Central Issues in Interactive Narrative Part 4 – A discussion with Chris Crawford

In part 4 of the discussion, Chris talks about his latests project Siboot and identifies “five killer tasks in IDN,” while I bring up the question how the audience understands IDN.

Part 1 part 2 part 3

Chris writes:

You’re right that the Holy Grail is beyond our reach; I have tacitly admitted as much with my Siboot project, which is half-game, half-storyworld. The interactive storytelling in Siboot uses only about three-quarters of the technology in SWAT; I have ripped out a lot of the more complex components. I dumped the ability to travel to stages; that’s all handled directly by the Engine. I limited Actor attributes to just three; I replaced the Tinkertoy Text system with an iconic system that is computationally simpler. Only two actors are allowed on a stage at once; no three-party interactions are allowed. There’s no hiding, no spying, and no automatic lie detection. I can’t recall some of the other simplifications I have made.

I do believe that we should be able to get something like soap opera working in a few decades. The trick is to purge our thinking of plot-related notions and concentrate exclusively on character interaction. Yet even soap operas have mini-plots. I’ve been thinking of the idea of a set of mini-plots that are abstracted enough to permit different actors in each of the roles. The engine would then launch a new mini-plot whenever the count of active mini-plots fell too low. We’d need to come up with an abstraction of the idea of plot; I have some ideas about how to do that. Then we’d stock the story world with a goodly set of mini-plots.

I have some dissatisfaction with the point you make about exploring new narrative structures. The source of my discomfort is the long history of that notion. I still recall a contentious Phrontisterion in which I was trying to coax the group into coming to some sort of conclusions, no matter how vague. A good number of the participants refused to support ANY overall conclusions, arguing that the field is simply too poorly-developed to admit any such thing.

My objection to this is that we have to start somewhere. We have long since learned that branching trees don’t work very well — yet we’re so open-minded that we don’t discourage newcomers from pursuing that goal; time and effort are wasted re-inventing square wheels.

I recently wrote a piece listing what I believe to be the five killer tasks in IDN, the central problems that must be solved before we can hope to accomplish anything. I place them in order of increasing importance:

1. An IDE customized to the system; IDN is simply too complicated to be built in a general-purpose language.
2. The display of faces with emotional expressions appropriate to the context.
3. A personality model that is USEFUL for guiding actor behavior, and SIMPLE enough to be practical.
4. A narrative engine for executing everything,
5. A linguistic user interface. You simply cannot meaningfully interact with actors without using language.

I’ll offer this argument to justify the replacement of boolean variables with floating points, which you accept but don’t seem to consider to be of great importance: how can you differentiate between the affections that one person holds for others in a group? Does John love Jane, hate Mary, love Tom, and hate Jim? If you don’t use floating points, how can you handle this kind of problem?

I agree that simulating the real world is all wrong. Instead, we must simulate the narrative process. Putting in floor plans for houses, weights and values of different objects, distances between different points — all that stuff is a waste of time. We’re  simulating PEOPLE, not things! So we simulate how people respond to different events, depending upon their personalities and histories.

The idea of the genius-led project is not to limit everything to a single genius. What I envision is a large number of projects, each with its own genius. The creative act is, for the most part, the act of one artist. There can be many specialists who assist that artist, but the artistic vision cannot be seen by any committee. The cinema has developed a good system of specialties that support the central genius: the director.

The problem we have seen with collaboration, as you have already noted, is the incompatibility of visions that makes such collaboration possible. Everybody has their own idea of how to do it; indeed, we don’t even have a common vision of what it is that we’re doing. Is it “interactive storytelling” or “interactive digital narrative” or “ludic narrative”? Think of all the different terms we have seen here.

Of course, this line of reasoning supports your argument that we need more exploration before we start nailing things down. But I grow impatient with what seems to me to be floundering. Sure, we’re exploring lots of opportunities — but we haven’t actually discovered anything yet! Perhaps my impatience is due to having struggled with these things for 20+ years. Perhaps it is due to the distant cloud of mortality that I spy on my horizon.

As to Siboot, I have far too much to say about it, and what I would say is constantly in flux. You could struggle through some of the essays in my Siboot design diary (http://www.erasmatazz.com/library/design-diaries/design-diary-siboot/index.html). It’s all rather disjointed, because it’s a diary, not a final design document. But reading a few of these might convey some idea of what’s happening with the design.


Hartmut replies:

You make many good points. I like your list of important tasks and your pluralist take on the genius. I am aware of the difficulties of collaboration, but have not given up yet. I’ll read the Siboot documents to learn more about the project.

I am also a bit more optimistic in regards to what has already been achieved. From Toni Dove’s work and Brenda Laurel’s projects to Balance of Power, The Sims, Facade, Dear Esther and the Walking Dead Game, there are quite a number of milestones.

However, I feel you circumvent the point I make about the audience. Let me put it differently, then.

Michael Mateas has always urged people to stay away from “cheap tricks” like random functions and instead favored complex computational processes. I understand this view and deeply appreciate it. However, things are different, once we switch to the audience’s perspective. A computer program is opaque to the majority of its users and as a result, members of the audience create their own abstractions and assumptions of the underlying mechanics. The perception of the output of a random function can be identical to one that results from a sophisticated algorithm. Audience members therefore might feel led on or “tricked” in the sense of cheap stage craft if they do not have an understanding how a particular reaction and turn of events has come about. At the same time, to the algorithm’s creator this can be a frustrating experience, as her/his work might not be fully appreciated.  Reacting to this problem, Michael Mateas in his 2010 ICIDS Keynote talked about the need to make the underlying complex AI mechanics more visible to the user. For a concrete implementation he showed a screenshot of Prom Week with many value meters (the published version has less of these indicators).

I feel that Michael’s comments indicate an epistemological problem here – so far, we have operated under the assumption that once we get the mechanics and the experience right, all will be good and people will just intuitively understand and appreciate what we are doing in enabling a novel way of expression that can communicate the complexity of the human condition in a post-modern reality.

