Narrative Lego: A GDC talk by Ken Levine

I've been actively researching some longstanding issues around narrative systems in game design. While exploring, I stumbled across this wonderful talk from GDC 2014 by Ken Levine on how to build Narrative Lego.

My mission and current (unannounced) project are pushing me to dig deep. I've been actively researching some longstanding issues around narrative systems in game design. While exploring, I stumbled across this wonderful talk from GDC 2014 by Ken Levine on how to build Narrative Lego. I figured my notes would be useful for anyone else who prefers reading to watching videos. Enjoy!

Narrative Lego by Ken Levine at GDC 2014 (posted 2016)

The pitch

By breaking narrative down to its smallest yet non-abstract elements and finding ways to combine and recombine them, one could potentially build a nearly infinite array of narrative opportunities out of these small building blocks. In this GDC 2014 talk, BioShock co-creator Ken Levine explains how it's time to start exploring 'narrative legos'.

Ken's background and context setting

Ken worked on linear narrative games for 19 years, with some very notable accomplishments such as Thief, System Shock 2, Freedom Force, and BioShock. But there are problems with linear narrative and he felt it was time to change.

Why move away from linear narrative games?

  • He didn’t want to keep building “bigger”. This type of content is time-consuming and expensive. Years built up for “One Big Moment” means no constant engagement with the audience.
  • Linear narrative puts boundaries between developer and audience. “Would you kindly only works once” (note: "Would you kindly" is a reference to an important narrative point in Bioshock). These days Ken believes the audience wants to be a part of the story.
  • Ken loves systemic games (eg. Civilization, X-COM). Narrative doesn’t lend itself to system.
  • To do something different, he felt we had to go back to the drawing board. When it comes to innovation, having time to fail for a long period is important. For that reason experimenting on a small team makes more sense. You can’t have a huge team staring at you asking what they should be working on.

What this talk is NOT:

  • A design for a specific game
  • A product pitch
  • A specific development plan
  • An intellectual property

What this talk IS:

  • Open Source
  • Ken’s way of contributing to a conversation people are already thinking about.
  • Ken’s initial scratching in the dirt to figure out a way to build player-driven replayable narrative gameplay.

Negative aspects of linear narrative design

Linear narrative approaches to game design have downsides:

  • Expensive to make
  • Pieces don’t speak to each other. Gameplay doesn’t have real meaning on narrative outcome.
  • Branching exists, but with limited states and interaction. Multiple endings still mean a fixed number of player states.
  • Doesn’t fully embrace the unique power of games. Replayable and player-driven demonstrate the unique power of games.
  • Not player-driven. You are seeing at most the X or Y versions of the story your friend is seeing.
  • Can only add ON, not add IN.

Ken's shout out: We don't want to say it's just bad. There is great and important work going on in this area (The Witcher, etc.). But Ken wants to take a fundamentally different approach.

My current metaphor for adding ON and not IN is a luxury train. Linear narrative designs eventually reach an end to their content. The train reaches the end of the line. Add that point you can only add more track ON to the end of the line. In that way you'll get more time on the train.

To contrast, systemic game designs let you add IN to the train itself. You can go in circles on the same track but add crazy upgrades to the luxury boxcars. Maybe you can make tea now? Maybe now you can have magic powers or speak new languages? The track stays the same but the travel experience becomes fundamentally different.

Approaches to solving the problem with more tech or AI are not working

The technical approach to solving this so far has been to get better at simulating people. In Ken's opinion, everyone is bad at that. They should be simulating characters not people (more on that later). A robust solution lies beyond any current technological or creative horizon.

In addition, overly ambitious solutions (let's simulate human consciousness!) can lead to paralysis. So what are some baby steps we can take?

For example, game physics wasn't built in a day. We had a progression of simple primitives that was built up over time: 2D circles -> rectangles -> 3D spheres -> cylinders -> axis-aligned boxes -> arbitrary polygonal shapes -> ragdolls -> cloth -> fluid

The important point is that people were still excited at the beginning. The 2D circles were pretty cool in their day! So when it comes to solving narrative systems we should stop trying to model everything. Instead we should model a limited set of believable and impactful things.

