Content uploaded by Hartmut Koenitz
Author content
All content in this area was uploaded by Hartmut Koenitz on Nov 20, 2017
Content may be subject to copyright.
Final Draft, to appear in R. Nakatsu et al. (eds.),Handbook of Digital
Games and Entertainment
Interactive Storytelling Paradigms
and Representations: A Humanities-Based
Perspective
Hartmut Koenitz, Department of Entertainment and Media Studies, University of
Georgia, Athens, GA, USA, e-mail: hkoenitz@gmail.com
Contents
A Field Between Computation and Humanities 2
Points of Orientation ...................................................................................................................... 3
A History of Influence .................................................................................................................... 4
Fundamental Differences Between Humanities and AI ................................................................. 4
A Multitude of Models ................................................................................................................... 6
Descriptive and Prescriptive Models ............................................................................................. 6
Media-Specific Aspects: Bordwell ............................................................................................ 7
The Humanities Turn in AI ............................................................................................................ 8
Requirements for a Dialog ............................................................................................................. 9
The Many Roles of the AI Researcher ......................................................................................... 10
Recent Developments .................................................................................................................. 11
Opportunities ................................................................................................................................ 12
Conclusion ................................................................................................................................... 13
Recommended Reading ............................................................................................................... 13
Abstract
Interactive Storytelling is an interdisciplinary field in which the humanities meet
artificial intelligence. Collaborations between scholars rooted in the humanities
and the computer sciences like the one between Brenda Laurel and the OZ group
at Carnegie Mellon University have had a major influence on the field. At the
same time, there are indications that the relationship is often tenuous, for exam-
ple, between models of narrative in the humanities and their application in
computational research projects. This chapter investigates the relationship, notes
challenges, and identifies opportunities for an enhanced collaboration. Additional
scrutiny in understanding context and scope of narrative models in the
humanities would improve access for AI researchers to the vast space of
1
2 H. Koenitz
available models and allow for the codification and re-use of adaptation strate-
gies. Simultaneously, the work of many AI researchers could be recast and
recognized as contributions to narrative theory. In this regard, film theory can
serve as a potential model for a narrative theory of Interactive Storytelling.
Keywords
Bordwell’s view of film • Descriptive and prescriptive narrative models • Inter-
active Storytelling (IS) • Computation and humanities • Descriptive and
prescriptive narrative models • Humanities vs. AI • Narrative models • Narrative
models • Narratology • Prescriptive and descriptive narrative models • Interactive
digital narrative (IDN)
A Field Between Computation and Humanities
Interactive Storytelling can be described as a marriage of computation and narration
that brings together perspectives origination in Computer Science and the Humanities.
The interdisciplinary aspect is both a chance and a complication. A chance, in that the
product of two distinct fields might combine the best of both worlds, and thus enable
expressive artifacts that speak of the human condition in novel and compelling ways.
And a complication, in that terminology and general approaches differ to a degree that
a serious translation effort seems inevitable. Without that, there is a considerable risk
of miscommunication. The situation is complicated further by semantic differences as
the meaning of the same term varies across fields. For example, the understanding of
“model” in computer science might diverge considerably from the understanding of
the same word in the humanities. Conversely, “story” in a more general sense
meaning “a narrative” clashes with the basic structuralist distinction in narratology
between “story,” meaning the “what” of a narrative, its content, in contrast to
“discourse,” denoting the “how,” the presentation, its “telling.” This example
illustrates the potential for miscommunication on a basic level. Even though
poststructuralist and postclassical narratology have refocused on other aspects of
narrative, the story/discourse distinction is still a cornerstone in academic training in
the humanities and no contemporary humanities scholar would use these terms
without reference to their specific meaning. When an AI researcher talks about
“interactive storytelling,” or just “story” without providing a specific definition or
acknowledging the etymology of these terms, a scholar grounded in the humanities
might wonder whether the colleague considers the narratological context and in
particular the distinction between story and its telling.
For this reason, I avoid the term “storytelling” in my own work and instead
use the abbreviation IDN for interactive digital narrative (Koenitz 2015). The
abbreviation is also intended to reduce conceptual problems inherent in the
compound “interactive narrative.” When we create the conceptual framing of
Interactive Storytelling Paradigms and Representations: A Humanities-Based... 3
narrative + interactivity, of interactivity as an addition to traditional narrative forms,
we are prone to misunderstand the challenges and opportunities of interactive
narrative as a specific form of narrative expression, in which interactivity is a
fundamental ingredient, not an additional feature.
The editors of this section prevent these terminological problems by clearly
defining their usage of terminology and distinguish between Interactive Storytelling
for the Research field, Interactive Narrative for the individual “product,” and Inter-
active Storytelling System for the technology. In this chapter, I will adhere to this
nomenclature.
Mapping the opportunities afforded by the humanities for computer science
requires a careful consideration of the etymology and semantics of both fields, but
foremost an effort in a clear descriptions of concepts. I ask the reader to remember
this aspect if some of the descriptions in this chapter might appear verbose.