However, our creations do not have the advantage of being easily perceptibly different from the moving image of TV. Yes, there is the controller and the clicking, but my point is there is not the immediate gut reaction to the first showing of film in the  late 19th century. With film, the difference in perception was undeniable, with interactive narrative, it is a lot less obvious. Roy Ascott already in the 1960s saw this problem when he urged artists to make cybernetic systems but cautioned that such works would be perceived as coming from the “old mold,“ as versions of the object art of old.

What does not help in this regard, is the way many in the field relate their work to earlier forms. The devil is in the comparison. If we aspire to make something that is like something else, the danger is that what we create might be perceived as iteration of the original and not as something that should be understood in a new way. Consider what you wrote “I do believe that we should be able to get something like soap opera working in a few decades.“ With “soap opera“ there is already a comparison. The problem is, once your system runs perfectly, it might look and feel like a soap opera. And to be clear, I am very aware that it is very difficult indeed to talk about what we are doing without resorting to comparisons with earlier forms.

Maybe the way out could be to escape the box. A work like Text Rain makes the difference obvious. Toni Dove’s earlier work did, too. Maybe the Holodeck is a good metaphor in exactly this way. Not in the sense of the perfect illusion that the Star Trek series depicts but as a way to differentiate from the TV box and the cinema screen.

There is work needed to give the audience a frame of reference other than the fixed narratives of old. We need not only get better in what we are doing, we also need to get better in making people understand what we are doing.


Am 25.10.2014 um 14:44 schrieb Chris Crawford <chrisc@storytron.com>:


Central Issues in Interactive Narrative Part 3 – A discussion with Chris Crawford

In part 3 of the discussion, Chris expands on the disciplinary divide in the field, before discussing a communal approach versus the single genius. Finally, he lays out his approach and talks about his current project Siboot. In turn, I talk he audience and the difficulty to extend something as well known (by almost everybody) as narrative.

Part 1 Part 2


Wow! That was a great piece. It’s sort of “A Quick Summary of the Field of IDN in 2500 Words or Less”. I don’t keep up with what others are doing, largely because I don’t have access to the printed academic literature and I can’t go to the academic conferences. On a deeper level, I’m a troglodytic lone wolf anyway. I *like* my cave!

I agree and disagree with everything you wrote — and that is my standard for perfection in writing. A good essay should include just enough agreeable content to command credence, and just enough disagreeable content to stimulate thought. I think you hit the perfect balance here.

Your characterization of the two schools of thought (AI versus arts) is an echo of the all-pervasive Two Cultures problem that has bedeviled our culture for more than 50 years. The fact that the former school dominates game design explains its artistic flaccidity.

Being an ornery cuss, I have deep criticisms of both sides. The AI approach seems overly boolean to me. AI has developed so many wonderful tools that AI people can’t help but want to apply those tools to the problem. Unfortunately, they’re all carpentry tools being used on marble. The whole boolean approach imposes an overly simplistic black and white mentality on the infinitely subtle processes of drama. Michael Mateas recognized this with Prom Night and used 4-bit numbers to obtain greater resolution, but I think we need to go all the way to floating point. To put it another way, I think we need to stop thinking in terms of AI and more in terms of simulation. I’m sure that’s a reflection of my training as a physicist, and I confess that this may constitute an unconscious prejudice on my part. But I could be right; doesn’t character simulation seem a more promising route than boolean branching?

The arts side, on the other hand, seems too unwilling to get its hands dirty with actual hardcore programming. I see all sorts of clever ideas emanating from that side that are obviously uncomputable. Many of the other ideas rely on a primitive approach to programming using an overly high-level language. I like to remind people that the Renaissance artists dissected rotting human corpses in order to better appreciate human anatomy; why can’t modern artists find the mettle to get their hands bloody with digital gore?

The fundamental thing that both sides miss, in my physics-centered weltanschaung, is the necessity of modeling character interaction using numeric algorithms. Boolean algorithms don’t have enough resolution to model the subtleties of human interaction, and all those damn fixed branching systems are equally ham-handed.

There’s also the whole verb-centered thing that I have been banging my drum about for decades, but let’s set that aside as old news.

Here’s a point that I think you didn’t quite notice, although you danced around it: the difference between a communal approach and a “genius” approach. You naturally assume a communal approach to solving the problem, and I can’t deny that, over the long run, that is the necessary approach. Interactive storytelling is, as I have already moaned, too damn complicated to be conquered by one person; it will take the combined efforts of many people to truly solve this problem.

But my own personal history inclines towards the “genius” approach, with ME (but of course!) playing the role of Genius. I think that I can lay claim to have made more mistakes than anybody in this business, which gives me a leg up. I believe that I am close to a working system; I have so many technologies working. Of course, that could well be the undying optimism of the troglodytic lone wolf; I clearly lack sufficient objectivity to make a reliable assessment. Then again, nobody else knows enough about my technologies to make a reliable assessment, either.

So far, despite numerous attempts, there is no central concept that has attracted enough credibility from others to provide us with a useful starting point. Just now, we are in exactly the position that the artists crave: a confused array of individual ideas. This is a phase we must go through, but at some point we must begin a crystallization process around one idea that provides the best foundation on which to build. The community can then expand and refine that idea, adding more technologies to it, while the artists act as scouts exploring the peripheries, as is their charge.

For the last six months, I have been toying with the idea of making my technologies open source. That would require some sort of PR effort. I detest PR and am correspondingly wretched at it. So for now, this is just a fantasy on my part; I’m in the process of wondering how that fantasy could be realized.

I know that many parts of my technology can be used separately. I would break it up into independent pieces as follows (I place them in rough order of easy adaptation to other systems):

The three-factor personality model with perceptions, accordances, and weights.
The core dataset of Verbs, Events, Stages, Actors, Props, and Quantifiers
BNumbers for computation
Verb-centered design
The Deikto metalanguage system
The reaction-based engine
The Sappho scripting language

I am certain that some of these technologies would be useful components of any successful interactive storytelling system, but I’m still daydreaming about how to proceed.

Now onto the question of “what interactive narrative is really about”. This question doesn’t stimulate a lot of thought in me, largely because I feel that it has been adequately answered. First off, I think that we can agree that none of the working systems we have seen so far are *really* interactive narrative; there’s nothing we can point to and say “That’s it! That’s the right approach! Therein lies the future!”

