In a train compartment, Rick Sanchez meets a man who’s traveling to the end of the line. He tells Rick he’s there to meet “the man in the white coat”—like everyone else on the train—to settle a long-standing score. “The man in the white coat” is Rick himself, who soon learns that everyone on the train has a story to tell about him.
The stories that follow sound like prototypical Rick and Morty adventures, except Rick can’t recall any of them. The reason, it turns out, (spoiler alert) is because Rick and Morty are trapped on a “Story Train” by their antagonist the Story Lord, whose power is to generate an endless stream of stories that are nearly impossible to distinguish from reality.
The episode itself (“Never Ricking Morty”, a reference to The NeverEnding Story) is a self-referential take on Rick and Morty co-creator Dan Harmon’s Circle Theory of Story, a story generation algorithm that Harmon uses to develop stories for everything he creates. Thanks to this algorithm—agonizingly developed over the first few seasons of the show—Harmon has claimed that the writers of Rick and Morty could churn out plots forever.
While the satire is fun, this episode left me with an idea that continues to bounce around in my head, mingling with other long-simmering thoughts about interactive storytelling. The idea, in summary, is that of a Turing test for stories.
Stories are important. Numerous studies have proven that templates like the story arc and three-act structure are not arbitrary creations of writers, but rather reflections of the fundamental process by which humans make sense of the world. Life is experienced as a confluence of stories that play out simultaneously. Additionally, the stories that form the building blocks of our lives are not constructed (theological considerations aside) by a single entity or storyteller, but emerge spontaneously through the combination of countless inputs from the physical and digital world. Some inputs are generated by us; most come from other agents—people, animals, and natural and man-made systems. These stories, for all intents and purposes, define reality.
In 1950, Alan Turing proposed an experiment which continues to mark the seminal goal of artificial intelligence. The purpose of the imitation game, or Turing test, was to determine whether a computer program exhibited intelligent behavior indistinguishable from humans. In the same vein, a Story Engine that is capable of creating completely original stories that are indistinguishable from those created by humans—much like the Story Train that Rick and Morty were trapped on—would pass the Turing test for stories.
Fair enough, you may think, but what’s the point of this exercise besides putting writers out of a job? For starters, trying to imagine a story that does not originate within a human mind, whose origin can’t be traced back to human imagination is, for me at least, a satisfyingly-recursive thought experiment.
But the potential for such a Story Engine is not just intellectual: it’s coming soon to a screen near you. Gaming technology has seen incredible progress since the early 2000s. A huge global market and the rolling promise of radical new experiences enabled by VR, AR and other tech continues to expand the horizons of interactive entertainment. Still, the potential of these applications—games—are often limited by the context of their story. Too little narrative and the game feels like a simulator; too much robs the gamer of their agency, turning the game into little more than an interactive film.
The concept of a game—or even a platform for game development—powered by a True Story Engine has preoccupied me over the past few years. Even before I saw the excellent “Never Ricking Morty” episode, seeds of this concept had grown into a tangled garden of ideas. The “Power to the Gamers” series published by The Lydion Magazine offered the perfect opportunity to cultivate this garden more scientifically, pruning the tendrils into a multi-part thesis.
In this first installment, I’ll take a closer look at existing methods for the creation of new stories. These methods support human storytellers by transforming a set of inputs into original narratives that comply with pre-supplied parameters.
By defining the different categories of “Story Generators” based on the entity or mechanism that generates the story, it is possible to home in on an approach for the creation of a True Story Engine. So far, I have identified three classes of Story Generators: Manual, Automated, and Natural.
- Manual: This is the most obvious method of artificial story creation—from the mind of one or more people. Examples cover a broad range:
- Traditional writers—from Shakespeare to Stan Lee—established a…
- Formula for sequel or continuity-based stories—explored and taken to extremes by…
- Television, which has also necessitated manual story generation algorithms—from Rick and Morty to Looney Tunes to procedurals (like CSI) to soap operas.
The rapid churn of plots often required a specific formula (as in a procedural) or a more generic algorithm like the aforementioned Circle Theory of Story.
My key takeaway from Manual Story Generators for a True Story Engine creation thesis is that once there is a proven algorithm for the creation of narratives that engage their audience, human storytellers can produce quickly and effectively at high volume for a long time.
