/Blog: Using emergent systems to improve interactive storytelling

Blog: Using emergent systems to improve interactive storytelling


The following blog post, unless otherwise noted, was written by a member of Gamasutra’s community.
The thoughts and opinions expressed are those of the writer and not Gamasutra or its parent company.


 

I was playing God of War (2018) recently and the first thing you do as Kratos, is transport a big log with your son. As you’re rowing a boat, you approach the dock. As you dock, the game smoothly transitions from gameplay to cutscene. Kratos steps out, the camera pans and focuses a bit on his son, then pans back to Kratos pulling the log out of the water. As the boy passes Kratos, the camera smoothly pans back to be behind Kratos, to that familiar 3rd person perspective we all know and love. A beautiful display of game cinematics that took decades of evolution to find a sweet spot between “only gameplay” and “cinematic cutscenes without gameplay”. But the player can still put down the controller, because it removes the control from the player. The story perspective shifts from 1st person, to 3rd person. You stop being Kratos and are now watching Kratos. The game falls back on an old medium, cinema, to deliver a story beat. We don’t often see movies falling back on the book medium whenever they want to convey something that is hard to convey in a movie. However, in games it has become the norm. Isn’t that odd?

In the movie industry they found out a while ago that they don’t need to convey thoughts literally, like books do, by text. They could show emotions to imply them, have them literally be read out loud as the character is thinking and various other similar techniques. These constant improvements required quite a bit of effort from people dedicated to innovation. People, who had to think “This is how we are doing it now, but could it be done better in a different manner?”.

Don’t get met wrong. I like God of War, it deserves the praises it gets and like many modern games, it blends cinematography and gameplay in a satisfying manner. We as an industry have progressed on this area greatly, evidently so. But I think we could use more effort on other techniques that might prove even more potential in the long term.

That is where emergent narrative systems come in. Emergent narrative might prove to be an effective solution, because emergent narratives arise from interaction, not from a writer’s vision of how you are supposed to experience the story.

If you haven’t heard of emergent systems, it’s when properties and/or behaviors arise out of components that don’t exhibit or imply these properties and/or behaviors on their own, but do when they interact within their wider whole. To give an example: Birds want to stay near each other, they want to go in the same direction, but they also don’t want to collide with each other. So if you see a single bird those rules don’t look like much, but as soon as you see a huge flock you’ll see the most interesting patterns and behaviors emerge.

So an emergent narrative is a story that is unscripted and emerges out of its components. In real life this is actually the most common narrative, it’s literally the conversations we’re having, the way the news talks about events, the way we react to the news and talk to each other about it, the way those reactions could lead to new newsworthy actions. Most of it is pretty mundane and not newsworthy though, while written stories can be pretty dramatic. So why should we want emergent narratives in games? Because of the interactive nature of games. Pre-written narratives can be great, but an emergent narrative fits the medium’s interactive nature better. I’m not saying all games in the future should feature emergent narratives, but right now we are working with narrative forms that don’t fully exploit the interactive nature of videogames. I say videogames specifically, because while there are some improvements with videogames this decade, everything still pales in comparison to the emergent narrative of Dungeons & Dragons.

Emergent narrative in Dungeons and Dragons

In Dungeons and Dragons every session features different story beats, even when two different groups pick the same adventure. But how? A simple answer is: “Because the players decide what happens”, but a clearer explanation to me would be: “Because players are people with unique personalities that respond to each of the other player’s actions in a unique way”. And even then we’re oversimplifying the immense complexity of the emergence occurring here.

The game industry has been trying to simulate the Dungeons & Dragons aesthetic since forever. Prevalent is the simulation of its RPG systems, NOT its narrative system. For the sake of simplicity I came to the conclusion that Dungeons & Dragons has emergent narrative, because people do what they want. What people want and do is largely based on their personalities (and other things like their backgrounds, environment, intelligence, physical- and mental health etc, but that would make this very complex). The reason I wanted to oversimplify is because I had an idea brewing that needed to distill this phenomena to its core.

I recently prototyped an emergent system that simulates personalities called Fabula Persona.

There are a lot of complex processes happening here, which are not obvious at first glance. The NPCs have a lot of unscripted behavior. I’ll explain the design of Fabula Persona and show interesting examples of emergent narrative.

Fabula Persona is Latin for story through personality, which is what I tried to achieve here, or better yet actually achieved here, depending on your definition of story. Narrative is defined slightly differently everywhere. I prefer the Dictionary.com definition: “a story or account of events, experiences, or the like, whether true or fictitious“. In my own words, any event you link together to form a cohesive whole.

