What follows is the text of a talk given by Dan Franklin for a seminar called Reading the Data: Informatics and Contemporary Literary Production, co-hosted by the Ambient Literature research project and Bath Spa University’s Centre for Cultural and Creative Industries

The title of this talk is ‘Should a Great Writer ever feed the dolphins?’

There’s a follow-up question: if they do, should they make and then eat the meal together?

Let me explain.

The questions I’m concerned with are: what roles can data play in the act of writing, in editorial workflows, and how can (or should) any data gleaned from audiences flow back into the act of literary production?

Or more fundamentally: Should a ‘great’ writer pay heed to their audience? Should they then go further and collaborate with that audience in the act of creative production?

There are numerous perils of writing in the age of audience development and amidst the new metrics of media consumption, such as the four-episode rule on Netflix (which is their benchmark as to whether a viewer will persist with a series until its conclusion).

In 2006, in an interview on NPR, the musician Tom Waits said: “I try to keep my audience a little hungry, you know uh, ‘Don’t feed the dolphins’ is my word. Next time you go out they’ll poke a hole in your boat.”

Tom Waits concert, July 2008, Prague.
Photo by Anna Wittenberg

For a twenty-first century example of this dilemma, Game of Thrones has deviated from the storylines laid out in the books, and publication of the next installment is stalled, but the television marches on as relentlessly as a White Walker.

George RR Martin finds himself in a fascinating 21st Century bind. It’s a bind defined by IP, exploited across multiple platforms and channels. Some characters are dead and some remain alive in his fractured universe, sections of the audience clamours for certain storylines and then reacts venomously when they are actually delivered.

How does data-informed storytelling worry the idea of “canonical” literature? Where fans’ wiki pages are sources of authority there is a strange reversal at work. In some cases, a book from an author might not be considered truly canonical if it doesn’t adhere to the strictures the fanbase places on it. Authors become a hostage to their own data and the way their audiences have interred it.

Here’s a mid-twentieth century example of why you might not want to do what your readers want: as related in Truman Capote’s In Cold Blood, Lowell Lee Andrews finished reading The Brothers Karamazov by Fyodor Dostoevsky, put the book down, shaved, put on a suit, took a .22 calibre rifle and a revolver, went downstairs and executed his mother, father, and sister. If that’s what your reader does after finishing your novel what element of that dataset do you want to carry forward into the act of literary production? How’s that for feedback?

The adjacent problem we are squaring up to is that we are seeing the advent of the mobilising and weaponising of robots that can write.

On Twitter, a platform which has achieved an Ouroboros-like characteristic of eating its own tail (a tail lately composed of vitriol, Brexit and Trump, and all the associated bots that whip up such fury), games writer and book editor Laura Michet proffered a pleasingly counter-intuitive argument against worrying about robots that write.

She said: “I am not convinced that total automation of written content will happen anytime soon in any industry. The biggest reason for this is that content automation takes time and effort, and written content is already practically worthless. The value of written content in and of itself has fallen so rapidly and so far that it is becoming almost impossible for most people to make a career of it.”

Conversely, in book publishing, break-out bestsellers are now few and far between – we often have one break-out new fiction title a year (if that). Arguably, we should not be asking whether the robots are a threat but whether the robots can save the industry.

Michet continues: “In the end, total content automation will be most interesting to people who want computers to say what no human would be willing to write or capable of writing. That is: words of immense brutality and unpopularity, words that are incredibly repulsive or boring to write, or words that no one will ever read, that exist only to trick or poke some other automated system or metric.”

She concludes: “This is even more disappointing and upsetting to me than the idea that writers will be completely replaced by robots.”

***

How about when a reader’s experience of a work is co-created with the author by inputting data in some way. Real-time thrillers in the forms of apps (and sometimes ebooks) have been around for some time, but will they ever ascend to become must-be-part-of-it cultural experiences? Do they need to?

A recent example is stopamurder.com, a new project from JA Konrath, who is a thriller and horror writer who had a lot of success as a self-published writer in the heady days of 2011-2016. He blogged prolifically about his experiences, his sales and income, and greatly successful career. He was also generous and even-handed about the experience and disadvantages of publishing this way.

It’s a symptom of the slowing down in the area as a whole that he’s largely stopped blogging for the last six months, but he did fire it up again to announce the new project last month.

Here’s the blurb:

This is unlike any mystery or thriller book you’ve ever read before. You play the sleuth, and as the story unfolds you will be tasked with solving puzzles to prevent a murder from happening.

In this five-book series, you’ll uncover the mind and motivations of a nefarious killer who is plotting to commit an unspeakable crime.

Each book contains an epistolary collection of emails, texts, and letters, delivered to thriller author J.A. Konrath, by a serial killer. This psychopath is sending detailed, cryptic puzzles and brain teasers that lead to clues about who will be murdered, why, when, where, and how.

