I’ve been guilty of binge watching both House of Cards and Orange is the New Black on Netflix while avoiding real work this summer–I’ve been hooked by the dark plots and HBO-style care put into each episode.
But amidst the hype, some believe that Netflix’s use of algorithms to create is an oxymoron, basically that creativity is threatened by adding a little math. That’s one of Andrew Leonard’s arguments in his Salon article “How Netflix is Turning Viewers into Puppets” and Evgeny Morozov’s echoing response in his Slate article “The Curse of ‘You May Also Like’.” As most are aware by now, Netflix created House of Cards as a result of compiling data that those who like BBC shows (this is a BBC remake) also like Kevin Spacy. Andrew Leonard (no relation) is suspicious both of the success of a data-oriented approach, but also “Can the auteur survive in an age when computer algorithms are the ultimate focus group?”
But critics have been spelling doom for the author at least since Roland Barthes. While “big data” can certainly reduce the human artist’s choice, I disagree that this sort of creation limits creativity, or necessarily results in a mindless repetition of themes or plots. Plenty of authors use a similar assessment of what is popular while sorting out their plots, weighing what’s popular when deciding the topic of their next work.
Edgar Allan Poe was famous for this self-conscious selection, and, while scraping to get by in an age when authors were paid pennies (if anything) for creative work, produced horror stories like “Bernice” that he claimed could be both popular and good. These two intentions, popularity and quality, aren’t as separate as snooty art museums and college courses on the classics might have us believe.
In his “Philosophy of Composition,” Poe revealed the “wheels and pinions” of his craft, and tried to dispel the myth that the artist creates with a wantonness for his audience, that he is slave to his own emotions rather than his reason. He described creating “The Raven” by deliberately choosing elements that would appeal to his audience: melancholy tone, check; the death of a beautiful woman, check; creepy bird, check. What Poe was doing wasn’t so far off from Netflix—they put together a list of popular items, mixed well, and served. He just didn’t have the benefit of big data to confirm his guesses.
That’s not to say that all art should be created in this manner, just that we shouldn’t feel afraid for the destiny of art by sometimes replacing human whimsy with cool calculation. In fact, maybe we should feel relieved. Formula fiction has been a staple in on the shelves at Barnes and Noble’s, and my guess is that the real formula for good art is more sophisticated than the human romance novelist or action movie director believes.
Take the “simple” fairytale.
In 1928 Russian narratologist Vladimir Propp methodically broke down the elements of the common fairytale and came up with 31 “functions” or events that happen in the fairytale and how they interact with each other: the hero probably leaves home, and, if he does, can victoriously return (with the help of a magical creature/device) (sometimes in disguise) and marry the princess (or milkmaid) (before the evil guy does). Even though there are only 31 functions, and most are somewhat dependent upon others, all the different combinations (there are sub-functions, and the tale can start or repeat in various places) make millions of tales possible. My husband, Brian, and I made a Propp fairytale generator based on these functions, which you can see here. Don’t expect to read this to a kid at bedtime through–it’s just the functions combined and recombined (if you press “refresh”) according to Propp’s rules, the bare bones of a story. I once tried to have Brian calculate out the number of possible fairytales in the Proppian model, without taking into consideration the “start over” clause Propp puts in, but it got too complicated and the answer was: ‘it’s a really, really, really big number.” In other words, we aren’t going to run out of fairytale plots anytime soon.
Knowing that should make us less afraid that a little recombination will result in more of the same; in fact, recombination with the help of big data can result in art the “auteur” may never have dreamed of, or may not have had the guts or support to do.
Take Dadaism. Dadaist poets using the cut-up technique would put jumble words or lines to randomly select for the creation of new poetry. In one sense this is the opposite of what Netflix’s big data is doing—where Netflix’s selection was calculated, this is random—but it’s also similar in that both strategies replace the human with something else, probability or popularity. The wizard behind the curtain isn’t human.
Replacement through pure probability has had mixed success. Nanette Wylde’s digital installation Storyland uses randomizing algorithms to create narratives that reset by refreshing the page much like my Propp generator. The results can range from nonsensical to intriguing, like this story I got: “On a very hot summer day, an angry man spoke out for a cause. The man was not of this world. Memories were rewritten. The man left a fortune for an average Joe. The average Joe also was not of this world and was filled with regret. When no one was looking, a retail clerk spoke out for a cause. The retail clerk was ruthless. Words were spoken. The retail clerk was scandalized by the man. The man forgave the average Joe.” A bit fragmented, but get Christopher Nolan to direct it, cast Benedict Cumberbatch as the angry alien, and it could be a hit. In fact, a little unexpected recombination could be much more original than the hackneyed, brain-dead romances, or beat-em-up action movies that are tried-and-true favorites. What if, just what if, big data puts two and two together in a way we didn’t expect? Maybe we’ll find out that people who like Russell Crowe also tend to like the Muppets? And these people also can’t resist a good Seafaring tale? Chances are no producer would take a risk on such a nutty sounding idea, but, hey, the math says so, so we might end up in a world where this is possible…
…and, as silly as it sounds, I’d much rather this sort of creativity be given license over another Nicolas Sparks movie.
But that leads me to my last reason why we shouldn’t be afraid of the loss of the human author: quality.
The quality of a piece of art isn’t solely on dependent on the combination of a few elements of plot or director/actor mashups, whether these are the result of a randomly generated algorithm, famous author, or popularity contest. A work of art takes more than that: both a winning combination and the skill to put it together.
Luckily, House of Cards and Orange have had both. House of Cards was indeed a remake, and not an original, and Orange is based on a memoir, but I think once Netflix gets comfortable with their formula, they could decide to take the leap and experiment with new plots, or other creative combinations. Netflix has the potential to learn one of two lessons from their recent technique, that, as Andrew Leonard fears, all it takes is the right combination of actors, directors, and a regurgitated plot to make a hit. Or, the much more subtle lesson that recombination based on big data can lead to new and unexpected results, that untested art has the potential to be wildly popular and break the mold, but only if sculpted with an artist’s patience and skill.