The filmed entertainment business is sceptical at the idea of having a script analysed by a computer. The criticism goes: if algorithms get to decide which screenplays will be produced, and an automated process cannot grasp something as fundamentally human as a story, then AI is fundamentally unable to spot diamonds in the rough. This concern essentially boils down to: to what extent is an AI capable of casting judgement upon human creative works?

At ScriptBook, we acknowledge that computers are unable to experience an emotional response when analysing a script. That being said, what we can do is detect the underlying patterns that elicit such emotional responses — and ultimately predict them, despite a machine's lack of emotions.

"What machines lack in contextual information, they more than make up for in the sheer volume of data they can internalise when trained."

Consider image recognition: a neural network receives a flat, two-dimensional array of pixel values and must decide which of thousands of categories of objects is correct. Neural networks have access to none of the contextual information humans use — and yet they match, if not surpass, human performance in visual tasks. Last year's DeepStack AI beat humans at poker — a so-called "imperfect information game" where human intuition and psychology play decisive roles.

In storytelling, the advantage of big data is formidable. While a computer may not capture why we find a certain movie enthralling, it can compare a script against tens of thousands of others and draw parallels to story structures associated with success or failure. A human reader uses experience and gut feeling — not necessarily bad predictors, but certainly biased ones. A machine correlates the content of thousands of scripts to financial and critical success in an objective manner.

As for originality: stories that stand the test of time tend to share common structural elements. Georges Polti observed in 1895 that all dramatic situations can be traced back to one of 36 distinct prototypes. Even original stories seldom deviate strongly from this common core. Structure and story are inextricably linked — and structure can be learned by an AI.

Use case: Get Out (2017)

As an example of a highly original screenplay that follows typical story structure, consider the Oscar-winning "Get Out". The film follows a textbook three-act structure: an inciting incident, a first act establishing the protagonist and the world, a second act where things unravel, a midpoint warning, escalating trouble in the third act, a desperate low point, and a final reversal. What makes it original are the setting and the details — not the structure.

When parsing the script with ScriptBook's AI and plotting production budget versus "artisticness" — a metric differentiating broad-appeal blockbusters from niche arthouse films — our system places Get Out slightly towards the arthouse side of the scale. This is accurate: the film was made for $5M and targeted a specific audience, while performing far beyond expectations.

📊

Figure 1: Production Budget vs. Artisticness for Get Out

Upload the Get Out scatter plot here

Figure 1: Production Budget vs. Artisticness — ScriptBook correctly positions Get Out near comparable arthouse films like Moon and Boyhood.

Use case: A Most Violent Year (2014)

For 2014's critically acclaimed "A Most Violent Year", we compare predicted audience rating against artisticness. Our AI manages, based solely on the script, to recognise that A Most Violent Year is more akin to films like Ex Machina and The Grand Budapest Hotel than The Dark Knight. To a human this is obvious — and our AI has learned to make the same differentiation, despite having no notion of what "originality" is, by analysing structural elements alone.

📊

Figure 2: Audience Rating vs. Artisticness for A Most Violent Year

Upload the scatter plot here

Figure 2: Audience Rating vs. Artisticness — A Most Violent Year correctly positioned near comparable films.

Conclusion

Creativity tends to follow structure. At ScriptBook, we do not try to emulate a reader. Rather, we extract the defining elements of a story — characters, structure, dialogue style, interactions, dramatic events — and represent them as numbers. ScriptBook is not free of error. But then again, nor are humans. The kinds of mistakes ScriptBook makes are fundamentally different from those a human being makes — based not on gut, but on learned patterns. This makes it perfectly complementary to human intuition.

References

  1. Moravčík et al. (2017). DeepStack: Expert-Level AI in No-Limit Poker. Science.
  2. Yorke, J. (2014). Into the Woods: How Stories Work and Why We Tell Them. Penguin.
  3. Polti, G. (1895). Les 36 situations dramatiques. Mercure de France.