- Automated: This method primarily manifests in video games. There are many different sub-categories, including (but not limited to):
- Sandboxes and Simulators: Games such as Minecraft and Flight Simulator provide a sandbox for the simulation of a bunch of activities, then leave it up to the player to create their own game-based stories that typically unfold through interactions with other players. Games like R*'s Grand Theft Auto, the Red Dead Redemption series, and IOI's Hitman series can be considered more structured sandboxes. They allow players to choose (and often alternate) between creating their own stories within the world of the game, or pursuing existing stories created by the game's designers.
- Emergent Gameplay and Roguelikes: Games such as XCOM, Valheim, Animal Crossing, Rimworld, and Among Us create simulators for players to create events and outcomes, through which a story emerges for the player—often unique and different from stories that emerge for other players of the same game.
Similarly, roguelikes such as FTL create a unique story for each run of the game through a combination of events generated by the user's specific choices within the simulation during that run.
- CRPGs (Computer Role Player Games): Baldur's Gate, Elder Scrolls, Diablo, and the recent masterpieces Disco Elysium and Elden Ring generate specific stories based on choices that the player makes about their character’s mental and/or physical attributes. They are not as flexible in their storytelling as Sandbox / Simulator or even Emergent games, but not nearly as rigid in the available story branches as Adventure/Linear games described next.
- Choose-your-Adventure: Games such as Myst led players down a pre-built "choose-your-adventure" determined by the kinds of player choices and puzzles that were foundational for the games industry. Recent titles such as Supermassive's Until Dawn and The Quarry have taken this concept to their technological extreme, letting the player generate interactive movies on the fly, based on specific choices.
- Linear: These titles include traditional sidescrollers or first-person shooters like BioShock or Call of Duty, which guide the player through pre-defined, linear narratives that act as context for the specific set of skills that the game requires (combat, platforming, etc.)
My key takeaway from the Automated Story Generators as seen in video games is that algorithms enable story generation with varying degrees of flexibility, and each sub-category is capable of creating compelling, engaging stories. The Sandbox/Simulation genre (a) has the most flexibility in terms of the specific stories generated for each player and the uniqueness of events that occur. This flexibility decreases across the spectrum with the Linear (e) stories offering the least narrative flexibility. Different players of a game like BioShock typically end up experiencing very similar stories. On the other hand, narrative fidelity and resonance increases from (a) to (e), with linear games like BioShock arguably delivering more impactful stories compared to Minecraft or Animal Crossing. The potential for each of these genres to achieve success along a continuum of narrative impact and flexibility offers us much to explore in the conception and construction of a True Story Engine.
- Natural: This is the method of story generation we experience in everyday life. The "stories" that comprise our lives are formed by myriad real-world outcome events, created by innumerable inputs that are themselves the outcomes of prior events. This recursive tree of cause and effect, compounded by millions of sensory and internal inputs—and subject to our own biased choices—supports the generation of the sequence of events we call “story”.
Herein lies the unexplored potential of digital games, and, more broadly, interactive experiences in general. Games, unlike books, television, or films, by their very nature require input from the player.
Typically, games use player inputs to select from among pre-defined outcomes—each input directs the game further along a path—explored in the "Automated" section above.
However, if we assume that our True Story Engine exists, we can also imagine feeding into this story generation Engine as inputs:
- Millions of in-game actions and events…
- Millions of choice-based inputs generated…
- By thousands of players, making independent decisions…
- And hundreds of trained storytellers, economists, and gamemasters generating "world events", independent of each other.
Conceivably, if the technology and infrastructure existed to enable the above, our True Story Engine would be capable of creating stories that don't come from a single mind or even a collection of minds, but are spontaneously generated by an almost-infinite combination of inputs that would be otherwise almost impossible to predict deterministically. Such a story might have a shot at passing our "Story Turing Test."
My key takeaway for our Story Engine thesis from the Natural Story Generators is that given the general prevalence of the Internet and networking technology, combined with advances in technology such as AR, VR and distributed ledgers/consensus (typically enabled by blockchain databases), it could be possible to simulate the scale, variety, and causality of inputs that spontaneously create events—and stories based on these events—that appear to emerge as naturally as the headlines of today's news.
The takeaways from an initial evaluation of existing Story Generators can serve as guidelines for developing an hypothesis for the creation of a True Story Engine. In the next edition’s installment of this essay, we will try to do just that. We will also attempt to construct a method to test, measure, and evaluate the effectiveness of this hypothesis.
I’m treating this essay series very much as an experiment and a dialog with our readers. As such, I am eager to hear your thoughts on the concept of a Turing test for stories, a True Story Engine, and early explorations on how such an Engine can be built. Please reach out with your ideas, comments, and feedback—these will be invaluable as I develop Part 2.