Over at SF Signal I wrote a piece "How Well-Read Are You in Science Fiction?" There are three databases that collect lists of popular science fiction books that try to statistically identify the best books of the genre, [offering] combined list that shows which books were cited the most. They use different sets of best-of lists, but their results are often similar. The final lists are, Classics of Science Fiction, Worlds Without End Top Listed, and Premiosylista Comparativas: Comparativas: Ciencia ficcion (Spain).
Interestingly, each list has a different book in its #1 position (though both "Dune" and "Frankenstein" make the top four on at least two of the three lists). But is this really a good methodology for determining the classic canon? What would be the best way to statistically identify the greatest sci-fi books? (And have you read any good science fiction novels lately?)

Comments

define your terms first

[Anonymous Coward]

You need to define best, greatest, and classic before you can go further in your quest.

Re:define your terms first

[Anonymous Coward]

And not in the least decide what the definition of the term science fiction is (it's not as easy as it sounds).

You can't

[110010001000]

Art/literature/music is subjective. You can't rank them, except personally. Next question.

Re:Not Dune

[__aaclcg7560]

Dune has faded badly.

What do expect from a novel written over 50 years ago? Go buy yourself a new paperback from Amazon. I usually replace my copy every ten years or so.

Re:Why Limit This Contrived Gimmick to Just SF?

[mwvdlee]

The problem is that you can't really state your preference.
If they offered a choice to say "thanks for the recommendation, but I won't be buying this. Ever", the recommendations could be improved.
On a site like Goodreads, you can state which books you like, and it uses that information to recommend others; worked quite well for me in the past.

Re:Why Limit This Contrived Gimmick to Just SF?

[smooth wombat]

Unfortunately, the local bookstores have gone out of business in my area. I can't spend an afternoon browsing the shelves to find something new.

Are you sure? Neil Gaiman just celebrated Independent Bookstore Day and we know he's an authority on such things.

Re:Looking backwards, not forward...

[__aaclcg7560]

So, the best you can do is just judge people by their individual merits instead of worrying about group X or Y.

Unfortunately, people will look at a book and decide not to read it because a women wrote it. As several commentators on Slashdot has already mentioned: "the best science fiction is written by men." In fact, some women writers wrote under a pen name because of this obvious bias.

http://io9.gizmodo.com/5077952/women-who-pretended-to-be-men-to-publish-scifi-books

Measure the subjective responses

[Bruce66423]

If a lot of people are wowed by product A and bored silly by product B, it is irrational to argue that the two can't be ranked as to which is the one most worth investing time in to read, given that we have a finite amount of time to spend doing so. Therefore to get people to vote for their 'favourite' seems a rational way forward, despite its subjective foundation.

Otherwise I know a wall with some paint drying that you can watch this evening...

Re:Measure the subjective responses

[Dutch Gun]

I think the trick in any sort of subjective ranking is to find a person or group of persons that tends you match your own personal tastes. Subjective ranking certainly isn't worthless when I recommend a book, movie, or TV show to my parents. That's because I can more or less accurately judge how well they'll like something since I know their tastes.

I think this is where most algorithmic approaches tend to fail (like on Amazon), at least from what I've been able to see. I think they tend to find general correlation - that is, "those who like book x would also like book y", but I think a much more effective approach would instead be to search for other customers who's general ranking patterns tend to match your own most closely, and build a personalized recommendation group from which to mine predictive data. In this way, the predictions would be more or less tailored for each individual customers based on similar likes and dislikes, rather than being based on general popularity trends.

For better accuracy, the algorithm should pay even more attention to statistical outliers. For instance, I generally love space-opera-y science fiction (Honorverse, Lost Fleet), as well as "harder" works, like The Martian. But I didn't care for the Heinlein I've read, so would wish to avoid more of him, and more important, more books in that general style. So, a clever algorithm would notice that trend of mine and find sci-fi fans who *also* didn't care for Heinlein, and give their recommendations a slightly higher bias based on that data point.

Generally speaking, any sort of non-personalized "ranking" is going to simply be a popularity contest within the target audience you select. There's nothing wrong with that, so long as people understand that's all it can ever really be. Which, of course, they won't.

Re:How many have been read by today's readers?

[werepants]

"Frankenstein", while undoubtedly having a huge impact on modern society is basically unreadable by modern readers

What? Seriously? Frankenstein, along with many similarly classic works like 20,000 Leagues Under the Sea, War of the Worlds, are often placed in the children's section of book stores and libraries. I know that I read many classics growing up for just that reason - I saw some cool covers with aliens and monsters and submarines and got my start in sci-fi that way. If I was able to read that stuff as a kid, it's frankly absurd to call it "unreadable" for current adults.