In this chapter, I will start with some general remarks about the relationship
between humanities and computer science before tracing some of the
developments between humanities and AI in Interactive Storytelling. On this basis,
I propose strategies for future research bringing humanities concepts to AI.
My goal with this chapter is threefold – first, to draw attention to the richness of
available perspectives in the humanities, particularly outside of the confines of
literary studies, which can serve as models for use in Interactive Storytelling.
Second, to reframe the adaptation of humanities models by AI researchers in this
field as genuine theoretical contributions and thus reconsider the relationship to the
original models and their historical and epistemological context. Third, to raise
awareness of the mutual benefits an enhanced collaboration between humanities
and AI would bring to both fields.
Points of Orientation
Marc Cavazza and David Pizzi (2006) provide a good overview of narratology
from the perspective of AI by considering a range of popular models of narrative,
including Aristotle, Propp, and Barthes. Another valuable overview is in a paper
by Pablo Gervas et al. (2006), originating in an interdisciplinary collaboration
between humanities and AI scholars focused on narrative generation.
Of particular value as a basic orientation are summaries of the diverse narrato-
logical approaches in the humanities that also consider the historical development.
Many good resources exist in this regard, a particularly easily accessible one is the
Living Handbook ofNarratology (H€uhn n.d.). In addition to gaining an understand-
ing of different approaches, it is also important to consider their underlying assump-
tions. Dominique Robert and Shaul Shenhav (2014) provide a current overview of
this aspect of narrative analysis.
I like to suggest that having this kind of context is crucial in finding appropriate
models, but also in avoiding fundamental assumptions potentially detrimental to an
AI narrative project. The latter point pertains to the web of influences not readily
4 H. Koenitz
apparent on the surface, but equally important. For example, a narrative model might
embed the structuralist story/discourse dichotomy without explicit consideration of
the structuralist assumptions governing this distinction.
A History of Influence
There is a long history of influence between the two fields, starting with one of the
earliest scholarly works on Interactive Storytelling, Brenda Laurel's 1986 doctoral
dissertation in theatre studies at Ohio State University, "Toward the design of a
computer-based interactive fantasy system" (Laurel 1986). Her thesis caught the
interest of the computer scientist Joseph Bates, who made her work a foundational
element of his research in believable characters. Bates founded the OZ group at
Carnegie Mellon University, which accumulated a number of researchers interested in
the same subject, to work on "technology and art to help artists create high quality
interactive drama, based in part on AI technologies" (description on the OZ project
website). The group's particular focus was on building believable agents. This
connection between OZ and Laurel's work was not merely conceptual, she actually
worked with the group, as reported in her book Computers as Theatre (Laurel 1991).
Michael Mateas, a member of the group, went on to create the first fully realized
interactive drama at Georgia Tech in 2005. Façade (Mateas and Stern 2005b) puts
the player/interactor into the position of a friend trying to save the marriage
between the two virtual characters Trip and Grace. Lauded as a milestone, the
work is an exemplary result of the combination of concepts originally developed in
the humanities and then realized with the capabilities of artificial intelligence in
computer science.
Fundamental Differences Between Humanities and AI
While the trajectory from Laurel's PhD thesis to Façade is certainly a compelling
success story, a closer look also reveals some underlying issues. A basic concern
shared between the two fields is for models; however, the concrete approaches differ
considerably. Broadly, they can be described as "concrete" versus "abstract" —while
AI models tend to be formal descriptions of concrete technical implementations, most
narratological frameworks in the humanities are abstract, aiming to identify the essence
of narrative, rather than any particular case. Seymour Chatman describes the approach
as follows:
[...] literary theory is the study of the nature of literature. It is not concerned with the
evaluation or description of any particular literary work for its own sake. It is not literary
criticism but the study of the givens of criticism, the nature of literary objects and their
parts (Chatman 1980)
Interactive Storytelling Paradigms and Representations: A Humanities-Based... 5
This perspective, rejecting the evaluation of specific artifacts, contrasts with basic
scientific practice that aims to generalize from the particular case. Where Newton
starts with the observation of a falling apple (a particular event) to explain the
mechanics of the universe, the narratologist Wladimir Propp starts with a collection
of Russian folk tales to identify shared characteristics. This comparison also embeds
another important difference in the status of the resulting categories and overall
concepts. Whereas scientific definitions conceptually originate in empirical obser-
vation, definitions in the humanities are deduced, created by the researchers, as
Chatman reminds us: "literary theory [...] should assume that definitions are to be
made, not discovered" (Chatman 1980).
Such tendencies —different starting points of the academic inquiry, as well as the
divergent nature and origin of rules and categories —are deeply engrained in the two
disciplines and therefore enter, sometimes unconscientiously, scholarly work.