Some of the puzzles are easy to figure out. Others are much more devious.

Do you like solving mysteries? Do you enjoy brain teasers or escape-the-room games? Are you good at spotting clues?

This series works best with an internet connection, using a color e-reader or app to enter answers on the killer’s website. A black and white e-ink device will work, but the interface will be smoother if used in conjunction with a computer or smartphone.

While each book in this series can be read and enjoyed on its own, the experience will be richer if read in order, and if the internet is used.

Over the five book series, you’ll need to answer more than seventy puzzles. When you answer correctly, you are rewarded with more clues that can stop a murder and reveal the killer’s identity.

Still paying attention?

Participating in this project requires flipping from the ebook (the body of the content, and where you find the emails and your riddles) and the website the murderer has set up for inputting your answers: http://www.stopamurder.com. The ebook is a series of exchanges that could be read as a standalone text, though played this way, you won’t know the final set of numbers that allows you to ultimately ‘stop’ the murderer.

It has clunky writing and is not brilliantly edited together. But there’s an interesting concept underpinning it. What’s intriguing is seeing a cross-platform experiment from the self-publishing scene like this, albeit one built on a solid track record of commercial success selling “regular” ebooks. Overall, the project is pretty absorbing and good fun. But it’s my job to find things like this engaging, so I don’t know how reliable I am as a critic.

One of my favourite ploys from Konrath is dropping in some of his other books in the serial killer’s emails to the fictional “JA Konrath”.

For example, the killer quotes a sample from Konrath’s own novel Whiskey Sour which forms the basis of a clue. There’s also this excellent comment in one of the earliest missives: “Your Jack Daniels novels (named after drinks, and I must say that’s a clever way to brand) all involve your sleuth outwitting the most depraved and heinous of criminals.” And then there’s this: “Do you prep, Joe? I read about it in one of your books. The one where Jack Daniels runs and hides in Wisconsin. Rum Runner, I think it was.”

If anything this kind of interactive, co-created thriller ‘where you are the detective’ is an amusing cross-marketing vehicle. It’s telling how often I’ve seen (and been involved with) ambitious paid-for digital product ideas re-engineered to become free, bolted-on lateral digital marketing initiatives. See also the pivot from product to services which so many book-based start-ups undergo.

There are some core questions lingering around this project. Do readers want to fuss switching between the browser and the ebook (which is essentially a website in a wrapper)? Do readers really want to break the box when ultimately it can be read as a standalone experience?

Why does this multi-platform stuff still feel so impenetrable and difficult to parse after all these decades of trying to integrate it into the conversation and data now allowing a flow-through of information. It’s STILL not obvious how to read any of this stuff. It requires patient explaining and even more patient understanding as the lengthy how-to blurb suggests. There’s not much room for patience nowadays.

***

The content data absorbed from reading a book (in print or digitally) and the environment in which you read it – content and context – amounts to an experiential cocktail.

I remember going to a production of The Winter’s Tale where an audience member was reading along with the play text in the front row and one of the actors jumped down and read his next line from it.

Scripts are interesting like this – the programme is the text of the play you’re about to see but the performance might include variants and glitches that deviate from the canonical version. Scripts are often products of teams in writers’ rooms, amended onstage or on set, and actors can improvise dialogue (famously in some cases). The ultimate, definitive version is slippery and its exact authorship can be questioned.

At this point, I should say something about the data we could be using when we track how people are reading. There are controversial editorial possibilities for authors willing to change what they’re doing in response to feedback gathered from digital devices.

Effectively the literary equivalent of test-screening, if you let a consumer panel read some books in advance of publication and they send you their reading data, it opens up lots of avenues for more mercenary plot-driven writers or instructional non-fiction authors to refine what they’re doing. (Business books are known to struggle to sustain their big ideas beyond the opening chapters.) What about corporate book clubs where business teams can read together and share comments inline, whilst the boss monitors everyone’s progress? You can’t pretend to have read the book if you haven’t anymore.

This is happening all over the web, and in contrast print books are perceived as more of a ‘black box’ technology where nothing enters and nothing leaves a dumb object.

But author practice is changing in this regard. Yuval Noah Harari, author of massive global bestsellers Sapiens and Homo Deus, has published drafts of his writing online in Hebrew for feedback, to create the marginal improvements that make his work better. Another example is (also massively-bestselling) Steve Jobs biographer Walter Isaacson, who in 2013 posted a passage of his work-in-progress on the origins of the personal-computing age first on Livejournal, then Scribd, then Medium where it blew up.  Authors have the vision, their audience the revisions.

