Gamasutra: The Art & Business of Making Gamesspacer
View All     RSS
October 21, 2014
arrowPress Releases
October 21, 2014
PR Newswire
View All
View All     Submit Event

If you enjoy reading this site, you might also want to check out these UBM Tech sites:

Framework for Systemic Storytelling, Part 2.
by Justin Keverne on 10/11/11 02:53:00 pm   Featured Blogs

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.


Building off the initial framework outlined in Part 1 these additional concepts serve to provide means of structure and control. The primary appeal of this model is that it marries dramatic character development with player agency while potentially allowing for more variation than can efficiently be achieved through the use of branching narratives alone.

For a possible manner in which the described concepts could be used within an existing game consider the myriad characters in Alpha Protocol with their conflicting goals and motivations. Instead of the increasingly complicated branching structures that were used the relationships between different characters and between each of them and the player could be handled systemically. For the player the observable outcome might well be very similar to that achieved by scripting each possible interaction, but by defining those relationships systemically and by allowing players inputs into that system numerous additional options are opened up and the range of player expression is increased.

General Concepts:

  • Abandon the use of plot as the overriding motivator for progression focus instead on character motivations.

Separating player actions from a scripted plot allows players to take actions based on their desired outcome, or at the very least their least undesirable outcome, rather than the outcome decreed by the original designer. In the Alpha Protocol example a similar structure to the one that was used could be encouraged by simply giving the player the objective of disrupting the plans of Ali Shaheed. Certain characters would be motivated to help, others to hinder based on their long term goals as defined by the designers and writers.

  • Rely on basic assumptions about player psychology.

Players will naturally apply human traits and motivations to characters and they will tend to continue following a path they find interesting. If your characters are strong enough players will want to see their arcs through to the end. (Unsurprisingly this sections requires more in depth analysis and study to ensure that any assumptions made are accurate and appropriate.)

  • Focus on character arcs over plot arcs.

Dramatic moments are subjective what is important to one character is a non-event to another don't try to imbue a scene with emotion if the characters the player is focusing on have nothing at stake.

  • Populate the world with characters that have non-aligned goals and motivations.

Two characters with directly aligned long term goals does not make for dramatically interesting conflict. Allowing the player to take sides, or not, based upon their actions immediately requires one or more of the characters to adjust their plans thereby creating conflict.

  • Allow events to unfold without player involvement.

If two characters are motivated to kill each other and the player or other characters do not act to stop them let them kill each other. The player doesn't need to witness such events but they should, like all other characters, be affected by the consequences.

  • Treat the player as another character.

Don't create special case interactions between the player and other characters.

  • Determine player choices based on the actions they take not through explicit decision points.

Defining players based on their actions allows characters to make judgements based on what they do and therefore react to them as they would any other character. Where the player goes, when, with whom and what they do there should all be used to determine other characters reactions to them.

  • Define player verbs by the characters, props and setting.

Don't allow the player to use a weapon or directly attack other characters if it is inappropriate to the setting. A political thriller calls for a range of characters and props that a fantasy adventure does not, player verbs should be defined accordingly.

  • Use characters, and setting to determine genre and theme.

If the cast of characters, settings and props are those befitting a noir story then the choices available to the player can be organically restricted to those that are thematically appropriate for such a story. Genre conventions in this sense are not necessary a flaw and in fact they can help players understand the range of options available to them.

  • Make it clear that motivations assigned to actions are character specific.

If players want to act a certain way to gain the support of a specific character let them. They are not gaming the system, they are manipulating particular characters.

  • Implement a wide variety of vectors by which to inform players of events and character motivation.

News reports, emails, diaries, gossip, all these methods and more can be used to impart information to the player regarding events in the world and the motivations of particular characters.

  • Track interactions of players with the various means of obtaining information to create a model of player awareness of world events.

By tracking the vectors through which players obtain information, assumptions can be made regarding what events the player may or may be aware of at any given time.

Structure and Dramatic control:
The following are methods of controlling the structure and flow of the player's experience and preventing potential combinatorial explosion. In general these rely on filtering possible character actions based on certain criteria. This criteria can either be defined at creation, or set up to change based on specific events.