The opportunity is to create a narratively driven experience where:

  1. Narrative elements are non-linear and interact with each other.
  2. All narrative elements trigger off Player action.
  3. Such triggers are generally transparent to the player.

Starting point: Skyrim

To ground the discussion, Ken reaches for a familiar game: Skyrim. Skyrim has these familiar things:

  • Factions the player can choose between
  • Quests
  • Character Growth
  • Crafting (weapons, armor, structures, vehicles)
  • Non-linear quest structure
  • Strong loop between non-combat and combat

Introducing Stars and Passions

A player standing at the center of four racial factions: Orcs, Elves, Goblins, Dwarves.
A concept for a fantasy world.
Zoom in on orc village: There are five Star NPCs each with a name.
Now we add something new: The Stars.

Only having a few Stars is important because we don’t want to overwhelm the player with characters to track. In this example, there are 5 but it just has to be a number that is trackable and not overwhelming.

So what is a Star?

  • A Star is an NPC with a set of Passions.
  • Passions are not full psychological models. Remember that the tech approach so far has been to simulate a person, not a character. This is a dead end.

Traditional media (eg. movie) narrative models passions, not people. An example is Luke Skywalker. What are Luke Skywalker's Passions?

  • Craves adventure
  • Has father issues
  • Wants to prove himself

By contrast, there are things that just don’t really matter to us about a character even if they exist in the person. What are some possible non-passions for Luke?

  • He's a vegetarian
  • He has tooth decay
  • He has OCD

So a Passion is what a Star cares about relative to the actions of player. It's crucial that the Passions be transparent to the player. Passions must respond to player action.

The Passions of Frank the Orc

A meter showing Frank's impressions of a player based on how much they hurt elves.
One of Frank the Orc's Passions. What a meanie.

Meet Frank the Orc.

He has the default passion shared by most orc Stars: Hating elves. As you do things that are bad for elves, he will like you more. As you do things that are good for elves, he will like you less. So that red/green meter is like a measurement of Frank's feelings for you the player.

The macro-passion meter

What if Frank the Orc has 3 Passions instead of 1?

Stars can have 3 passions, each of which players can impact.
Stars can have 3 passions, each of which players can impact.
The Macro Passion is a function of the 3 Passions of a single Star.
The Macro Passion is a function of the 3 Passions of a single Star.
Macro Passions also have thresholds. When the player crosses one, something consequential happens for the player.
Macro Passions also have thresholds. When the player crosses one, something consequential happens for the player.
Example of a player getting to the second positive threshold.
Example of a player getting to the second positive threshold.
Example of the player losing reputation with Frank.

The important point on that last slide is that you don't quickly change relationships with Frank. You have to build it up over time, and you can break it down over time. This adds a lot of opportunity for maintaining delicate social balances, changing your mind, reacting to new goals, etc.

Game equilibriums we need to know

A non-zero sum game involves a mixture of resources. You have all the pink moons and I have all the blue circles. We can trade and now we each have a bit of both. Everybody is happy.

A zero-sum game involves a single pool of one resource. If there is a single pool of 10 gold, when I get 3 gold, that’s just 3 less for you. I gain, you lose. You can't please everybody.

Zero-sum games raise the stakes in this setup.

Zero-sum game with Orcs vs Elves.
Zero-sum game with Orcs vs Orcs.
More drama! The love that dare not speak its name! 

Within villages, there is no monoculture. Each character has a specific hope and dream. Not all Orcs share the same passions.

Introducing Drones

We need to populate the world without overwhelming players. There can only be a limited number of Star characters or the players will lose track. The solution is to create Drones.

Drones are simplified NPCs without passions. So their overall feeling towards you is determined as a function of the Macro Passions of the Star characters in the same village or faction.

Ideas for things you could do with the higher level macro line

Dramatic Events

Let's inject more drama with an "Unaligned Star".

The Red Dragon appears in the skies above Fantasy Land!

The Red Dragon is an Unaligned Star. All he cares about is death and destruction on his Macro Passion bar. But there is another dragon. A Silver Dragon. He just wants you to do nice constructive things.