Another distinction pertains to the scope of a given framework. The humanities
scholar works from the premise that the sought-after model needs to accommodate
a wide range of already existing narratives, while AI researchers evaluate models
based on their concrete output. Chatman makes quite clear that the detailed critical
description of the individual works in the canon used to derive a narratological
model from is a task outside the realm of narratology. He sees this task as best left
to the field of literary criticism. In contrast, the AI researcher needs to scrutinize
any particular output to judge the success of a given system.
In a similar vein to Chatman, Gerald Prince states unequivocally that aesthetic
evaluations have no place in narratology:
For narratology, there are no great, or beautiful, or profound narratives (there are only
well-formed ones). (Prince 1982)
The perspectives of Chatman and Prince demonstrate the fundamental scope of
narrative models in the humanities. The aim for universal descriptions has conse-
quences if directly translated into AI models. An AI formalism built on
narratology might create well-formed narratives. However, these narratives might
still be boring, haphazard, and trivial. An additional layer of "aesthetic evaluation"
is necessary to produce interesting and aesthetically pleasing pieces.
It is for this reason narratological frameworks that cover semantic and
pragmatic aspects seem to be a more promising avenue for Interactive Storytelling.
Prince (1982) does indeed propose such frameworks, but cautions that semantic
and pragmatic interpretations should take care to remain generic, and thus be
applicable to a wide range of works. Therefore, narratological frameworks are
most likely insufficient for the creation of successful Interactive Narratives
because of the limited coverage of necessary aspects.
In addition, most models in the humanities do not contain formalisms in a form that
could be applied directly to interactive narrative. Gervás et al. in their discussion of
this problem mention the lack of "logic language (e.g., predicate calculus) or other
structured representation, including tables, graphs, etc." (Gervás et al. 2006).
6 H. Koenitz
A Multitude of Models
A further obstacle to the direct adoption of narrative models from the humanities in AI
is their sheer number, which make the respective advantages and disadvantages for
application in interactive storytelling difficult to judge, due to widely variable
approaches. The identification of basic elements, structural relationships, and termi-
nology vary considerably between models origination in schools of thought as far apart
as Marxism and Psychology. Casting the net even wider to include East Asian or
African traditions yields cyclical multi-climatic structures in some African oral narra-
tive traditions or models of narrative forms without conflict in Japan (Kishotenketsu).
What is commonly referred to as narratology is but one of many available
analytical perspectives in the humanities. In other words, narratology, as a field
understood to contain scholars such as Propp (1928), Barthes (1975), Chatman
(1980), Prince (1982; 2003), or Herman (2002), is a specific subfield of the different
ways of analyzing narratives in the humanities. In addition, semiotic (Eco 1989;
Kristeva 1980; Simpkins 2001), Marxist (Eagleton 2006, 2013; Lukács 1983),
feminist (Butler 1999; Moi 2002; Ruthven 1990), and psychological (Lacan 2003;
Schram and Steen 2001; Sugg 1992) frameworks have been used to analyze narra-
tives. AI scholars approaching the humanities might not immediately be aware of this
plethora of approaches. Frameworks outside of the narratological core provide a rich
—and so far mostly untapped —source for AI research in Interactive Storytelling.
An example for a successful application of this approach is McCoy et al.'s
"Comme il faut" engine (McCoy et al. 2009), the platform for their work Prom
Week. These researchers use dramaturgy, and not literature, as a starting point and
move beyond models directly concerned with narrative artifacts by drawing from
knowledge about the authoring process. Finally, McCoy et al. base their approach
on models from outside the humanities, namely the social sciences.
Descriptive and Prescriptive Models
Frameworks for narrative in the humanities are mostly descriptive in nature, con-
ceived to analyze artifacts, but without any intention to provide the necessary detailed
information an author would need. In contrast, prescriptive frameworks are also
meant as models for artists to create works and thus are more appropriate as models of
AI. Aristotle's Poetics can be seen as a rare example of a prescriptive framework, with
detailed perspectives on practical aspects, including the poetic meter for language
used and appropriate topics. The prescriptive aspect of the Aristotelian framework is
one reason for the lasting appeal of this conception of narrative. However in practice,
the distinction between prescriptive and descriptive is not be entirely clear-cut, as
artists might appropriated descriptive frameworks, and in this way adapted them for
use as prescriptive formalisms. Artists at times have also developed formal
descriptions of their work. Examples include the Russian film director Sergei
Eisenstein (1969) and the German playwright Bertolt Brecht (1960). Eisenstein
critically reflected his work and thus is not only a pioneering film artist
Interactive Storytelling Paradigms and Representations: A Humanities-Based... 7
but also one of the first film theorists. Brecht, developed an activist theatre in
theory and practice. His Japanese-influenced Epic Theatre provided a model for
Arturo Boal's "Theatre of the Oppressed" which in turn has been proposed as a
model for activist video games (Frasca 2001).
A further complication for the application of existing frameworks in AI lies in
their media-specific basis. Aristotle's Poetics is originally concerned with stage
drama (and the specific subcategory of tragedy), while many twentieth-century
narratologists focus on literary narrative. Consequently, these frameworks do not
automatically extend beyond the specific media they are conceived for.