For me, there’s an easy answer to the question: interactive narrative is a system that permits the player to engage in dramatically significant interaction with interesting characters. Note that this perception commits me to the character-based school of thought on writing. Some writers are plot-centered: they design a plot and then fill in the details. Other writers are character-centered: they create an interesting set of characters, throw them together, and see what happens. I believe that the plot-centered approach is inimical to interactivity; if you already know where you’re going, you can’t give the player any freedom to interact. The character-centered approach, on the other hand, fits hand-in-glove with interactivity: the player is simply one more character in the cast and does the same thing that everybody else is doing.

This means that we have to throw away the whole idea of the drama manager (whom I call Fate) built around plot. Instead, I think we must think more in terms of soap opera: the open-ended series of interactions among characters. Fate umpires the interactions but does not provide long-term planning. (Although there are some easy ways to insure long-term development without intruding on the free will of the player).

An interesting aside: I clash with academics on terminological issues. Most academics prefer to shanghai a term that’s close enough to their meaning to be plausible, yet far enough away to permit them to assign it a precise meaning for their discipline. I prefer to take use terms from closely related disciplines, such as fiction and theater. Thus, academics say “agency” where I say “free will”. Academics say “character” where I say “actor”. Their “locations” correspond to my “stages”, and their “physical objects” are my “Props”. I wonder what it means.

I’ll conclude by observing that I agree with  you that a modular approach is most promising, but I fear that we’ll never be able to agree on any modules, no matter how simple or fundamental. I believe that the first two items on my list above are no-brainer modules that provide us with an excellent starting point, but I very much doubt that others will accept them.

My current estimate is that the only way to get the ball rolling is to get something actually working, something impressive enough that people will sit up and notice and be so impressed that they’ll embrace the underlying technology as something to build on. That’s my goal with Siboot.



I agree with much of what you are saying, but maybe I can emphasize and clarify some aspects. Let me start at the end: I’ve grown more and more skeptical of the Holy Grail of the big, wonderful work that convinces everyone how amazing and necessary interactive narrative is. Kevin Brooks’ system didn’t do it, Toni Dove’s performances didn’t do it, Erasmatazz didn’t do it, Façade didn’t do it, GRIOT didn’ t do it, Prom Week didn’t do it. And all of these are wonderful, groundbreaking works.

The one masterpiece to bind them all, open their eyes and let them see the wonders will not come. But I am not convinced anymore that we actually need it. What we need instead is an audience who understands and appreciates procedural, varied, and reactive narrative. An audience that is literate in digital media. This audience is growing up now and will develop an interest for deeper characters and more involved narrative development than the average AAA game can offer. Its members come from Diabolo and Skyrim and Bioshock and The Sims and many other games and will understand the intricate mechanics of Façade much better than many mainstream journalists who wrote about it. It will be a gradual process, but eventually, the digital generation will demand space for procedural masterworks in the galleries and library collections.

But we are not there yet. The majority of the potential audience is far behind in comprehending what we are working on. Sliding Doors, a movie with one (!) branching point, is a good indication in this regard. Most movies that attempted to show more branches/parallel narratives were not very successful (e.g. Cloud Atlas). I am NOT putting the blame on the audience here, but we need to acknowledge how fundamental a change is we are working on. What we are trying to do is to extend – and thus change – narrative, something that most people feel they have a pretty good understanding of, one that is based on thousands of years of tradition. This is why I maintain the best way to convince people that interactive narrative is something worth looking into is by exploring new topics, modes of interaction and sequencing.

An IDN work might perfectly simulate a stage drama, a novel, or a short story and as such will be a major achievement; however the full extend of this accomplishment will only be appreciated by the initiated few. For the rest of the audience, it will be derivative of the original form, an „interactive add-on“ to something already known. I am convinced that „interactivisation“ of existing forms is not the way, instead we should boldly go where narrative has never dared to venture. IDN will come into its own, once it presents narratives that traditional media platforms cannot. Characters that are nasty or nice and everything in between depending on the participants decisions and behavior, the experience of consequences resulting form the participant’s decisions, cross-session memory, a plurality of views between which the users can switch at will. And then it comes to forms and genres we are only at the beginning.

Early on in the development of AI characters, the researchers determined that life-like behavior and intelligence was not the main goal, but instead believability. I feel the same is true for the overall product of interactive narrative, which means as long as the participant had an engrossing and satisfying experience, the underlying technology has shown its value. Yes, I can absolutely understand how the move from binary two 4bit and floating point dramatically improves the range of possible expressions. However, I am not convinced that this development will automatically yield better characters.

The problem with the simulation is what to simulate. Narrative has always been the art of exaggeration and compression. A protagonist who stands in for many others, a village that represents the aspirations and psychological afflictions of a whole country, the space of a century covered in a book. I think it was you who has classified the The Sims as „life with all the boring parts put in.“ I tend to agree with this one, especially given an acceleration mode, so you do not have to wait forever just to see you SIMS sleep. Janet Murray suggests in a chapter in the upcoming book I co-edited to look at dramatic narrative situations in classical texts such as the King Arthur sagas as the „source’ for the simulation, and not real life. Such characters do not have to have the fidelity of real life humans’ emotional palettes, but they have to be convincing in their role.

I am skeptical of the Genius approach, the single leader to the promised land, and that might also has something to do with coming from Germany. Yes, I believe in a modular and communal approach. And I do not think we need to agree on an overall vision for narrative or specific modules to be able to have technical standards that allow us to combine different people’s work. I envision something more akin to a plumbing system, a TCP/IP networking layer. A draft exchange standard, hosted on a web server, together with a repository of components. The different components of your work could find a place where, as well as the works of many other researcher. This central platform could also solve the PR problem.

I feel given the potential gains we should make collaboration a priority. I’d be happy to see your knowledge on verbs applied in in another researcher’s or artist’s work.

Can you say more about Siboot?