Before I’m able to explain how the emergent narrative in Fabula Persona works, I need to propose a way of looking at narratives in games. Narratives, especially emergent narratives, differ greatly in user experience depending on how it is delivered. Having names for phenomena is the first step towards becoming more literate, something we as an industry could very much use. It’s baffling how much words are needed to explain abstract concepts or how most game design docs contain very convoluted ideas.

We could think of narrative as being straight forward content like dialog and story text or more contextual like environment or facial expressions (implicit vs explicit).

We could also look at narrative as being either something you experience within yourself or some external event (intrinsic vs extrinsic).

Any event could be placed within this framework having element one of both categories. Note: this is compatible with both interactive and linear storytelling.

                            

An extrinsic/explicit event is very simple to grasp, it’s whenever something explicit happens that we perceive externally. It’s when characters say something or whenever any event happens. Like the dialogue in the Last of Us, beating Bowser or causing mayhem in GTA.

Extrinsic/implicit events are more contextual and open ended, but still external stimuli. For example, many art pieces can be interpreted in multiple ways, because the meaning is “implied”. The aligned cannons in Breath of the Wild’s Akalla Citadel Ruins heavily imply that a battle took place, it tells a story implicitly, in this case, through the environment.

Intrinsic/explicit events are how the player experiences the story. “As the ball bounced back, I noticed it flying at an almost impossible angle and speed towards me. With my sharp reflexes I barely managed to block the ball, by what seems only 1 pixel, causing me to make the winning point!” That’s the narrative of a Pong game. It’s how the story actually happened, is experienced and how it is retold later. Dwarf Fortress causes these experiences often, its abstract art has the player’s imaginations run amok. There is a whole website dedicated to these stories https://dfstories.com/ (Note that these stories are now made explicit/extrinsic by sharing it with others, but they came from an internal experience).

For the intrinsic/implicit narrative I’ll use the same example. The explicit narrative “As the ball bounced back, I noticed…etc.”, is the content, it’s literally what happened, but the intrinsic/implicit narrative is contextual. It’s what the player felt, it’s what their subconscious thoughts were, it’s not literal, but it is definitely part of the experience. As you read the explicit narrative of the pong game or if you experienced something similar, you probably also experienced the implied narrative. The feeling of excitement, redemption, victory and the quick non-verbal millisecond thoughts that arise. I know this form of narrative seems vague, but it is part of the experience and it is also certainly an account of events manifesting.

The narrative forms are not mutually exclusive within a storytelling experience, the experience can evoke all of these, but you could use this way of thinking to define which phenomena you experience is what.

It is virtually impossible to get the same intrinsic narrative from different people. With Dungeons & Dragons, there is this input/output system going on where the intrinsic narrative of one player gets shared with the others, and thus becomes extrinsic, which creates new intrinsic narratives, which in their turn all supply the overarching extrinsic narrative.

Fabula Persona’s narrative type

In the case of Fabula Persona, I would define it as mostly having an extrinsic/implicit narrative. The NPCs definitely provide the content of what happens, but what it actually means isn’t objective.

I was aiming for a more explicit narrative initially, but this implicit style of narrative that I’ve achieved was already very ambitious. Innovating storytelling in games is hard.

Now that I’ve summarized the underlying thoughts, explaining how Fabula Persona works becomes more palatable. Knowing how I designed Fabula Persona could help you come up with another emergent system that might prove useful for the future of storytelling in games.

I was heavily inspired by playing Dungeons and Dragons with people that have diverse personalities, to simulate personalities for NPCs. The idea was to have stories emerge through the dynamic interactions that the NPCs with unique personalities have. To have their personalities interact, I decided that they should be able to talk. I also gave them a simple navmesh system so they could walk around without bumping into everything.

Mechanics/systems

To ensure emergent narratives manifesting, the criteria while designing mechanics was:

  • A mechanic should lead to an interaction that could trigger another mechanic, to stimulate chain reactions. Example: The wandering mechanic is there for NPCs to (accidentally) get into each other’s personal space which leads to an emotional response, which can also have multiple outcomes.
  • To differentiate from emergent gameplay the outcomes should imply some meaning. Example: Emotional responses and conversation topics could imply someone getting angry over a food debate.

Emotions

Early on I realized the NPCs needed emotions to convey what was happening to them. To not make the project more complex than it already was, I chose to only go for the four basic emotions: happy, sad, angry and scared.

Interests

The NPCs needed things to like and dislike to be able to become happy or sad. So the topics they could talk about (money, music, art, games, drinks, food, love and the weather) got assigned randomly to either the like- or dislike list for every simulation. The topics are arbitrary, they don’t have different mechanics and are only there because of their implied meaning.

Personal space and awareness

To become scared or angry NPCs have a personal space. If another NPC decides to approach too closely, they’ll either approach the NPC angrily, run away scared or do nothing, depending on their personality and stress meter.