Media-Specific Aspects: David Bordwell
Film theory provides a valuable lesson in this regard. The film theorist David
Bordwell considers the dominant basis of narratology in literary studies, predicated
on verbal forms of narrative as particularly problematic. For him, examples of
narrative in film, but also music or dance are testimony to the contrary, that
narrative can exist outside the confines of (mostly written) language. Bordwell thus
understands narrative as a "preverbal phenomenon" (Bordwell 2007) supported by
a framework in human cognition that enables the understanding of narratives
without requiring language. The insight that narrative is not exclusively linked to
literature has important consequences for analytical perspectives and resulting
models, Bordwell reminds us. When language is taken as the starting point,
elements and structures identified in literary studies should be available in other
forms of narrative as well. However, if other media forms are taken as having
specific affordances and providing unique opportunities, we should not expect to
successfully apply the same elements and structures:
If language sets the agenda for all narrative, then we ought to expect all media to follow
along. So in a film the analyst will look for equivalents of first-person point of view, or
something analogous to the voice of a literary narrator. But if we think that language is on
the same footing as other media, a vehicle for some but not all more fundamental narrative
capacities, then we might not expect to find exact parallels between literary devices and
filmic ones. Different media might activate distinct domains of storytelling. Perhaps, that
is, filmic point of view might be quite different from literary point of view, and there may
be no cinematic equivalent of a verbal narrator. (Bordwell 2007)
Bordwell’s view of film as a specific form of narrative also applies to narrative
forms in interactive digital media. While film narrative is setting up opportunities
for interpretation as Bordwell observes, Interactive Storytelling adds a further
dimension in providing opportunities for interaction. Seen in this light Interactive
Narratives are even further away from literary form than film, by including the
exploration of virtual worlds, choices and the experience of consequences, and the
ability to replay.
On the basis of his observations, Bordwell defines film narrative as comprised
of three-dimensions: story world, plot structure, and narration, in contrast to the
8 H. Koenitz
structuralist dichotomy of fabula/syuzhet or story/discourse that is still prevalent in
literary narratology.
The Humanities Turn in AI
In summary, AI researchers will most likely encounter frameworks in the
humanities that are either descriptive in their intent or media-specific and therefore
inappropriate and incomplete as the basis for practical implementations in digital
media. This does not preclude the usage of such a framework for AI in principle,
but necessary adaptation and additions of missing elements mean that the result is
actually a different framework, with a complex relationship to the original. I like to
suggest that this fact alone does not constitute a fundamental problem and rather
should be understood as standard academic procedure. However, the tendency to
foreground the original framework is problematic as the assumptions on which it
was based are no longer valid in their entirety. The underlying problem here might
very well be an attempt to gain additional "academic prestige" for interdisciplinary
work by referencing major theoretical frameworks originating outside a researcher's
home discipline.
In particular, Aristotle via the Poetics can be considered the "undying ghost in
the machine," of the field, a frame of reference that is often taken as a given
without proper contextualization of the relationship to the ancient text. In this vein,
Cavazza and Pizzi consider the importance of Aristotle's Poetics as a model for
Interactive Storytelling as having been overstated:
[...] the Aristotelian model seems to have been primarily used as an inspiration, a theoretical
framework in which to describe narrative concepts, rather than a source of narrative
formalisms, let alone their implementations in IS [interactive storytelling] systems. (Cavazza
and Pizzi 2006)
Further investigating the relationship between models and implementations, the
authors classify the connection between the Poetics and many concrete
implementations as tenuous, quoting Wilks's (1992) insight that "systems do not
always work by means of the formalisms that decorate them."
Maybe one aspect that AI researchers could take from the humanities is that existing
frameworks are not sacrosanct and narratological concepts are constantly evolving.
Later additions are valuable and welcome. For example, instead of emphasizing the
connection to ancient drama, Mateas' and Stern's Façade can be more aptly described
as based on the theoretical framework developed by Mateas himself (2001), modifying
Laurel's (1991) reinterpretation of Neo-Aristotelian concepts by implementing Janet
Murray's (1997) analytical and phenomenological categories. The reference to the
rather distant —several times removed —original notion is more distracting than
helpful. Indeed, Aristotle's unmodified concept is inappropriate for Interactive
Storytelling in its basic demand of a plot that is perfect if nothing needs to be added or
can be removed, a perspective which is in fundamental disagreement
Interactive Storytelling Paradigms and Representations: A Humanities-Based... 9
with the idea of interactive narrative that needs to offer excess material for
alternative paths and optional elements.