Central Issues in Interactive Narrative Part 2 – A discussion with Chris Crawford


In part 1, the discussion started around the question of authorship in interactive narrative, this second part concentrates on the issue of computational story models and how different perspectives in this interdisciplinary field affect the discussion and research agenda. I also bring up the question of enabling collaboration through exchange standards.

Chris continues

I am pessimistic about the prospect of making interactive storytelling authorship available to Everyman. This was in fact the goal of the Erasmatron, and it was a spectacular failure — the system required was far to complicated. With Storytron I made some aspects of the technology more accessible, but the underlying model grew more complex.

The central problem, however, is the difficulty of creating a computable model for storytelling. Every effort I have seen relies on a different conception of the structure of a story and the process of storytelling. We cannot collaborate because we all see storytelling differently.

I often illustrate this problem by telling two stories:

Itsy bitsy spider crawled up the gutter spout.
Down came the rain, and washed the spider out.
Out came the sun, and dried up all the rain.
And the itsy-bitsy spider crawled up the spout again.

This is a complete and well-formed story in four sentences. It has a protagonist, an antagonist, a crisis, a resolution, and even an edifying moral. Furthermore, any 4-year old child can immediately grasp it.

Once upon a time there was a handsome prince who lived in a shining castle atop a hill.
One day he leapt atop his white charger and galloped out of the castle, over the drawbridge,
and into the forest.
Where they fell into a hole and they both died.

This story contains many of the components of stories: the handsome prince, the castle, the white horse — but it is most definitely NOT a well-formed story. Any 4-year old child can recognize that.

Challenge: write a computer program that can accept these two stories as input and determine which is well-formed and which is not.

What boggles the mind about this challenge is not that we think it difficult — we can’t even think of HOW to solve it. The processing capability that any 4-year old child has is beyond our ability to conceive.

So yes, we desperately need some kind of abstraction that permits us to design systems for understanding and controlling the storytelling process. We need a science of interactive storytelling before we can have a technology of interactive storytelling.

But science did not spring fresh out of the minds of great scientists. The first true scientist, in the modern sense, was Galileo, and he often worked backwards from engineering and technology to induce the underlying physics. If you read his book on physics, you realize just how hands-on this guy was; the book is chock-full of examples taken from real-world construction techniques.

So we need some storytelling engineering before we can develop the storytelling science that will help us build good interactive storyworlds, The snake eats its tail.

I have walked a long way down one path, thereby generating a lot of experience. I can tell you a million ideas that don’t work, because I’ve tried them all (but there are billions more ideas to explore).

Here’s another kicker: because storytelling is so complicated, anything we build will necessarily be huge in order to adequately reflect a minimum level of fidelity to storytelling’s complexity. That cannot be accomplished by a single person. Yet there are no Apollo projects in interactive storytelling. We need an Apollo project that starts off with a Mercury program (just getting an astronaut into orbit), then a Gemini program (long duration space flight, orbital docking), before we can have an Apollo program. I think of my Siboot project as a Mercury project. It will get something working that just barely qualifies as interactive storytelling.

Perhaps the onus is on me to take the time to get the funding to build the team to complete this Mercury program.


Hartmut replies:

Fundamentally, there are two approaches to interactive narrative. One comes from science, the other one is based on arts/humanities. This became fully apparent to me at the final panel of ICIDS 2009 in Portugal, where Marc Cavazza and Martin Rieser discussed the next steps in the field. Marc emphasized how the focus should be on the “hard work“ in AI in making better algorithms, while Martin talked about the need for experiments to explore new forms of narrative.

This is a very foundational difference in perspective. In theory this is what makes this field interesting and vibrant, the interdisciplinary nature. Unfortunately, the reality is more complicated, since the two sides might not fully understand each other. As a humanities-schooled scholar and an artist, I am on the side represented by Martin here; yet as a systems builder I sympathize with and understand Marc’s perspective.

For the sake of discussion, let me abstract (and purposefully exaggerate) the two positions:

Science/AI: We need to work more on understanding and simulating existing stories/stage dramas/improv actors to build computational systems that can output narratives as convincing as the aforementioned forms.

Arts/Humanities: This is an opportunity for novel ways of expression that go beyond established forms of narrative. We need to experiment with this wonderful new artistic toolbox to see what we can come up with, but it is clear that we can do things impossible in non-interative media and that the results will not look and behave like traditional forms.

So this is where the crux of the problem lies: what interactive narrative is really about. The artists/humanities position (again exaggerated for the sake of discussion) feels that the science/AI people spend their time modeling OLD narrative forms, while what drives the artists is the opportunity for NEW narrative forms. In other words – the artists feel that the science people are busy modeling BEFORE we even know what we want to model. Artists are very skeptical regarding the underlying notion that structures and tropes from non-interactive media can be successfully transferred and will yield interesting results.

Like you said, successful, convincing narrative is hard. And while we like to think of narrative as a general concept that transcends time and media, when it comes to practical training, writers, filmmakers, game designers, and oral storytellers go to different schools for a reason. In other words, successful implementations are platform specific. Film had to develop its own concepts from the development of montage to continuity editing to specific framing to time-lapse to slow motion to the opportunities offered by color film, CGI, and 3D. Currently, practitioners are in the process to find similar conventions in interactive narrative. What kind of topics seem to work best? What kind of structures emerge? What kind of genres? What modes of engagement are satisfying for users?

In practice, these positions are less clear cut, but some broad trajectories can illustrate the point about different perspectives further. From Brenda Laurel to the OZ group to Façade, interactive drama focused on making (neo-) Aristotelian drama interactive by means of AI. Michael Mateas explicitly draws the connection to the ancient predecessors by writing that each walkthrough should have the emotional force of Greek drama. Marc Cavazza was involved in building an interactive version of Madame Bovary, i.e. an interactive novel. Post Façade, Michael has focused more on simulating “social games“ (as in Prom Week) and has lately advocated (FDG 2014) to read books about people’s behavior in society as a basis for interactive narrative. Brian Magerko works on understanding and simulating improv actors in his research lab as a basis to build more convincing virtual characters.