There is a zero-sum game between the Red and Silver dragons.

We can scale everything we have so far out to more factions, each with their own Stars and Drones. Everything interacts with everything else.

Overview of the entire system so far. Things are getting wild.

Scaling out further: From the north come the White Walkers!

This type of story event is spawned by a confluence of Macro Passions. This event has the effect of adding a new Passion to every Star which reflects their feelings towards the new zombie threat.

Major Events like this change the ground underneath you.

Adding more factions and a major story event for greater scale.

Wait... wasn't this a talk about narrative?

Yes. But replayable narrative first needs a system (recall that replayability was a key goal set by Ken at the beginning).

We've built a web of nearly infinite relationship states. Now we can allow changes in those states to "fire off" professionally created and crafted content.

Player triggered passion thresholds launch fancy content moments.

Detailed example of a story using this system

In the following slides Ken lays out a detailed example. In the example, the player is considering whether they should try to marry Veronica the Elf or Betty the Dwarf. The player can't marry both and has to make a choice (remember zero-sum games are more dramatic!).

At a high level, the player is understanding that Veronica is mechanically about combat and Betty is about crafting.
The player ultimately completes their quest and marries Veronica.

The player ultimately completes their quest and marries Veronica. They receive massive buffs and rewards and castles and stuff. Betty hits a massive low and the Player can no longer craft crops in the Dwarf Village.

A real problem for you.

And there is another problem. This whole thing caused a tension between mechanical benefits vs the player's personality preferences. You now have to live with this tension between love of personality and love of systemic game reward.

With good writing, players should have to make this balance of game vs emotional benefits. This leads to false friends, leading others on, betrayal, and reconciliation.

Things start to get pretty Game of Thrones 🙂 – but it’s all driven by you.

Hidden Passions

What if some passions were not transparent from the start? These hidden passions would work like a resource that the player can work to reveal by getting to know someone. You could potentially even get access to this resource from other players somehow.

Revealing a new hidden passion for a character has a ripple effect. What if orc reinforcements arrive and one of them is Juliet! This player would have a dramatic complication because they know Romeo so well, while another player would not have that layer of complication.

Closing thoughts on replayability and multiplayer

Generally the games Ken has made have been ADD ON only, not ADD IN (remember my train metaphor?). With this system, there are many ways to encourage replay.

Player's would have plenty of choice. They could support different factions or Stars. They could support the same factions or Stars but to different extents.

We could make a passion pool for each Star. We could write, record, and produce 10 passions, but on each playthrough, you only draw 3 at random. NPCs would be the same people, but their focus would change.

We could add new Stars, Passions, and Quests over time. With this design, a new faction is an ADD IN. Adding werewolves with Stars and Passions would slot into the existing system and world but create more possible collisions.

What would co-op look like?

Early thoughts here, but imagine Ken played with his brother Stu. Romeo sends the elves to attack Ken. Ken asks his brother Stu to help him. But Stu values his relationship with the elves in the game.....

The impact of this betrayal would be HUGE. Ken still remembers a Diplomacy incident he once had with a real-life friend. The friendship remains but the betrayal stings to this day.

Final thoughts from Adam

A few things stood out to me studying this talk, and I have some clear next steps.

  • This would be worth modeling in code to see how complex it gets.
  • I'm not convinced that Passions would cleanly aggregate up into one Macro-Passion bar, but if it worked it would simplify things a ton.
  • This talk is old now by game design standards so I want to learn about games that have rolled out systems similar to this. I bet they are out there.
  • Centering a narrative system on characters makes a ton of sense.
  • I completely agree with the premise: stop trying to model human consciousness and start simply. Let's steal ideas from screenwriting and make characters a set of basic passions. This idea is described in Apocalypse World in a delightfully vulgar way.

Thanks Ken, I learned a lot!

Become a better Game Master, Worldbuilder, and Content Creator

1 Homebrew. 1 Highlight. 1 Hack. Once a month.
Adam shares original TTRPG content, cool finds, and lessons learned with 1000+ members in the Sword & Source newsletter.
Unsubscribe anytime. Your email will never be shared with others. Read the privacy policy.