Seen in this light, Mateas' and Stern's claim of an "Aristotelian tension arc"
(Mateas and Stern 2005a) in Façade might be misleading. Given the lineage outlined
earlier, an unwieldy but more correct description would be of a "Mateas/Stern tension
arc based on Laurel's reconceptualization of Neo-Aristotelian concepts modified to
accommodate Murray's media-specific categories." This perspective does in no way
lessen the appreciation of the work of the involved scholars. The crucial insight from
this discussion is the realization that many — if not all — AI researchers working in
Interactive Storytelling are no longer just that. Instead, by adapting and further
developing existing theoretical frameworks for their needs, they engage in activities
typical of humanities scholars. We might even say that AI researchers working in
Interactive Storytelling — at least partially —become humanities scholars
themselves, for pragmatic reasons, since few, if any, narratological models can be
implemented as is. Yet, the effort that goes into making the necessary adjustments is
more than just "technical preparation work" and has so far not been fully recognized.
Indeed, when Szilas and Richle (2013) pinpoint the fuzziness of the concept of
dramatic tension in narrative theory or Gervás et al. (2015) identify commonalities
between the different categorizations of plot, they also work in a humanities capacity.
Existing perspectives that emphasize common ground between AI and the humanities,
for example, Gervás et al.'s notion of "shared abstract models" (Gervás et al. 2006)
are therefore only a starting point. It is necessary to recognize the contribution of AI
researchers to narrative theory.
At the same time, AI researchers are encouraged to more fully document their
process in adapting and augmenting frameworks originating in the humanities.
Once the assumptions underlying this translation process are available for research
as well as the strategies in conducting the translation, they could be consciously re-
applied, modified, and critiqued, and thus help end a practice of uncritical re-
application, described by Gervás et al.:
Often, the most significant reason behind a particular choice of theory is that AI researchers
find it easier to work with models that a previous researcher has already translated to AI
jargon and applied in some previous computer program. (Gervás et al. 2006)
An important aspect of the availability of such "translation rules" would be as
tools for the adaptation of additional frameworks, by making the plethora of
models in the humanities more accessible.
Requirements for a Dialog
AI scholars working in interactive narrative can only gain from an enhanced dialog
with the humanities. Those who ignore the context of humanities frameworks do so
at their own peril —instead of gaining additional insight by means of an interdisci-
plinary dialog that could help with the concrete implementation. Yet, such a dialog
10 H. Koenitz
requires at least a minimal acknowledgement of narratological models and
vocabulary.
From this perspective, McCoy et al.'s otherwise outstanding work on Prom Week
is marred by the lack of a clear definition of "story." For example, in one of their
papers, (McCoy et al. 2013) the authors variously refer to story as "personally
meaningful" game experience, as "campaign," as "gameplay," as a "collection of
levels [...] where the player can take social actions," or they talk about "player
agency at the story level" and the "player's path through the story." Each of these
descriptions depicts "story" as a slightly different concept. In a related perspective,
Markku Eskelinen fundamentally questions the value of contributions missing a
concrete definition of narrative and deems them "non-academic" (Eskelinen 2012)
While I consider Eskelinens's critic exaggerated, the lack of a definition is still a
serious omission. In any case, it represents an obstacle to interdisciplinary dialog.
How a scholar working in the space of Interactive Storytelling can integrate a
meaningful discussion of relevant humanities terminology and definitions is exem-
plified in Michael Nitsche's Video Game Spaces (Nitsche 2008). This author
dedicates considerable space in his volume to the discussion of narratological
concepts and their adaptation for interactive media.
Bordwell's approach towards film, outlined previously in this chapter, can serve as
an example for AI researchers interested in developing models for interactive
narrative. While mindful of the existing tradition in literary studies, his model of film
narrative reflects the specific characteristics of the medium. Similar, Interactive
Storytelling researchers might be best served by developing specific frameworks for
narrative in interactive digital media that are mindful of the existing narratolog-ical
tradition. Such an effort would have the added benefit of providing a common
ground for discussion with scholars in the humanities.
The Many Roles of the AI Researcher
Earlier, I have pointed out some of the fundamental differences between the two
fields. This discussion also embeds a perspective on the role of the AI researcher in
interactive narrative in contrast to the humanities scholar. Interactive Storytelling
research encompasses aspects of both narrative analysis and literary/aesthetic criti-
cism. And while it is true that many leading figures in the humanities have also filled
both roles —for example Roland Barthes — AI can still profit from an awareness of
the differentiation, and its different dimensions. However, this perspective does not
yet cover all the roles filled by many AI researchers. As I have argued earlier, there is
also a creative dimension (Koenitz 2014). In many cases, the researcher also becomes
the author using her/his system to create an artistic work. Thus, AI work in Interactive
Storytelling combines aspects of several different roles from a humanities perspective,
including narrative theorist, literary critic, and creative author.