On the other hand, artists like Toni Dove, Martin Rieser, many members of the ELO organization or the Tale of Tales duo create interactive narratives as a way to experiment with new narrative forms, with little regard for existing frameworks and concepts. Pamela Jennings in 1996 already denounced the Aristotelean model as unfit for the digital medium, although her substitution (cyclical narrative based on African oral traditions) is not automatically more convincing. Janet Murray’s fundamental understanding of digital media as defined by specific affordances and phenomenological qualities points in the same direction (and this aspect might be more important than the vision of the Holodeck). Her description of the “Cyberbard,“ a new type of procedural creator also marks the departure from the traditional author. Similarly Ian Bogost’s consideration of “unit operations“ offers a fresh perspective on understanding narrative mechanics in digital media – as does your book.

Marie-Laure Ryan and other scholars focused on the analysis of interactive narrative can be seen as representing a third perspective, which could help to bridge the gap between the other two, especially once the question of work vs. artistic statement is explored more. In literature studies and any related related discipline, it is well understood (Barthes “Death of the author”) that works can and should be analyzed by themselves, regardless of author’s biography or claims regarding the work. Michael Mateas might very well see Façade as a neo-Aristotelan one act interactive drama, but good arguments can be made that it is altogether a very different work, one that is more narrative experiment than it overtly announces.

With this perspective in place, let’s look at your examples. You are absolutely right, the discussion about which is the right kind of model will continue indefinitely as long as we are focused on this aspect. However, the artistic/experimental side might find neither example particularly interesting and instead ask you for accessible tools to experiment with, irrespective of your particular model. Instead of the complete solution, they might ask you for building blocks and connections. Once people start using an incomplete solution, they will figure out what they can do with it and they will ask for the specific enhancements they want. This is my experience with ASAPS and it means that I sometimes work through the nights to accommodate my users’ wishes.

I don’t see why different approaches cannot be combined to great effect. I feel what needs to happen is a perspective change that allows different people to work on different things. The world wide web is a success story of how one can build an infrastructure that grows over time. The initial technical capabilities might not have been that impressive and were clearly inferior to Ted Nelson’s vision of hypertext. However, the capabilities have expanded greatly and during the more than 20 years now, the WWW has survived browser wars and attempts at hostile takeovers. Altogether this might be a better model than the Apollo program, which, while successful has not be sustainable and as far as I understand could not even be replicated right now, as the construction plans have been lost.

What I envision is a modular approach, where you could offer a component that speaks to your particular strength, maybe one that allows artists to explore verbs in interactions, where Brian Magerko’s actors could be NPCs, Michael Mateas’ drama manager would control the sequencing and ASAPS could supply a meta-narrative structure. And in another instance, Aylett/Louchart’s agents would take over, combined with Michael Young’s work, Mark Riedl’s sequencing and Jichen Zhu’s story generation. Ideally, artists could supply elements in this way, too, for example some of Toni Dove’s work. Many more scholars and artist could be mentioned in this regard, this list is only meant as an illustration, and the potential for growth is huge.

For this to happen, we should work on the HTTP/HTML standard equivalents for interactive narrative. And we should think about the equivalent of a W3C organization to manage the standard and help it grow. This could also be a clearing house for ideas and it is here where your vast experience in things that work and millions of things that don’t would be invaluable.

Maybe we should try to get funding for this organization.


Central Issues in Interactive Narrative – A discussion with Chris Crawford

I wrote a paper summarizing central issues in interactive narrative, to be published in the Springer Lecture Series in Computer Science: Five Theses for Interactive Digital Narrative.

My intent was to start a discussion on the state of affairs in the field, on achievements, pertinent questions and areas for future research.

Chris Crawford reacted quickly and we entered in an ongoing discussion, which Chris and I decided to present here, inviting readers to comment and offer their views.

As it has grown long (and is still continuing at this point) I decided to present it in installments, although the complete discussion will be available, for those who want to read it in its entirety.

Several overarching themes appear in the discussion:

– The state of the art of the field: Where are we, what has been achieved

– Science vs. Art: Interactive digital narrative is an interdisciplinary effort that includes computer scientists, artists and humanities scholars bringing different perspectives and goals

– Genius vs. Community: Is it the single genius/beacon of light project that will continue to drive development in IDN. An important sub-topic here is the question of collaboration

– Concepts and Practice: How do conceptions of interactive narrative influence research and practical implementations? How do they influence the audience?

Enough with the intro. Here comes part 1:

part 2

Chris’ initial email :

I just read your paper and I have one immediate reaction to your Point #5 (IDN Needs to be Author-Focused). This was fundamental to my work, first with the Erasmatron and later with Storytron. I now believe that I erred in designing the system for less technically-inclined authors. The killer truth here is that storytelling is complicated. It is not possible to build a clean, simple, easily-grasped system for interactive storytelling; any workable system is necessarily big and complicated. That in turn imposes a steep learning curve on the author.

A second killer factor is the primacy of algorithm creation. Storytelling is fundamentally about human processes. Drama is not a static statement of human facts, but instead a description of human emotional processes. Such processes can only be described using algorithms, and algorithms can only be expressed adequately in mathematical form. Even boolean math is too simplistic to capture the subtleties of human emotional processes. Fully-fledged numeric algorithms expressed in algebraic form are required.

This imposes a huge constraint on the author base. Let’s face it, there are damn few decent storytellers who can also sling around algebraic expressions.

I have come to the conclusion that, for the nonce, we must place all our hopes on that tiny number of people who combine artistic talent with mathematics nous. I think that the ideal arrangement is something like Joseph Brook’s proposal in The Mythical Man-Month: the surgical team, in which a single expert surgeon is aided by a team of enablers. I have not been able to assemble such a team; perhaps you can.

The necessity for such an arrangement will, I hope, be temporary. Once we have trained a cadre of talented individuals who grasp both sides of the problem, we will be able to build standard tools enabling a larger group to tackle the work. Until that time, standardization is premature. We need broad exploration, along with many failures, in order to discover the best standards.

Moreover, we must recognize that building interactive storyworlds is a complicated task requiring considerable expertise, expertise that can be developed only through years of training and practice. As yet we cannot ask talented people to make the huge commitment required for this; we need some pilot projects clearly demonstrating that such efforts can yield worthy results.

