Anxious about it all

[Nicolas TroisCentQuatorze]

Hi guys:

Somehow the whole idea sounds like an Orwellian nightmare coming to life.
Let me elaborate:

No matter what criteria you pile up to measure books, there are some things that will escape you. No metric can capture spirit, creativity, mood, depth of the feelings or character development. A book is made of much more than measurable quantities. How much it affects your spirit in terms of giving you ideas, making you feel sad, lonely, desperate, happy, proud, cannot be measured by any rational process, unless you can enter my mind and decide that the mess you see will match perfectly well with any given story. Books definitely have sets of measurable quantities of their own, but beauty remains forever in the eye of the beholder.

Let us take another look at it: imagine you get it right, you actually have some kind of measurement that people trust for book recommendations. Your system gets to be the prime book chooser for lots of readers, and a new author would only get read if it proves to score reasonably well on your list of criteria. Guess what happens next: authors writing books only aiming at getting the best score ever, regardless of the story they tell, the characters they develop, or the situations they invent. Next step is an algorithmic production of books: generate the instructions for an author to fill in the gaps with any kind of story. And this is where it gets Orwellian but we are not there yet, are we? 

Believe it or not, this is what happened to the world of fine wine-making. Most pricey bottles you will find today have a very similar taste, because the number of wine tasters has dramatically reduced to a handful who all happen to share the same tastes. You just cannot produce a wine today that does not please these guys, and most producers have understood it clearly. Net effect: spend 100$ on a bottle of wine and you will never be disappointed. You will never be surprised either. Fortunately, these guys do not taste the whole wine production on the planet so wine afficionados still have some places to get to.

Back to books: on the constructive side, such evaluation tools may prove very useful indeed once I get a correct feeling about what they measure. I can easily produce a list of say 5-6 books I loved, which I find similar in any way that pleases me, and match them against your criteria to see how they compare. If your classification can tell me why it would choose to put them together or not at all, I can start using it as a compass to steer towards new authors.

Similarly: a film critic who likes all the movies I liked is a good indicator for me. So is a critic who hates everything I loved, like a South-pointing compass. If I cannot find any pattern in a critic's decisions about movies, it is of no use to me. I need to understand the way the guy works to be able to draw something out of his articles.

Give me a database of objective measurements and let me use it as one of many indicators about books I have not read yet, but do not aim at telling me "You liked X you will like Y" because you have no idea why I liked X to start with.

Hope I am being constructive here 
Cheers

[Prince Mu-Chao]

I disagree that mood cannot be captured, it just obviously isn't at this time. By comparing entire tag clouds created by thousands of users and identifying tags specific to style and mood, an algorithm for books at a site such as LibraryThing could work. Currently, LibraryThings recommendations need a lot of work, but not as much as this one, which said that it was pretty sure I would enjoy reading the Patriot Act if I liked reading 1984. As knee-slapping as that is, it is not a very good recommendation.

[Aaron Stanton]

Nicolas, I want to respond to your comments (which were very constructive, don't worry) in depth, but before I do that I just wanted to throw this out there - 1984 matching with the Patriot Act is an easter egg, a bit of a joke on our part.  We haven't analyzed The Patriot Act, nor would it match against 1984 if we did.    The Patriot Act data is only returned when you do a match against 1984, and doesn't really exist in the database.

I would recommend you read it, but entirely because I think everyone in the world should know what sorts of things we've been passing into law... but not because you're going to have fun doing so.



Aaron

[Stephen Rollins]

Let us take another look at it: imagine you get it right, you actually have some kind of measurement that people trust for book recommendations. Your system gets to be the prime book chooser for lots of readers, and a new author would only get read if it proves to score reasonably well on your list of criteria. Guess what happens next: authors writing books only aiming at getting the best score ever, regardless of the story they tell, the characters they develop, or the situations they invent. Next step is an algorithmic production of books: generate the instructions for an author to fill in the gaps with any kind of story. And this is where it gets Orwellian but we are not there yet, are we? 

A valid point, but there's one major problem; there isn't one universal score that all writers can strive for.  There will be (potentially) millions of people using the database, and they'll have the same tastes in books that they would without using BookLamp.  Writers won't be able to aim for one specific style of writing because BookLamp doesn't rate books on how 'good' they are.  The only way a writer can take advantage of the system like that is if they manage to get a 100% match with every single book in the database; not going to happen. 

Otherwise I'd agree with you wholeheartedly. 

[Andy]

Nicolas I disagree, I think that the way it analyses the books does give a good pointer to the kind of book, this along with user ratings and reviews would give a good idea about the book and the feel of it.

I agree with Stephen as well;
There is no set good book, as Aaron as said somewhere it dfoes not tell you whether a book is good or not, just whether or not it will be suited to you if you like certain book, I like some fantasies like LOTR and harry potter, but also like other books like Stephen King and Dan Brown, also I quite like biographies from some of my fav celebrities. If I searched any one of these books I am unlikely to get a similar outcome, even between Stephen King and Dean Koontz, who write in the same genre but in a tottally different style.

That saying, it would be possible for writers to atract an audience by trying to hit certain criteria but that certainly wouldn't suite everyones tastes nor would it make their book an instant best seller, user reviews would still count a lot. I dont think anyone would completely trust a computer system to recommend books, they would look at reviews and stuff as well, before getting something, or as someone has said before they would read a small except out of the book.

All we can do is wait and see really!

[Nicolas TroisCentQuatorze]

Replying to Stephen and Andrew:

Selling books is a business, a big one at that. You can be sure that any successful future-predicting set of criteria will be used and overused as a measure for new authors who want to get published. Edition is a risky business, anything to reduce risks will be taken as a blessing, with potentially disastrous side-effects like described above.
As a simple reader, I would not mind having a set of statistical tools at my disposal to run on various sets of books I already know. What I do not need is a scale from worst to best book, for any kind of criterion.