In fact, given the technical nature of AI work, the creative role of the AI
Interactive Storytelling researcher can be further broken down in creative and
technical aspects, thus opening yet another dimension (Table 1). As the technical
Interactive Storytelling Paradigms and Representations: A Humanities-Based... 11
Table 1 Roles of typical AI researcher as seen from the humanities
Roles in the humanities
Dimension
Narratologist
Structural
Literary critic
Aesthetic
Creative author
Creative
Printer
Technical
Roles in AI Interactive Storytelling researcher
programmer, the AI researcher is also the equivalent of camera technician doubling
as movie director. In a literary setting, this would equate to a consideration of print
technology and the materiality of books. While edge cases do exist in regard to the
treatment of artists' books (however these are often seen as an entirely separate form
from the perspective of fine arts (Drucker 2004)), such aspects are not normally
considered in literary studies. Technical aspects are more likely to be covered in
areas such as film studies, where they translate into particular aesthetic presentations
and resulting effects on the audience. For example, the film scholar Robert L
Carringer foregrounds technical aspects —such as wide angle lenses, specific light-
ning techniques, and camera movements — in his volume on the making of Citizen
Kane (Carringer 1996).
In this case, the humanities can help the AI researcher to better understand the
many different dimensions of work in this field. These can be used as research
directions, or as the basis of defined roles in a research team.
Recent Developments
Since about 2005, AI researchers have increasingly investigated and employed
models outside the literary domain and moved away from the focus on narrative
structures. One such effort attempts to identify structures in improvisational theatre
for use in Interactive Storytelling (e.g., Fuller and Magerko 2010). The goal here is
to model the cognitive understanding shared between performers and apply the
results in a system with virtual actors. In a similar vein, research into emergent
narrative (Louchart and Aylett 2004) centers on virtual autonomous actors. Other
efforts focus on modeling the author/director using AI planning methods (Porteous
and Cavazza 2009; Riedl and Young 2006; Young et al. 2004). Another approach
has instead focused on particular aspects of narrative, for example, the already men-
tioned research by Szilas and Richle on dramatic tension (Szilas and Richle 2013) or
O'Neil and Riedl's work on modeling suspense (O'Neill and Riedl 2014).
As already mentioned, Michael Mateas, the creator of Façade, has focused on the
notion of "social games" in the more recent work Prom Week (McCoy et al.2011).
The model of social situations used is derived from research in the social sciences.
A trend in recent years is the development of specific computational models of
narrative. Researchers in this vein seem to have decided that the differences in
scope and methodology make the adaptation of traditional narratological models
impractical. Their efforts amount to the creation of a specific field of computational
narratology, often based on cognitive science's understanding of narrative —here
they intersect with the humanities again in the form of narratologists like David
12 H. Koenitz
Herman, who sees narrative as a basic cognitive frame and understands cognitive
sciences and narratology as two different perspectives on the same issue. Examples
for work in this manner are, for example, George Lakoff and Srini Narayanan who
propose a "Computational Model of Narrative" (Lakoff and Narayanan 2010)
based on a shared "cognitive structure of human motivation, actions, goals, and
events." Another example is Chris Martens et al. (2014) who explore a concept of
"Proofs as Stories" based on ideas by Bosser et al. (2010) in an application of
linear logic for narrative, while Tory Anderson (2015) explores the implications of
the concept of Episodic Memory.
Opportunities
Finally, I like to offer suggestions for future work exploring humanities models
and concepts for AI approaches to Interactive Storytelling. I have already
mentioned the idea of documenting the process of adapting humanities models to
AI implementation, with the hope to identify "translation rules."
Janet Murray (2015) reminds us that the humanities can also provide partial
models —narrative concepts of particular dramatic situations, genre-specific lan-
guage, or character pairings. Accordingly, she proposes a strategy to abstract the
"women between two suitors" situation found in literature as a formalism for AI
characters.
In addition, there are also many complete narrative models that present opportu-
nities for AI treatment. For example, French neoclassical drama of Moliere and
Corneille appears to be a particularly good candidate with its strict set of rules. On
the other end of the spectrum, the theatre of the absurd (Beckett, Ionesco, Pinter) is
testimony of successful narrative communication that purposefully undermines
causal and temporal relations and conventional narrative development in order to
expose the futility and absurdity of everyday life. Implementing the latter would
mean a considerable shift from the causality-based models that have typically been
used so far, but also from the goal of completeness; Clara Fernández-Vara's (2014)
implementation of narrative dream puzzles is an important pointer in this direction.
Similarly on the trajectory of causal breaks, but well-specified through the detailed
writings of its creator, Epic Theatre represents a particular opportunity for AI. In the
1930s and 1940s, the German playwright Bertolt Brecht developed a concept of non-
Aristotelian drama (Brecht 1960). Influenced by the Japanese No theatre, Brecht was
motivated by Marxist calls for the education of the masses. His form of drama was
designed to make the audience reflect the depicted circumstances and ultimately turn
to political action. With the experience of pity and fear and the resulting catharsis,
Aristotelian drama arguably reinforces the status quo by demonstrating the futility of
human effort in the light of fate. Brecht's concept wants the opposite, an audience
educated about the need for action. And while Aristotle demands a complete action, a
plot that needs nothing and cannot have anything taken away, Brecht's plot is episodic
and open to continuation —ideally in the form of political action outside the theatre.