Thus, I think we need to focus our energies on a few “Manhattan Projects” (“Apollo Projects” for the less bloody-minded) to demonstrate the power of the medium. We need a Hiroshima (or a moon landing) to motivate the world.

Hartmut’s reply:
I absolutely agree on the need for more authors and that they need to be educated about procedurality. However, my approach is from the other direction, the low end of procedurality, and rather than the “Tiger Jump” approach a la Façade/the Manhattan Projects/, I have an iterative approach, slowly building up capabilities.

This has to do a lot with my background – before I entered the field I worked in tech support and as a journalist writing articles for computer magazines. In both capacities I learned how little most people know about computers. I’ve walked countless people through seemingly fail-prove software installs, learned that even highly intelligent people could find computer tasks challenging and that without extensive training programs, computers would go to waste even in seemingly well-run companies.

Contrast this with the our field. All of us are highly-skilled computer geeks (yes, this includes the artists and humanities scholars) FROM THE PERSPECTIVE of the average person out there (yes this includes most prospective authors). To be entirely clear – we might NOT SEE OURSELVES as anything approaching technical geniuses, but for the average Jane and John we might as well be. This fact creates a huge barrier to overcome, maybe much more psychological in nature than related to the actual lack of skills.

Many existing tools are still an order of a magnitude too complicated for the target audience of prospective authors. There is even less of a chance that specialized programming languages like ABL will become commonly used. This is why i created my ASAPS system on the low end of computational complexity. I can teach a college student to use it and produce an interactive narrative in one term. Right now I have more than 90 works created with it, including one published the AppStore for iPads. I believe in slowly building up the capabilities of the system, but also in fostering collaborations with other researchers and their systems.

Jay Bolter told me once that everybody wants to have their own system in our field. That practice seems wasteful and unsustainable in the long run. I think we should focus more on how our tools can work together. Nicolas Szilas et al have published an exchange standard (Oparis) and that is a good start, but we need more. In May, I started to work with other researchers to hook up my ASAPS system to their AI system. The idea is that ASAPS would provide a front-end, but also a meta-narrative layer for the autonomous agents in their system. What we discovered in our discussions was that there is a need for an exchange layer between the systems, one that contains abstractions that both sides understand. I would be very open to more collaborations.

I do believe that advanced computational algorithms will play an important role in future IDN systems. However, the task is not done when that algorithm has been created and implemented. Nor is it done when it has been demonstrated in a single milestone project. The (mostly overlooked) challenge is to find an abstraction and concrete user-facing implementation that allows the average author to grasp the concept and use it well in her/his projects. In a sense what we have right now is the equivalent of the computer before the gui – we have neither the abstraction of the desktop metaphor nor Steve Jobs’s genius in polishing and packaging the parts and brining them to market.

The question of abstractions is crucial and it is here, where theory and humanistic perspectives enter the picture. Roy Ascott already in the 1960s (!!!) saw the opportunities offered to artists by digital technology, but crucially he understood that most people will fail to see the true novelty of systemic/cybernetic/procedural works and therefore will describe them as variations of non-procedural artifacts. This problem continues to this day. The static, fixed artwork still provides the lens through which procedural systems are understood. Only very slowly, through institutions like the ZKM and ars electronica, through the electronic literature foundation, curators and specialized art dealers, the perspective, the lens, is changing and the constant comparisons to „proper artworks“ become less prominent. However recent discussions wether video games (referenced here) can be art are testimony that the process is far from over.

In interactive narrative, we are only beginning to create abstractions that allow people to grasp the difference and specific opportunity in comparison to traditional fixed forms. As long as interactivity is understood only as an „add-on“ feature, the attraction will be low. Most people are quite fine with movies, novels, and the stage play as they are. They do not see the need for interactive fiction, interactive movies, or interactive drama. The point is that IDN is a new expressive art form, and needs to be understood and promoted as such. Instead of shoehorning concepts or metaphors of old (Freytag’s arc, Aristotle’s concepts of the well-formed plot and complete action) into the world of computers, we need to work on fitting metaphors and terminology that marks the difference.

This work is as important as the work on the technical side. People cannot see what they do not understand. Michael Mateas’ frustration with the fact that the average user did not understand the brilliance of the underlying AI in Façade and his determination to make it more visible in Prom Week is an expression of the problem. Most people might never fully understand AI, but they could appreciate a framework of abstractions that explains to them what an reactive procedural work is that lets them try different courses of action and experience the resulting consequences and outcomes.

I do not believe that any Apollo project will solve the abstraction problem. Audiences might be in awe, but still not understand. Is Façade a game? Not to me or you, but that’s how it was described in most publications. Not that I have a big problem with using the term “game“, but it creates the wrong frame of reference.

We need more bridges between he technologists and the abstraction people (humanities scholars and artists), who in turn need to start emphasizing on what is different. We need to connect our tools because the problem is too big for any single person or research group. Finally, we also need to educate people, conceptually and with easy tools, so they can move on and start creating works.

part 2 part 3 part 4

Save the Date – Cross-session Memory, Metanarrative and a Challenge to Endings

Save the Date is a real gem from the perspective of interactive digital narrative. This game puts the affordances of digital media to great use by demonstrating how procedurality and interactivity can extend narrative. At the same time, this work challenges long-standing conventions in both games and narrative.

So what does this game do? Save the Date reflects the player’s growing knowledge in consecutive replays though an evolving narrative. In addition, this work extends the notion of metanarrative in an important way and challenges the player to consider abandoning the game to reach a successful ending.

At the start of the game, the player invites a romantic interest – Felicia – to a dinner date. For this purpose, the player can chose from a variety of restaurant options. However, regardless of the particular choice, the date will invariably die. Hence, the goal of the game is to literally “save the date,” which turns out to be rather difficult. For example, taking Felicia to a Thai restaurant will have her die from a food allergy to peanuts.