  • Use dramatic filters to modify or influence possible character choices.

The use of specific constraints on character behaviour can be used to promote certain aesthetic experiences. A theme of tragic romance can be promoted by limiting character actions to those motivated primarily by emotional considerations over practical ones. (Requires codification of dramatic and thematic concepts, needs further examine in a later post.) This should not be necessary except in very specific circumstances because the characters and setting will have been designed initially to be ones appropriate to the theme.

  • Use player knowledge and player awareness to filter character choices.

Limit the ability of a character to take actions with wide ranging consequences if the player only has limited awareness of that character. This can be overridden by a dramatic filter for example if the player seems likely to meet this character in the future, allowing their influence to be felt before their make their presence known might be more appropriate.

  • As time progresses limit the influence of characters the player has had little or no interaction with.

Focus the story down to those characters the player had shown themselves to be more interested in. This will help to prevent events from occuring unexpectedly.

  • If necessary create a "Fate" character to allow actions to occur beyond the control of all other characters.

If designers desire certain events to occur such events can be instigated by a general purpose "Fate" character, who effectively serves as a designer proxy. Ideally such a character would never be needed but the possibility exists to allow this framework to work in support of a more scripted story.

Stories can be created that revolve around character emotions and desires rather than objects. What follows is an incomplete list of potential story beats possible with the techniques described. All these moments would occurred dynamically based on player actions and characters' reaction to, and interpretation of, those actions. In all these instances player actions could lead to a variation upon or a complete reversal of events. Consider the possibilities offered by such events occurring dynamically in a game like Alpha Protocol or Deus Ex.

  • A character might lie to the player to get them to perform a specific task so they can avoid being implicated.

Because the transmission of information is modeled it might fit a character's motivations for a certain action to be performed but not attributed to them. The variables that govern a desire for a certain outcome and a desire to maintain positive relationships with certain other characters might both be high.

  • A player could act as a puppet master exploiting the desires of multiple characters to bring about a specific occurrence.

With an understanding of the way in which knowledge of events propagates, players could manipulate the flow of information to convince characters to take actions on their behalf. This is in essence a reversal of the previous example.

  • While working for a particular character the player is betrayed because their previous actions, which had been unknown, come to light.

Information about actions occurs through interaction between characters so it is not instantaneously. It would be possible therefore, for a player to take an action detrimental to a charcter's desires and then start working with that character, only for them to then discover the player's earlier actions.

  • Characters with competing motivations and long term goals join forces because the actions of the player have disrupted both their plans.

Certain actions on the part of the player could make them a more immediate problem for two otherwise competing characters, leading them to both take actions to deal with the player, thereby either directly or indirectly helping each other.

Related Jobs

Petroglyph Games
Petroglyph Games — Las Vegas, Nevada, United States

Illustrator / Concept Artist
DoubleDown Interactive
DoubleDown Interactive — Seattle, Washington, United States

Game Designer
Zindagi Games
Zindagi Games — Camarillo, California, United States

MOBILE Art Director
Treyarch / Activision
Treyarch / Activision — Santa Monica, California, United States

Senior UI Artist (temporary) Treyarch


Jonathan Lawn
profile image
I like what you're doing here, but it feels like you've a way to go. You seem to have made some interesting decisions already though.

I can see how this can lead to a story of a crowd of characters, rather like Facade ( was. I'm not sure how you can ensure the player a central role in this. By giving them a role in the scenario that keeps them central? Or by giving them powers that other characters don't have? Or perhaps you don't feel a need to make them central: taking any role in the overall story could be equally interesting (in which case maybe this can be an MMO not AI run)?

Similarly, I can't see anything on dealing with an large world of characters. Can you track every character? Can you tell which will be important when you create them. In M&B Warband, every character seems to have an attitude to you, but they limit the cast to a few hundred.

It sounds fun to be able to play with the sort of interactions you describe in your goals, but if the characters are going to be able to compete with the player in this, the AI is going to have to be good (or at least convincing). Projects to produce AI for the game Diplomacy (e.g., have shown how difficult this is.