Characters also have a different function in the
Interactive Storytelling Paradigms and Representations: A Humanities-Based... 13
epic theatre. Instead of being fully immersed in their role, they are supposed to
"stand beside themselves", and comment on the character roles. Brecht likens this
new kind of acting style to a street scene in which a motorist explains an accident
to bystanders or the police. While she might act out certain parts of her report, she
would also comment on them: "This is how I looked in the mirror and then this car
came out of the side street..."
Several aspects make epic theatre an interesting model for Interactive Storytell-
ing. The removal of the requirement for stringent causal connections between
events is already helpful. However, a major change is in the treatment of characters.
It is no longer the aim to be "in the role" to create a believable likeness, but instead
to comment on the circumstances, both in and outside the role. As the creation of
believable characters has been a stable of AI work in Interactive Storytelling, this
change in perspective provides an interesting avenue for future work. While the
emphasis on commentary would put higher demands on the character's understand-
ing of the world, the effort for maintaining believability would be considerably
reduced.
Conclusion
AI researchers in the field of Interactive Storytelling still have much to gain from
the humanities - an enhanced understanding of the subject matter of narrative,
models outside of literary studies, applications of additional fields such as
psychology to narrative, but also considerations of authorship and knowledge in
performance studies applicable to the understanding of interactive processes.
However, the potential gains will be even greater when the relationship is defined as
bidirectional. After all, both sides often work on the same key problems, as Gérvas
et al. already pointed out in their pioneering collaboration of AI and narratology
experts (Gervás et al. 2006). In this regard, the work of many AI researchers on
aspects of narrative deserves proper recognition that can be framed as contributions
towards a field of a media-specific narratology, "Interactive Storytelling Studies" in
line with "Film Studies."
Recommended Reading
T.S. Anderson, From episodic memory to narrative in a cognitive architecture, in 6th Workshop
on Computational Models of Narrative (CMN’15), eds. by M.A. Finlayson, B. Miller, A.
Lieto, R. Ronfard. Atlanta, GA, USA, 2015 pp. 2-11. doi.10.4230/OASIcs.CMN.2015.2
R. Barthes, S/Z. (trans: Miller R). (Hill and Wang, New York, 1975)
D. Bordwell, Poetics of Cinema (Routledge, New York, 2007)
A.-G. Bosser, M. Cavazza, R. Champagnat, Linear logic for non-linear storytelling, in
Proceedings of the Nineteenth European Conference on Artificial Intelligence ECAI, vol. 215,
eds. by H. Coelho, R. Studer, M. Wooldridge. (ECAI, Lisbon, Portugal, 2010).
doi.10.3233/978-1-60750-606-5-713, pp. 713-718
B. Brecht, Kleines Organon f€ur das Theater (Suhrkamp, Frankfurt, 1960)
14 H. Koenitz
J. Butler, Gender Trouble (Routledge, London/New York, 1999)
R.L. Carringer, The Making of Citizen Kane (University of California Press, Oakland, 1996)
M. Cavazza, D. Pizzi, Narratology for interactive storytelling: a critical introduction, in Technolo-
gies for Interactive Digital Storytelling and Entertainment (Springer, Berlin/Heidelberg, 2006),
pp. 72–83. doi:10.1007/11944577_7
S.B. Chatman, Story and Discourse: Narrative Structure in Fiction and Film (Cornell University
Press, Ithaca/London, 1980)
J. Drucker, The Century of Artists’ Books (Granary Books, New York, 2004)
T. Eagleton, Criticism and Ideology: A Study in Marxist Literary Theory (Verso, London/New York,
2006)