Bildschirmfoto 2014-06-24 um 23.44.01In subsequent attempts the player will now find an additional option to warn the date about the peanuts and thus avoid death at this level in the game. In this way, the game reflects earlier experiences, which means the narrative changes accordingly. And it keeps changing to include more warnings and ways to avoid earlier disasters. Yet, as the player quickly learns, using this knowledge directly is also risky, as Felicia will find the unexpected intimate knowledge rather creepy and as a reason to break up contact.

Bildschirmfoto 2014-06-26 um 02.26.56Eventually, the player will encounter a set of choices that reveal her as a player of a video game to the date character. In this moment, the game narrative becomes meta-narrative. In more traditional media forms, the meta narrative involves a character addressing the reader as such. In Save the Date the roles are reversed – the player addresses the in-game character and makes the date aware of her role as a character in a game. The subsequent conversation has Felicia reflect on her role and speculate with the player on how to proceed and on the nature of narrative.


An abstract representation of Save the Date shows subsequent sessions as substantially different from earlier ones:


Cross-session memory is rarely used as a narrative strategy, yet holds much promise in expressing complex and interdependent consequences. Save the date successfully demonstrates how cross-session memory can be used to enrich the narrativ experience by reflecting the player’s growing knowledge. The procedural aspect of digital media puts the ability to retain and reuse memory states at creators’ disposal, as yet another tool for narrative expression.

At the same time, the work explores a meta-narrative dimension that is unusual by being initiated from the player’s side. The game has the player abandon the in-game role to engage in a discussion of the nature of game play and narrative. The discussion centers on the question for a the winning state/successful ending. The only way to win, the Felicia character poses, is to stop before the official ending and use imagination the create a better ending. This proposal is not an easy one to follow as it violates yet another deeply entrained convention narrative and games share – to continue to the end.

Bildschirmfoto 2014-06-25 um 00.36.39

Save the Date is not only a great example of interactive narrative, but the work also constitutes a challenge for analytical perspectives that assume fixed structures on replay and stable endings.

Try it :



The Future of IDN: ICIDS 2014 Workshop

The Games & Narrativ group invites participants to a workshop on the future of Interactive Digital Narrative at ICIDS 2014 in Singapore.

After more than 25 years of fruitful research, starting with Brenda Laure’s 1986 PhD thesis, and productive practice in interactive digital narrative, it is a good time to consider future directions amid a maturing research field and a growing market for narrative-based interactive media. The Games & Narrative group invites participants to discuss ongoing issues as well successful methods and projects with us. On this foundation the workshop will enter into a phase of “futuring” – productive speculation – how will IDN look like in 5, 10, 25 or even 50 years? In addition to this, we will also take this opportunity to debate concrete initiatives, like joint research proposals, exploratory projects, forums for interdisciplinary dialogue, a central repository for projects or an academic/professional organization. The results of the workshop will be made available on the Games & Narrative website.

We invite a broad variety of participants to this workshop, including researchers, professionals, artists, publishers and critics. Our aim is to assemble a wide range of perspectives, in order to imagine future scenarios for our field that take into account the highest number of points of view.

Contact and registration:

futureworkshop2014 at gamesandnarrative.net

Defragging the Divide: panel at DiGRA 2013

Is Narrative an integral part of computer games, or is it just an accident?
The Games and Narrative group is happy to announce a panel at the DiGRA 2013 Conference in Atlanta, titled Defragging the Divide: narrative practices in current videogames (and how to understand them). Join un on Thursday 29 August at 11.45am for a discussion also featuring Janet Murray and Espen Aarseth.

A decade ago, the new discipline of games studies/ludology vigorously denied any connection between games and narrative (Aarseth 2001, Eskelinen 2001, Juul 2001, Frasca 2003). For example, Markku Eskelinen wrote that since a well accepted descriptive methodology for traditional games – he references The Study of Games (1971) – does not consider narrative as part of the ludic field, computer games should follow the same route. If they do not, as he implicitly suggests, it is more because of marketing strategies than of intrinsic characteristics.

Indeed, game companies are recently boasting narrative as a key selling point for their products. Steve Papoutsis of Visceral Games, best known for their “Dead Space” franchise, stated “Our goal […] is to apply our key design principles – immersive, challenging gameplay, intense narrative and focus on quality”. Rockstar Games recently announced Grand Theft Auto 5 defining it “a bold new direction in open-world freedom, storytelling, mission-based gameplay and online multiplayer”. Dan Connors of Telltale Games, underlined the “narrative design” in The Walking Dead games as a feature that appeals equally to hardcore gamers and to more casual gamers who are fans of the franchise.

To interpret these industry perspectives solely as marketing ploys seems unconvincing. Consequently, what we are faced with is a curious disconnect between theory and practice. Whatever the academic argument regarding narrativity in videogames, the industry clearly believes they are using narrative in some capacity. While industry did not concern itself with the outcome of the ludology-narratology debate, the game studies community should, as it concerns the understanding and expressivity of games as a creative medium. We are not satisfied with the outcome of the ludology-narratology debate from last decade, as specific theories for understanding narrative elements in computer games have not yet emerged. Perhaps the best openings into this puzzling situation is provided by Gonzalo Frasca’s perspective that the debate on narrative in video games never took place (vs. the debate of whether games should be understood as yet another text and be analyzed with the toolbox of narratology) (Frasca 2003a) and Marie-Laure Ryan’s identification of a pragmatic position regarding the application of narrative in video games (Ryan 2006).

From these vantage points we want to re-open the debate on games and narrative in a panel that includes academics on both sides of the debate, as well as industry insiders. Has radical ludology been refuted, not by narratologists but by game producers themselves? Or are new technological possibilities making narrative features emerge more prominently?
We will ask participants how their positions have developed, where they see narrative in games in 2013 and in what directions they see academia and industry heading.

Award for Haunted Planet Game

We are very happy to announce that Haunted Planet Studios, the company run by Games & Narrative Group member Mads Haahr, has received two awards at the Irish Games Festival for a game based on Bram Stoker’s (author of Dracula) life and narrative world.

Bram Stoker’s Vampires won in the categories:

  • Best in Gameplay
  • Best Original Innovation in Gaming

Bram Stoker’s Vampires is available on iTunes and Google Play.

