A few days ago, I was having dinner with a publishing friend in San Francisco, and we started talking about Google Book Search.  We were both tired, but enjoying our conversation, and my friend made a comment that engaged me; I concurred in its finding.

Neither of us use Google Book Search.  

That's interesting, because both of us read quite a bit.  Neither of us, notably, are actively engaged in academic research, which has a very unique pattern of search and discovery.  However, we both read a lot of serious non-fiction, and we do not find books, or investigate books, through GBS.  Generally, when I am looking for something to read, I am abstract- (aka, book reviews) and metadata- (aka references, citations) bound.  I do not search for books to read, or purchase, by doing open-ended full text searches, even when I have specific topics in mind.  

I think full text search is a wonderful thing, and the world is poorer without it.  My experience, and that of my friend's, suggests that it is not that relevant for purchasing.  Rather, the essential characteristics that Amazon and Powells offer (as does GBS) are sufficient: cited reviews by trusted third parties, and by readers; a browse-inside type functionality to determine whether the author can write with clarity and engagement; and straight-up search on basic metadata such as author and title.  

As I mentioned earlier, I have no desire to draw too many conclusions from this.  For research, navigating through full text search will become a normative expectation.  However, even there, the nature of the "book" as an object frustrates our naïve, idealistic visions of books talking to each other.  A moment's thought draws into critical light the relative worth of deep content interlinking.  

Books - even non-fiction - intrinsically present poorly structured information.  They are about as far as one can get from a RDF tuple: meaning must be interpreted and mined.  As a searcher looking for information on a topic, full text search against books generally serves me poorly; it may be a very useful adjutant for deep inquiry, but for drive-by or casual queries, it is a frustration.
 
For example, if I was interested in a general topic such as Ronald Cartland's role in the anti-appeasement discussions in the House of Commons in 1939, I benefit from finding books that discuss this.  However, if I am particularly interested in the date of Ronald's Cartland entrance into the House, or summary information on his relationship with the famous romance writer Barbara Cartland, then a search against the main Google index is far superior.  

There are a couple of forks from this.  The first is that the amount of effort to intertwine book content with other data, such as temporal, geographic, or biographical thesauri, might have relatively little real-world usage (oh, to see the logs from GBS ...).  Aside from the Curious and a few deep content divers, how useful is mixed-content integration for discovery against books?  Is integration perhaps more useful against other forms of content, such as the inherently more structured and reference-like material found in open web pages?  

Second, what does useful cross-text navigation actually look like?  Going back to my previous example on appeasement, if I was interested in the outcome of the House debate on Germany's invasion of Norway and Great Britain's failure to respond aggressively, how would that actually appear to me, as I navigated my way through a GBS product?  What, in other words, becomes a target for hyperlink expansion, and what are its targets?  If I assume that "appeasement" was a topic that I wanted to navigate, what UI would I expect?  

What is difficult here is intentionality.  It is extraordinarily difficult to determine what a user's intentions are as they navigate and browse through a sea of text.  It is relatively easy to give them intellectual "snack food" - places cited in this book; a timeline; historical figures.  Those might drive clicks, and ultimately ad sales, but they might not actually help the user in their quest.  

We are dumb animals after all, most of the time.  We click on bright shiny objects, and are easily distracted.  Designing a product to best meet the diversity of a user's intentions is very different than designing a product to maximize revenue.  

There's one final question for which I have no answer.  Presumably, as Google in fact previously announced but has not advanced, it would be possible to commercialize GBS as a bookstore, with PPV or POD services; it might also serve as a source for licensed content and services.  Either or both might, from Google's perspective, be a desirable offering.  

But what is the relative value compared to integration with the main search experience?  

This is not a trivial question.  Google has to make very hard choices about what to maintain as a search silo, and what to integrate into the main index, at what level, and through what metrics.  Is the user better served by collapsing book content into the main index?  What are the consequences for GBS functionality?  We remember that index-level integration is the choice that Microsoft made with Live Books when they vacated the book digitization market.  

Google is under no compulsion to continue GBS as an independent entry point in the absence of market incentives that justify the myriad expenses associated with its development.  In the absence of sufficient direct-driven revenue, GBS might actually dilute advertising based income from the main index.  Because book "data" is so poorly structured, and browsing behavior does not always translate into tidy crystallizations of intent, advertising CPMs against GBS might be less than when books are combined with the main index.  (I'm not saying they are, just that certain types of user behavior might not generate valuable advertising targeting profiles).

I'm left with a lot of questions and no answers, but I think we need to consider carefully, as publishers, libraries, and search firms, what we are trying to achieve, and importantly, for whom.