In April, data-driven publisher Inkitt announced that it is partnering with Tor Books to release the first novel selected by an algorithm for publishing. Here, Inkitt's founder and c.e.o. Ali Albazaz explains why he believes that AI is reaching a tipping point in the industry - but not in the way you might think.
In March a novel co-authored by an artificial intelligence (AI) algorithm moved into the second round of submissions for a national literary contest in Japan. What may have seemed like momentary buzz suddenly gave the publishing industry pause. Is technology capable of replicating the human process involved in creating something as powerful as the written word?
While a world where robots rank on the New York Times bestseller list is still light years away, the industry is starting to acknowledge the impact that AI is having on publishing.
AI may summon images of robots and false humanoid consciousnesses, but its influence will in fact prove both less glamorous and more pervasive. Technology is often accused of decentralizing systems currently in place, but with publishing, the opposite is true. In the next five years the use of aggregated data will empower the publishing pipeline to allow industry authorities to make more informed and fairer decisions based on what readers want.
Ultimately a data-driven process, rather than producing works of inhuman genius (although that may of course happen too...) will change the industry by empowering readers to drive what type of literature is chosen and released to market in an unprecedented way.
Unpacking reader preferences
It’s a difficult task determining what readers want. Traditionally literary agents and publishers selected novels that they perceived would resonate with readers as it wasn’t possible to track reader preference at scale. The popularity of book clubs and the use of focus groups have provided insight and direction over the years, but the numbers in these pools are not large enough to determine a broad range of reader preferences.
Everything changed with the introduction of e-readers and online reading forums. Suddenly people were not only reading, but engaging online, critiquing and discussing literature in group settings. Authors had a direct line of communication to readers and vice versa.
And with connected devices it finally became possible to track reader behavior at scale - allowing parties involved in the publishing industry to make more informed and smarter decisions.
For the last 400 years the publishing process has been built on the knowledge, experience, and instincts of editors and literary agents globally who diligently sift through millions of manuscripts to determine which one has the potential to become the world’s next best seller.
But this traditional system is by no means infallible, as those who work inside it would be the first to admit. The first Harry Potter book was rejected 12 times and Twilight 14 times; Stephen King’s Carrie received 30 rejections before being published. And how many great novels never hit bookshelves because of the decisions made by industry experts? How many authors give up after the first few rejections?
Publishers were using their gut feeling as there has been no tool - other than the inevitably biased and naturally risk-averse human brain - that could do better. But now technology has broadened the opportunity to ensure future best-selling novels are not overlooked. Data will act as an equalizer, giving every author a fair and equal opportunity to get published.
Data-driven publishing is just the beginning of turning the book selection process at publishing houses into a science. Instead of going through the pitch-rejection process, data will be able to identify genre, style and subject preferences for publishing houses based on reader response.
Imagine a world where you were guaranteed that a book would sell before it’s even published?
Publishing where everyone benefits
Publishing has always consisted of two parts: picking the right book and great marketing. With smart data we can optimize both.
All parties involved in the publishing process benefit from data-driven decisions when it comes to selecting a book.
Authors have a better chance of knowing if their novel has traditional publishing potential, avoiding the arduous process of sending manuscripts to individual agents and publishing houses, often to face multiple rejections. Readers have access to the type of literature they like and their preferences dictate what becomes available in the market. And publishers are guaranteed to have a best-selling novel through a selection process that tracks reader engagement and response.
Marketing also becomes streamlined and more effective with access to reader data that helps identify demographics and preferences when it comes to reading patterns.
The possibilities of technology in the publishing industry are endless. Progress is being made to uncover new ways to understand how readers respond to literature, to identify the specific passages that cause people to fall in love with a book and to unearth the world’s next best selling novels. One area we could guarantee accuracy in would be measuring the hormone levels of readers, which may sound like science fiction but is already being slated as a possibility.
Using behavioral data to select books for publishing is the best toolset we have available to us in 2016, and in the next few years the accuracy of the tools we have at our disposal will increase to the point where they become truly, well, intelligent.