They also have an awareness circle which allows them to notice if they are near things that make them happy or sad or notice the emotions of other nearby NPCs.

Goals

There is also a system in place for them to pick a random goal, which could be “Talk to NPC 5”, “Walk to Object 2” or “Wander around” etc. This gets overwritten by any emotional response. So “become angry at NPC 3” or “Walk away from nearby fight” would take precedence over their initial goals. When their stress levels have diminished they would restart their goal orientation phase.

Personality system

To have every session be diverse, the NPCs need to have different personalities.

Without going too in-depth into the psychology of personalities, I’ve simply chosen The Big Five personality traits as a blueprint as it is most commonly used by psychologists nowadays.

The below image is a nice TL;DR version of it, although in my design I replace “Emotional stability” by the more commonly used “Neuroticism”, so the low/high description is inverse.

I gave every NPC a randomly assigned score of 1 to 5 on every of these traits, which resets on every new simulation. The traits in Fabula Persona translate into the following:

  • Extraverted NPCs have a higher chance to approach a person instead of an object, talk more, talk longer and have a smaller personal space. Introverts approach objects more than NPCs and are happier with less social interaction.
  • Agreeable NPCs don’t become angry easily, they tend to pick a more submissive stance like becoming scared. They are easily influenced by others in their awareness circle. Being near a heated argument influences their stress levels easier.
  • Conscientious NPCs have a higher chance of picking an interest to approach and have a lower chance of wandering aimlessly.
  • Neurotic NPCs are prone to negative emotions, the stress meter fills up quicker, and cools down slower. So based on their agreeableness they either become scared or angry when this happens.
  • Open NPCs like more things, are more easily bored and seek out new things quicker.

And there you have it, an extrinsic/implicit narrative system where interesting situations can emerge.

Interesting cases

So here are some interesting cases that occurred:

 

I hope Fabula Persona can inspire others to experiment with something similar. I was able to make Fabula Persona thanks to the Creative Industries Fund NL. Without them I wouldn’t have had the chance to try this out. I’m grateful funds exist, because they stimulate innovation like this. I hope that designing emergent narratives in the future isn’t seen as risky anymore.

If you’re looking to do something similar, maybe some of the design techniques I’ve used could be of value. However if there is one other tip I could give, it would be: Solve bugs immediately before adding more features. Due to the interrelated nature of the system, it easily becomes a tangled web of bugs. It quickly becomes difficult to see if the mechanic is working as supposed, and if it isn’t, what is causing the bug. It’s also not as clear at first hand if the newly introduced feature isn’t breaking something else you’re not paying attention to at the moment.

If you have any questions or remarks about Fabula’s Persona’s design, emergent (narrative) systems, or something similar, feel free to leave a comment or to contact me otherwise.

An interesting approach I purposefully overlooked in this article is the use of neural networks. There are many interesting projects to pick from but a good example of a qualitative explicit external emergent narrative system is AI Dungeon. AI Dungeon is an infinitely generated text adventure powered by deep learning. I also find it a fitting example for this article, because it’s very Dungeons and Dragons like, in the sense that the dungeon master is an A.I. Every line with a “>” is user input:Afbeelding

Machine learning still has flaws, but the future of procedurally generated stories is looking promising. It is one of the reasons I focussed on simulating personalities, because in the future we will need systems that understand how people work on a deeper level. Currently, procedurally generated stories tend to have characters that are very inconsistent, personality wise, and would probably be considered mentally unstable in the real world.

If I had more time, I would’ve loved to see if I was able to make the narrative more explicit. Having them interact with the mechanics, anti-social NPCs resorting to picking up weapons when angry, agreeable NPCs taking care of wounded characters, dialogue that is generated by A.I. that takes into account the emotional state and personality and that adjusts its vocabulary accordingly. Or lying to the NPCs who are unreliable narrators, because they can also lie or they simply remember events wrongly.

It could, ideally, transform into a game where the NPCs function as people with real personalities that react to the player’s actions accordingly, like a Dungeons and Dragons game with thousands of players. To make these events less random and chaotic, the dungeon master in Dungeons and Dragons manages the dramatic tension. So an A.I. that registers events and manages the narrative arc would likely increase the quality of the narrative.

Looking at cinema for awesome cinematography has served us well, but there are so many other places we can look towards as inspiration. Emergent narratives happen all around us. I’ve used the personalities of people playing Dungeons and Dragons, but you might look at narrators at a sports event, improv theater, freestyle rap battles, the way fake news spreads or observe how people react to accidents. In the meantime I will return to God of War, listening to Kratos telling stories to his son with the worst delivery possible, hoping that emergent narratives will soon achieve Kratos’ level of hilariously bad storytelling.