Immersive Theatre and Westworld, Part Two: Audience Agency
This article is the second in a series of four. I recommend starting with Part One.
“The newcomers are looking for the same thing we are: a place to be free, a place to stake out our dreams. A place of unlimited possibilities.”
– Season 1 Episode 1, ‘The Original’
In Part One we looked at some different approaches to narrative structure: On Rails, Branching Path, and Open World. I framed these only in terms of physically moving through a space to encounter different scenes, with the audience’s ability limited to what scene they see next. However, these narrative principles can also apply to the way each scene plays out in an interactive show, where the audience is given some degree of agency as to what happens next in the story.
Westworld handles this in a very interesting and unique way, but to get to that let’s first see how audience agency might be handled within the structures discussed in Part One. For the sake of simplicity, we’ll continue with the example from Part One of Dolores dropping a can of beans.
First, Dolores drops her can of beans in an on-rails show with peripheral interactivity. The story requires an audience member to pick it up and give it to her, so she stalls for time until someone does, maybe saying something like “now where did my can of beans go?” The show is likely in a holding pattern, with the audience and characters unable to move on until the can of beans is picked up, as this is the only narrative path for the story to follow. It’s only lightly interactive, as the audience is participating in the story, but they don’t really have any agency in how the story unfolds.
In a branching path show, the can of beans might land on the ground and roll until they’re halfway between Dolores and a hungry child. Now you’re faced with a choice between two options: who do you give the beans to? Here the audience member has agency to, through their actions, choose one of two predetermined paths for the story. However, any other action that isn’t one of these two paths (keeping the beans for themselves, say) will be disregarded or blocked, as it doesn’t fit one of the two paths. Again, the scene will likely stay in a holding pattern until the beans are given to Dolores or the child.
In an open-world videogame, the can of beans falls on the floor, but then the options of what you can do with it are limited by the different controls available to you — the classic set would be something like look at it, take it, talk to it, shoot it, use it, walk away from it, or give it to someone, each of which might lead to a different story response (some of which might be that nothing happens). However many options there are, though, there are a finite number, because the game designers can only script and program so many responses to the player input.¹
And this is where we can see how Westworld is different. The audience member doesn’t have a set number of options to choose from. They can do anything they like with the tin of beans. They can eat it, roll it, steal it, punch it, kiss it, sing about it, sell it — whatever they want, so long as they’re physically able to do it. To paraphrase Laura E. Hall, their body is the game mechanic. Their response to the offer could take any form, which opens up a whole sphere of possibilities.² Within the TV show Westworld, this is one of the park’s selling points, what gives it its immersive quality — the guests have complete freedom to do whatever they want, without artificial narrative constraints.
But hang on, how do you possibly write for that? How can you incorporate that level of audience agency while still providing narrative cohesion, without the whole experience being completely improvised? Let’s look, in turn, at how all each of these different approaches might be applied to writing The Beans Scene into an experiential theatre show.
Let’s start by making an on-rails show. You write the dialogue that leads up to Dolores dropping the beans, you write the dialogue that takes place after she gets the beans, and maybe a couple of different filler lines for Evan Rachel Wood to drop in, depending on how long it takes for someone to pick up the beans. Great.
For a binary choice branching paths show, you write the dialogue leading up to the moment of the choice, then write two different endings, one for each of the options that the audience can choose, and the scene plays out based on their choice. If there are multiple choices throughout the show that affect each other, you’ll need to script even more scenes — three interdependent binary choices create eight (2³) possible endings.
Now let’s imagine that our show limits our audience members to the classic set of videogame options described above. If the audience has a dozen different things they can do to the can of beans, then we need to script a dozen different scenes in response. This quickly becomes impractical if we apply it to the many choices that they might be able to make throughout the show — three interdependent choices of twelve options creates 1728 (12³) possible endings. One way to deal with this would be to group possible audience responses based on the impact they’ll have on the world. For instance, giving the beans or talking to Dolores might start a romance scene, taking the beans or walking away might cause her to just ignore you, and shooting or talking to the beans might mean she calls you weird.
What we’re doing here is reducing a dozen options down to a more manageable number of outcomes, which we can more easily script. The choice of different story paths is dependent on whether certain circumstantial conditions are met by audience actions, rather than the audience member consciously needing to make an active choice.
But none of these different formats is quite what Westworld does. In Westworld, the hosts (characters) are improvising the scene with the audience. The audience can go anywhere within the circle of possibilities, and the hosts can go with them. So how do we script this?
The solution that the (in-fiction) narrative designers of the Westworld theme park have come up with is very useful. For each narrative moment, they start by scripting the audience responses that are most likely. If the can of beans falls on the ground, say, the most likely outcomes are that a passer-by will either pick it up and give it to Dolores, or they’re going to ignore it completely. So, the narrative designers start by writing scripts for scenes that respond to both those options. Then, if the audience does something outside of those expectations, the hosts start improvising. If the passer-by sings a song about beans, or starts cartwheeling, or shoots Dolores’ horse, she’ll respond by improvising — what would her character do in this situation?
But these improvisations also incorporate and repurpose sections of their scripted dialogue. We see Maeve’s ‘This is the new world’ monologue a number of times throughout the series, taking on different meanings in different contexts. Or, when Dolores is being interviewed by Jeffrey Wright’s character, she talks about loss, saying:
“You think the grief will make you smaller inside, like your heart will collapse in on itself, but it doesn’t. I feel spaces opening up inside of me like a building with rooms I’ve never explored.”