Telling Ghost Stories with Physical Space, Part 2

In my last post, I discussed the location-based augmented-reality game Bram Stoker’s Vampires. I outlined the basic game concept and discussed how locations that were relevant to Stoker’s life as a student in Trinity College Dublin were used for geo-locating characters related to his novel Dracula. One of the characteristics of the game that I didn’t get to fully discuss in the first post is its narrative structure. I will do that in this post and I will also explain the flexibility in the narrative structure that our game engine affords.

Players of our games take on the roles of paranormal investigators, i.e., detectives who solve paranormal mysteries. For this reason, our basic narrative unit is a case (as in a case solved by a detective), which corresponds roughly to a quest, mission or level in many other games. A case typically takes 45-90 minutes to solve and should offer some degree of closure at the end. While the case is our main overarching narrative unit, it consists of smaller narrative parts, or encounters that are geo-located and linked together. During play, a player investigates one encounter at a time. Completing one encounter unlocks one or more in the chain. Our engine allows the encounters to be linked in a sophisticated fashion. The resulting experiences can be non-linear and personalized, such that different players playing together will get different experiences, which also improves replay value.

We distinguish between the following types of encounter structures:

  • Sequential: One encounter is active. Completing the encounter unlocks the next.
  • Parallel Inclusive: Multiple encounters are active and can be found in any order, but all must be completed to unlock the next.
  • Parallel Exclusive: Multiple encounters are active and the player chooses a single one, which when completed eliminates the others.

The launch version of our game Bram Stoker’s Vampires contains a single case with a total of eight encounters. The first three encounters are the three vampire sisters that I introduced in the previous post, followed by four encounters with Count Dracula and one with Bram Stoker himself. We structured the encounters with the three sisters as a parallel inclusive structure, which means that the player can find them in any order. In game terms, this means that the player will see three blips on their paranormal radar and can choose which spot to investigate first. These parallel encounters are linked sequentially into the overall progression, meaning that the player must find all three sisters in order for the subsequent encounters (with Count Dracula) to be unlocked and the game to progress. As a narrative structure, this is a simple 3-way branching structure where the three paths join immediately. The figure below shows the map used for the launch version of the game that began on 26 October 2012 in collaboration with the Bram Stoker Festival.

As noted in the figure, each of the subsequent encounters are linked in a sequential structure. This means that after the encounters with the sisters, the experience is linear: Each player will encounter Count Dracula and Bram Stoker in the same order.

The third possible structure, parallel exclusive, was not used in Bram Stoker’s Vampires. If encounters are linked in a parallel exclusive fashion, the player sees multiple blips on their radar and chooses one to investigate. When the chosen encounter has been completed, the remaining encounters (and corresponding radar blips) from the parallel set are removed from the radar. The player who sees multiple blips on their radar is not given any cues that will help them determine whether the next encounter is a parallel inclusive or exclusive and hence does not know whether the decision to investigate will eliminate the other encounters. What I call parallel exclusive in this post is the typical branching structure used in many interactive narratives: The player makes a decision that removes other possible story avenues at that point in time. (The branches may of course join again later on.)

Falkland Palace Entrance

An example of the parallel exclusive structure is found in our game Falkland Ghost Hunt that was set in the gardens of Falkland Palace (see photo above) in Scotland. In this game we used an important character from the castle’s history, Mary, Queen of Scots as a quest-giver who was in opposition to a fictional figure, the Banshee. In our story, the Banshee had awoken the castle’s spirits, which the player could then encounter in a series of cases set in the castle grounds. In this fashion, the story about Mary and the Banshee worked as framing device in which a varying number of other stories could be set.

Historical Map of Falkland Palace

The map above is an artist’s depiction of the castle gardens. The red lines show the framing story and the yellow lines one of the framed stories called “The Grey Lady.” Grey ladies are of course classic ghost characters found in many old castles and mansions throughout the Western world. In this particular instance, we decided to compose our own story about a woman whose young daughter died in a fire in the castle. Historically, there was a fire in the castle during which the East Wing burned down (see photo) but we don’t know if any children died. In “The Grey Lady,” the mother roams the garden, distressed and shaken, searching for the daughter. When the player encounters the mother (marked 1 on the map), she begs for help to find the missing child. The next two encounters (marked 2a and 2b) use an exclusive branching structure. Both serve a similar function in the story in that each constitutes a trace of the missing child: one (2a) a shoe and the other (2b) a poppet (doll). After the initial encounter (with the grey lady) both of the parallel exclusive encounters appear on the radar. The player has to choose which to investigate and the decision precludes exploration of the other encounter. After either the shoe or the poppet has been found, the third encounter is the child itself, who is of course placed by the ruined East Wing of the castle (see photo).

Falkland Palace East Wing

The examples in this post have shown the three main narrative structures available in our game engine. As a theoretical point of interest, we note that it is possible to represent an N-ary parallel inclusive structure as an N-depth parallel exclusive structure where the first decision point has N choices, the second N-1 and so forth. The different structures have different effects on the experience. A chain of sequential encounters feel somewhat like a tour or a mystery: the player is led from one encounter to the next, following a trail of story segments. A parallel inclusive structure implements a collection game mechanic: the player has to complete all three encounters but can do so in any order. The parallel exclusive structure is intended to make the experience feel more dynamic through the introduction of apparent uncertainty: the player doesn’t know what happened with the radar blip that disappeared — whether the ghost moved or disappeared for other reasons.


Katsiaryna Naliuka, Tara Carrigy, Natasa Paterson and Mads Haahr, ‘A Narrative Architecture for Story-Driven Location-Based Mobile Games,’ Springer LNCS, Third Workshop on Story-Telling and Educational (Serious) Games (STEG’10) , Shanghai, China, 8-10 December 2010.

Tara Carrigy, Katsiaryna Naliuka, Natasa Paterson and Mads Haahr, ‘Design and Evaluation of Player Experience of a Location-Based Mobile Game,’ 6th Nordic Conference on Human-Computer Interaction (NordiCHI 2010), Reykjavik, Iceland, October 16-20, 2010, pp 92–101.