Compare that to literature prizes: prize winners are not necessarily good books, "good" being defined by every individual for themselves. If I know that a book got a prize, I also want to know who gave it and on what criteria. Have I already agreed with the jury previously? Do I have a way to estimate how they rate literature and if this rating corresponds to mine? I guess what I am trying to say is that absolute measurements are fine, but relative rating is bound to have flaws.

[Aaron Stanton]

First, I want to point this conversation here:
http://beta.booklamp.org/forum/index.php/topic,25.0.html

Discussing a lot of the same issues, probably doesn't hurt to bring them together.  The line that made me think of it was your statement, Nicolas, about not needing a system to tell you good from bad.

Simply put, the line I wrote in the other post is that BookLamp will never attempt to tell you a that a book is good or bad.  What it can objectively identify, such as an amount of dialog, it will.  However, it does not tell you that the amount of dialog is good or bad.  There will never be a BookLamp award, for example - I'd encourage you to read the thread I linked to, because it explains that a little more.

But first, I want to reiterate that I think it would be very difficult to write to the formula without knowing how the formulas are calculated.  For example, one of the metrics we take into account is the pattern that different metrics score from scene to scene, meaning having a certain dialog rating isn't absolutely going to get you matched, but having a certain dialog rating in relation to the scenes that come before and after might.  What is or isn't considered a match changes depending on each user.

My point is, the level of abstraction that you'd have to operate in - without knowing each formula - would be very difficult.  The best you could do is take a book that is already popular and attempt to emulate it.  That's already happening.  Ideally, BookLamp attempts to spread out the reader base, so that those that want a specific, unpopular writing style can find it, even if no one else likes it - and hopefully create a market for an author that's currently only getting fringe attention.  BookLamp attempts to lower the barriers between trying new authors.

In my opinion, the far more threatening numbers are the ones that talk about what percentage for total book sales come from the top few authors.  I just Googled for it and couldn't come across the reference, but I remember it was very high.  By far, the top echelons of the writing community command the majority of the money spent on books, which leaves many, many deserving authors ignored because they don't have advertising dollars or name recognition.

I'd suggest that the infrastructure you're concerned about is already in place, except that it's not based on emulating an author, it's based on being that author.  Writing like the author won't do you any good, because the marketing structures of the publishing industry already prevent that from being effective.  Case and point, Stephen King wrote as Richard Bachman for several years, producing writing of equal quality but with far less success than Stephen King.  It wasn't until it was leaked that the two were the same author did Bachman's books really start to sell.  Then, advertising dollars and marketing took hold.  Ideally, two books of equal quality would find their way into the hands of an equal number of readers - the idea that an author could emulate style and sell books because of it... honestly, that seems like a step up to me.  Not perfect, but better than what's there.  *shrug*

I wonder if the existing stagnation is more damaging than a system that allows readers and writers access to greater amounts of data than is currently available.  I don't know.  But I think that you're interpretation is too simple, and only accounts for certain factors.

For example, I listen to a far greater variety of music by artists I've never heard of because of Pandora.com - I have listened to and have bought music by people I wouldn't have otherwise.  But it still has the same issues that you're worried about.

Anyway, all that was really just to say that relative rating is bound to have flaws only if you're judging on a scale of good to bad, which we're not.  If you have two piles of stones, and one has 3 stones and the other has 10 stones, there is no flaw in the statement, "The second pile has more stones that the first pile."  What would make that statement wrong - and one I'd morally disagree with - is if you were to add, "...and as a consequence, the first pile is worse than the second pile."  Letting you select whether or not you want to look at piles with 10 stones vs. 3 isn't inherently harmful, because we never attempt to pass judgment.  In the same manner, it's not inherently wrong to tell you that Book A has more passages in it without any dialog than Book B, though I understand your concerns.

You also make the comment that you have to have a way to judge if a rating corresponds to your own rating - well, that's easy.  We have no inherit rating system.  Until you provide the system with information about what you want to find - either by selecting a writing style to match or by manually adjusting your preferences - there simply is no way to make a match for you.  You have to tell us what you want, and then we'll use that data to find more like it, or ones different from it, or anything you want.

Anyway, that was a novel.    Those are my thoughts, at least some of them.  It's a complicated issue and I'm glad to see the conversation taking place.  We here in the team have had dinners discussing similar concerns - makes for good dinner conversation.

Let me know your thoughts,

Aaron

[Stephen Rollins]

Whew... amen to all that!  Took me a while to read, but I think it's an excellent answer.  You really can't answer a question like that anyways without a long post...  And I wasn't bored the whole time.    Er, the whole time, I wasn't bored... that makes more sense.

[Therin]

What, just bored some of the time?   

Seriously, I do have to agree with Aaron though.  The kind of situation that Nicolas describes does already exist in a way.  There are so few publishers that they are in effect your wine tasters.  So many writers are barred from publication because their work does not suit what these few publishers are looking for and suddenly the writer begins to tweak their work to better fit what they perceive as best-selling standards. 

Maybe the publishers project will actually broaden the tastes of the wine tasters by encouraging the flow of particular styles to more boutique publishers.  It could be this will allow for a flourishing of lesser known writers whose works will not be disadvantaged within BookLamp for want of word of mouth.

For all we know this idea could be an example of pure market forces - items being channeled and read/bought/published because they are what the reader really demands as opposed to what the publishing houses think we want. 

Then again, it could be a site that lets people know what book to read next...