– Episode 2, ‘Chestnut’
To which Jeffrey Wright replies, “That’s very pretty, Dolores. Did we write that for you?” She answers “In part. I adapted it from a scripted dialogue about love.”
So we can see that, from a line-to-line writing point of view, we can start by scripting sections of dialogue with specific scenes and giving our actors permission to adapt and reuse that dialogue as required. But this principle can also be extrapolated to look at the overall narrative structure of the park.
The host characters’ days are structured as a 24-hour loop. Left to their own devices, with no guests in the park, the hosts’ lives would play out the same every day. Dolores wakes up in the morning, talks to her father on the porch, goes into town to get the groceries, meets Teddy (who picks up the can of beans and gives it to her), spends time in the wilderness with him, then goes home to find that her father has been seized by bandits, who shoot Teddy and then her. Charming. All completely scripted..
But now let’s add the audience in and give them narrative agency. They can make choices about what happens, which will affect the loops. Say that a guest starts talking to Teddy before he can reach Dolores. This means that, instead of meeting Teddy, she moves onto a second variation of the loop, one where she goes back home, then takes her watercolours out to do some painting in a field. Dolores’ story is still scripted, but it’s following a different branching path. If she meets Teddy, move to loop 1A, if she doesn’t, move to loop 1B, etc.
As an aside, we can see here the same principle of ‘circumstantial conditions’ I described earlier. Audience members may only be making a few active choices, but these choices have knock-on effects that passively impact other story loops. The conditions become quite complex as well, with multiple endings of multiple stories all becoming interrelated.
Anyway, for each choice the audience can make, the narrative designers plan for the choices they think audience members are most likely to make, while simultaneously making room for improvisation. It seems that in Westworld, the narrative designers intend, or at least prefer, for audience members to choose one of the scripted options, as this will keep the story closer to the established loops, providing more guarantee of a satisfying narrative.
One way to do this is to provide audience members with instruction from the hosts. For instance, in Episode 1 Dolores tells a family of guests she meets that she has to go, and that they should too, because “There’s bandits in these hills.” Another approach is to lead the audience to take certain actions without them realising they’re being led. You enter a room, and at the other end of the room a desk lamp shines on an open book. (See the discussion of passive invitations in Part One) You’re not being told to go over and read it, you’re just subliminally encouraged to. Jeff Wirth calls this process backleading³ — invisibly guiding an audience member to make certain decisions (For more on this, see the discussion of the 1–9–90 principle in Part Three). There are also many overlaps here with other disciplines including exhibition design, user experience, choice architecture, and behavioural psychology.
In theatre, the understanding of the most likely (or perhaps, predictable) audience choices and responses is deepened and refined over the run of a show. The more people you get making each choice in performance after performance, the better you understand what the most likely outcomes are. In my experience, this is most keenly observed by the actors, as they’re the ones seeing each decision being made. It’s crucial that this knowledge is shared among the cast and crew of the show, and that the narrative architecture stays a living beast, so that adjustments can be made as we learn more about how people are actually navigate the story we put in front of them.
At any rate, in Westworld, if the audience member happens to make a choice that doesn’t fall into one of the scripted paths, this choice is still accepted and followed by the hosts. It’s frequently mentioned in the show that the hosts are only supposed to make minor improvisations within their loops, presumably to smoothly facilitate a natural conversation and relationship with a guest. If an audience member wants to take them out of their loop, then they will improvise to a more considerable degree, but their improvisation is being led by the audience. For instance, The Man in Black (Ed Harris) kidnaps several characters and takes them well outside their loops, so these hosts improvise significantly, but only in response to the audience member’s choices, filtered through their character’s backstory. They don’t suddenly have free rein to improvise anything.
In an experiential show, the cast may be inspired by a strong offer from an audience member, and similarly improvise away from any planned narrative paths, However their capacity to do this entirely depends on how much flexibility and plasticity is built into the show. Some shows have fixed endings that need to be reached, while others are wide open for new endings to be found. Both are completely valid, however as Jason Warren writes in Creating Worlds, the art in designing for interaction is most often a balance between the amount of power the audience are given in deciding what happens in a story, and the amount of work required by the cast and crew to support those decisions and make the story happen. (For more on how to find this balance, I strongly recommend getting a copy of his book.)
All of this comes back to one of the promises of the Westworld theme park: the guests have complete freedom to make any choice they want in an immersive world. The hosts are fully realised characters, who bring the consequences of those choices to life. In Part Three, we’ll look at the tools that Westworld uses to design its host characters in ways that support these audience choices.
This series of articles (and all the theatre I make and things I write) are supported by my beautiful patrons. Won’t you join them?
Diagrams in this series are by Tom Noble.
¹ This is sort of an oversimplification, as there are lots of interesting games coming out increasingly exploring emergent gameplay and AI-driven responses.
² The Circle of Possibilities’ is a variation on Keith Johnstone’s concept of the Circle of Expectations in improvisation. For more on his approach to narrative I recommend his fantastic ‘Impro for Storytellers’ (1999).
³ The name comes from ballroom dancing, where traditionally the ‘man’ leads and the ‘woman’ follows. However, if the follower is a more experienced dancer than the leader, then the follower can backlead, using their knowledge of the moves to create space for the ‘leader’ to step into.