I'm sure you're looking at all this and more (especially the "Structure and Dramatic Control" section), and look forward to hearing more about your progress.

Tiago Costa
profile image
Well just like in real life, our actions make people follow you or leave you.

In this case no one would follow the player, when the player made a choice that some NPC would agree with they would be more keen on following the player, and the other way around as well.

There is a setting, some will grow to the occasion some will not, the only problem is that the player may not choose to be the leader... and the game would never unfold.

Unless the other NPC's actions would unfold anyway...

Justin Keverne
profile image
"If two characters are motivated to kill each other and the player or other characters do not act to stop them let them kill each other. The player doesn't need to witness such events but they should, like all other characters, be affected by the consequences."

At the heart of such a system is the concept that NPCs are active agents that do not wait for the player to initiate action. The player can become the most important character in the world by inserting themselves into the lives of every other character, or they can choose to only focus on a limited subset of characters.

The degree to which reactions to player actions are weighted can be left under designer control, a factor I think I discussed in Part 1.

Bart Stewart
profile image
Interesting concepts effectively presented. I think I'd really enjoy a game designed to implement most of these ideas.

Actually... that may be happening now. Brian Green (Meridian 59) and Stéphane Bura, among others, are working on a concept called Storybricks that would create a shared persistent world in which NPCs act according to detailed sets of goals and motivations. Uniquely, a core element of this design is that players are to be given the ability (through a easy-to-use but highly expressive tool) to create characters and assign to them these goals and motivations, allowing players to participate directly in storytelling development. Jonathan, I think this also speaks to many of the suggestions you've made here.

The Storybricks system is, in a way, a Narrativist response to the combat-and-loot-and-levelling emphasis in virtually all MMORPGs today. It's definitely one of the most exciting things I've seen in MMORPG design in years, and is directly relevant to the storytelling-through-NPCs concepts that Justin has proposed here.

More details at for those interested in this kind of thing.

Aubrey Hesselgren
profile image
Excellent write up. I've wanted to do something pretty much exactly like this for years, but always struggled when it came to execution, especially on the AI side, but also because this kind of gameplay feels almost like meta-gameplay: more about AI playing with provided systems than anything else. For it to work, it seems like you need to try and fail quite a few times to establish the gameplay-space you need for these higher level rules to be effective. I honestly feel like the only reason we don't see more of this stuff is because of that massive barrier to entry. It otherwise feels like a natural progression of narrative in games.

Translating a character's high level motivation into base level verbs is tough, but not insurmountable. The concept of AI trees could be applied to this.

Tougher, is giving the NPCs the ability to consider how their actions will affect other NPCs, without going to infinite/recursive yomi loops. How do you give an NPC the ability/foresight to see what their ripples will cause? How will they plan to use these ripples to work in their favour? If they could plant a lie anywhere in the system to manipulate someone else into helping them, how do you work out the logic for where they should place that? I hope that'd be the same thing as the AI tree ("I want this outcome, so the causal route is self evident") but I don't really know enough about AI to be sure.

In the same way, how do give the player perceivable consequence for their actions when it comes to "nudging" an NPC's behaviour in a way that benefits them?

It'd be great to be able to, say, intercept information between NPCs in order to affect the behaviour of one for your own means, but keeping it simple enough that the player can get the intended result is tough. Would the game end up feeling like a big chaotic lake, either through complexity (one agent's choice affects every other agent, which affects every other agent) or obfuscation (the player can't see the results of their actions, or the full chain of events)? I think your point about feedback via newspapers, gossip etc. is key. In effect the act of registering information in such a way helps to slow the chain effect of causality, putting it almost into turn based territory. i.e. every turn, or every X seconds, a piece of information can flow from one character to another.

As with anything, I think this stuff needs to start really simplified and small: not just for development, but also for the sake of the player being introduced to a fairly conceptual game mechanic. I've always talked to a friend about something like a Manor House Murder, where the agents are limited but the relationships relatively deep. You could increase the number of agents each time the player solves a murder, so that the player becomes accustomed to the complexity of the interactions. Start in a cottage, move up to Buckingham Palace.