Artificial intelligence is the future of interactive media

The fifth Black Mirror season – made up of a single episode called “Bandersnatch” – represents the latest mainstream offering in interactive storytelling. Released in December 2018 by Netflix in a choose-your-own-adventure format, Bandersnatch allows viewers to make decisions at various junctures – these choices then determine the story path down which the episode proceeds.

The result is what has been described as a network of “five endings and one trillion story combos”, including, apparently, some scenes that nobody can find. The Independent described this storytelling format as “groundbreaking” while The Guardian proclaimed that the “TV of tomorrow is now here”. On Twitter, fans gushed about its “amazing” storyline and how #Bandersnatch is “a genre-defining piece of art”.

But the format is actually not a particularly new idea: recent (and not so recent, as we will see) history is littered with precedents. Netflix itself presented similar interactive episodes in its 2017 children’s shows Puss in Books and Buddy Thunderstruck: The Maybe Pile. While well-received, they remain in the sphere of children’s shows, with the format presented as a way of gamifying TV, arguably “to make kids TV shows even more addictive”.

But this interactive format appeared even earlier. Back in 1967, the Czechoslovakia pavilion at the Montreal Expo exhibited a film called Kinoautomat. Created by Radúz Činčera, the film was screened in a specially built cinema hall with a green and red button installed at each seat. At various points during the film, the reel stops, and a moderator appears on stage to ask the audience to choose between two scenes. The audience’s votes are then reflected in green or red lights around the screen, and the scene with more votes is played.

No matter what choices were made, the film always had the same ending (a burning building). Arising out of 1960s Communist Czechoslovokia, Kinoautomat thus not only made a statement about the validity of branching story path structures – but also a profound one on the validity of political choice and the nature of democracy.

More recently, the interactive web series, Try Life, created in 2012, presents similar episodes on teenage life. Taking on problematic life issues such as abuse, drug use and violence, a choice could quickly end up in chaos or redemption. Like Kinoautomat, the format underscores how choices are not frivolous entertainment. A mistake could lead to terrible consequences: one cannot “try” life.

Even the time-bar in Bandersnatch – a rectangular block presented alongside each choice which shortens in length to signify the amount of time left to the viewer to choose (before Netflix’s algorithm does it for them) – has been seen before. In 2016, CtrlMovie AG released a smartphone app on which viewers can play an interactive movie called Late Shift. It displays exactly the same feature.

All about algorithms

So what is different about Bandersnatch? Its interactive storytelling format has been done before – and better. Instead, it’s the less glamorous issue of Netflix’s back-end programming that makes Bandersnatch an exceptional storytelling experience. What Bandersnatch really shows is how Netflix’s algorithms are able to deliver an unprecedentedly seamless experience of processing and presenting branching story paths for individual viewing experiences on a platform streaming to more than 100m subscribers worldwide.

When you think back to Kinoautomat in the bespoke cinema hall – where at each stop a projectionist in the backroom had to set up the chosen reel in a hurry to play the relevant choice – it’s clear how far we’ve come.

And this is where interactive storytelling will make its biggest breakthrough in the next decade – not by presenting interfaces of choices for viewers, but in how computational algorithms will be automated to such a degree of sophistication that they will be able to process and produce audiovisual media with which to tell our stories, in all the ways with which we as humans use stories to laugh, cry, think and make sense of our lives.

There is already evidence of algorithmic narrative power to demonstrate this potential. In 2016, IBM produced the first film trailer created entirely by artificial intelligence. To make the trailer for 20th Century Fox thriller Morgan a database of thriller trailers were fed into the IBM Watson computer. Through pattern-finding and other functions, the algorithm then selected music and scenes from the film to piece together a credible trailer.

Thanks to the accuracy of today’s computational language generators, it’s hard to work out whether some writing is created by humans or computers.

As consumers and citizens, we need to understand how computational automation is able to process language, emotion, morality, personality and other fundamental human traits. We need to do this quickly, because increasingly it will be these algorithms that are telling our stories in future, while we become increasingly passive partners. And we need to work out whether this is what we want.