In fanfiction, comments are currency. When an author willingly shares a work-in-progress they submit it up to the crowd, often one installment at a time. They hone their craft emulating others’ work in the same way bands develop by doing cover versions until (maybe) “They’re even better than the real thing”.

But ultimately this crowd-edited experiment has to end somewhere. In this case, the buck stops with the traditional figure of the author: “You can take this too far,” Isaacson said at the time. “There has to be someone in charge.”

This is a compelling refiguring of Roland Barthes’ 1967 essay ‘The Death of the Author’. (Yes, I’m dredging up this first year English literature undergraduate degree text in front of this audience, but forgive me). His theory posited that the meaning and import of a book is how it is received by its readers. Barthes expressed this pithily: “a text’s unity lies not in its origins, but in its destination”. But in the case of the feedback-influenced book, the book is shaped by the readership in the form of the crowd – sometimes editing and often responding to the text – but it is ultimately realised by the unifying figure of the author. Or, to put it another way, the buck stops with the showrunner.

***

The story we tell from raw data is more conclusive than the data itself. This is as clear as ever in the fractured, post-digital period we now occupy. Not only is the reliable, evidence-based story known formerly as the “truth” hard to grasp but all kinds of data about an intended audience can be targeted at will (notice how the language of online advertising borrows heavily from that of warfare) by bots and any and all forms of Artificial Intelligence.

Algorithms sharpen each other up, Artificial Intelligence improves by way of adversarial networks (effectively overseeing and vetting each other’s work) and their aim is to improve their output to the point where the role of editor – one which has always scored highly in surveys of the kinds of employment work that AI is unlikely to supplant – seems to be in peril.

But algorithms get it wrong all the time. The Bestseller Code, subtitled ‘Anatomy of the Blockbuster Novel’ sorted bestsellers by an algorithm created by the authors and determined that The Circle by David Eggers perfectly fitted the criteria to be the ultimate bestseller – this was a book which in reality had middling sales and provided a box office flop despite the film adaptation starring Tom Hanks and Emma Watson.

The proliferation of social media and use of digital platforms has probably narrowed and homogenized the culture we consume in western society. It’s like the iceberg is cracking so we group together on outcrops of ice, reaffirming each other’s possible survival and shared experiences. To slightly re-engineer this metaphor, the mainstream is becoming more main, while its tributaries and divergent passages become more various and less sustainable.

This means, in turn, a recycling in literature as much as in other forms of culture and media. All of this torrential data and how it might shape, inform or reflect culture is bearing down on us and it feels like we react by retreating back to known ground. Simon Reynolds put it beautifully in his book Retromania: “We’ve become victims of our ever-increasing capacity to store, organise, instantly access, and share vast amounts of cultural data” he writes. “Not only has there never before been a society so obsessed with the cultural artefacts of its immediate past, but there has never before been a society that is able to access the immediate past so easily and so copiously.” Data’s very capaciousness itself drives us to ration our intake and fall back on what we know, what feels comfortable.

***

Data overwhelms, it overflows. Perhaps this also accounts for the retreat back to the material world. In the face of all of this data is it coincidental that the publishing industry has rekindled its love for “beautiful books”?

Data is also manipulated and coerced, and what we are currently experiencing is a reckoning where non-human actors (for example bots) controlled by nebulous forces are influencing the discourse online. This moves from the spreading of misinformation, the derogation of honest, robust debate and the generation of crass and damaging content on platforms such as youtube.

There is useful data that flows and enriches like oil, and data like a monstrous sewer-bound fatberg which clogs up information systems. If writers are expected to make sense of the world by telling stories and truths about it, then unclogging this data blockage will be one of those challenges.

And then we come back to the audience. How real is the “audience” likely to be going forward? Are we in danger of creating artificial intelligence which makes authentic critical opinion, or simple and authentic audience feedback, impossible to glean? Is the future fake audiences posting fake reviews for fake books by fake authors?

There is a fast approaching, intertwined crisis of authorship and audience which we must face down. We are only just beginning to see the impact of data on the act of writing and how it is consumed, and the feedback loop that follows.

The crisis might become so grave that we cannot identify what is real and what is true – writers trade in truths and a corollary of that truth is the authenticity of the authorial voice. In this way, the teeming data of the ecosystem around our online selves threatens to stymie the development of literature in the digital realm. It is far simpler for the merchants of literary culture to adhere to the old ways: to grasp as tightly as possible to the boards of their hardbacks, inhale deeply of that delicious Vanillin scent and avoid the problem altogether.

We are left with the troubling possibility that data does not facilitate but instead slows and halts the development of literary culture itself.

But we shouldn’t give in to fear of what might occur, and instead, I’d urge us to continue to make work that shows the creative possibilities of new forms of literature. Show, by example, that the cultural benefits outweigh these potential risks.

Long live the author.

Thank you.