T. Eagleton, Marxism and Literary Criticism (Routledge, London, 2013)
U. Eco, The Open Work. (trans: Cancogni A). (Harvard University Press, Cambridge, MA, 1989).
M. Eskelinen, Cybertext Poetics (Bloomsbury Publishing USA, Indian Trail, New York,
2012) S. Eisenstein (1969). Film Form: Essays in Film Theory. (J. Laeyda, Trans.). San
Diego, CA: Harcourt Brace
G. Frasca, Videogames of the Oppressed. (2001)
D. Fuller, B. Magerko, Shared mental models in improvisational performance. Presented at the
The Intelligent Narrative Technologies III Workshop, (ACM, New York, 2010). doi.10.1145/
1822309.1822324
P. Gervás, B. Lönneker-Rodman, J.C. Meister, Narrative models: narratology meets artificial
intelligence. Presented at the International Conference on Language Resources and
Evaluation, Genoa, Italy (2006)
P. Gervas, C. Leon, G. Méndez, Schemas for narrative generation mined from existing
descriptions of plot, in 6th Workshop on Computational Models of Narrative (CMN’15), ed. by
M.A. Finlayson, B. Miller, A. Lieto, R. Ronfard (Dagstuhl Publishing, Atlanta, GA, USA,
2015), pp. 54–70. doi:10.4230/OASIcs.CMN'15.54
D. Herman, Story Logic (U of Nebraska Press, Lincoln, 2002)
P. H€uhn et al., (ed.), The Living Handbook ofNarratology. (n.d.). Retrieved 5 Oct 2015, from
http:// www.lhn.uni-hamburg.de
H. Koenitz, Five theses for interactive digital narrative, in Interactive Storytelling: 7th International
Conference on Interactive Digital Storytelling, ICIDS 2014, Singapore, Singapore, November 3–
6, 2014, Proceedings, vol. 8832, eds. by C. Fernández-Vara, A. Mitchell, D. Thue. (Springer
International Publishing, Cham, 2014). doi.10.1007/978-3-319-12337-0_13, pp. 134–139
H. Koenitz, Towards a specific theory of interactive digital narrative, in Interactive Digital
Narrative, ed. by H. Koenitz, G. Ferri, M. Haahr, D. Sezen, T.I. Sezen (Routledge, New York,
2015), pp. 91–105
J. Kristeva, Desire in Language (Columbia University Press, New York, 1980)
J. Lacan, Ecrits (W. W. Norton & Company, New York, 2003)
G. Lakoff, S. Narayanan, Toward a computational model of narrative, in Computational Models
of Narrative Papers from the AAAI Fall Symposium. Palo Alto:... Symposium: Computational
Models of Narrative. (2010)
B. Laurel, Toward the Design of a Computer-Based Interactive Fantasy System (Ohio State
University, Columbus, 1986)
B. Laurel, Computers as Theatre (Addison-Wesley, Boston, 1991)
S. Louchart, R. Aylett, Narrative theory and emergent interactive narrative. Int. J. Cont. Eng. Edu.
Life Long Learn. 14(6), 506 (2004)
G. Lukács, The Historical Novel. 1937 (University of Nebraska Press, Lincoln, 1983)
C. Martens, J.F. Ferreira, A.G. Bosser, M. Cavazza, Generative story worlds as linear logic
programs, in Intelligent Narrative Technologies 7: Papers From the Workshop. (Intelligent
Narrative Technologies 7, Palo Alto, 2014)
M. Mateas, A preliminary poetics for interactive drama and games. Digit. Creat. 12(3), 140–152
(2001). doi:10.1076/digc.12.3.140.3224
Interactive Storytelling Paradigms and Representations: A Humanities-Based... 15
M. Mateas, A. Stern, Procedural authorship: a case-study of the interactive drama Façade.
Presented at the Digital Arts and Culture, (2005a)
M. Mateas, A. Stern, Structuring content in the Façade interactive drama architecture. Presented
at the AIIDE, (2005b)
J. McCoy, M. Mateas, N. Wardrip-Fruin, Comme il Faut: a system for simulating social games
between autonomous characters, in Proceedings of the Digital Arts and Culture Conference
DAC. Digital Arts and Culture 2009 (Irvine, CA, USA, 2009)
J. McCoy, M. Treanor, B. Samuel, M. Mateas, N. Wardrip-Fruin, Prom week: social physics as
gameplay, in Foundations of Digital Games (ACM, Bordeaux, 2011). doi.10.1145/
2159365.2159425, pp. 319-321
J. McCoy, M. Treanor, B. Samuel, A.A. Reed, M. Mateas, Prom week: designing past the game/
story dilemma. in Foundations ofDigital Games 2013. (2013)
T. Moi, Sexual/Textual Politics: Feminist Literary Theory (Routledge, London/New York, 2002)
J.H. Murray, Hamlet on the Holodeck: The Future of Narrative in Cyberspace (Free Press, New
York, 1997)
J.H. Murray, A tale of two boyfriends: a literary abstraction strategy for creating meaningful
character variation, in Interactive Digital Narrative, ed. by H. Koenitz, G. Ferri, M. Haahr, D.
Sezen, T.I. Sezen (Routledge, New York, 2015)
M. Nitsche, Video Game Spaces (MIT Press, Cambridge, MA, 2008)
B. O'Neill, M. Riedl, Dramatis: A computational model of suspense. Presented at the 28th AAAI
Conference on Artificial Intelligence, Quebec City, 2014
J. Porteous, M. Cavazza, Controlling narrative generation with planning trajectories: the role of
constraints, in Interactive Storytelling (Springer, Berlin/Heidelberg, 2009), pp. 234-245.
doi:10.1007/978-3-642-10643-9_28
G. Prince, Narrative analysis and narratology. N. Lit. Hist., 13(2), 179-188. (1982). Retrieved from
http://www.jstor.org/stable/468908
G. Prince, A Dictionary of Narratology (U of Nebraska Press, Lincoln, 2003)
V. Propp, Morphology of the Folktale (University of Texas Press, Austin, 1928)
M.O. Riedl, R.M. Young, From linear story generation to branching story graphs. IEEE Comput.
Graph. Appl. 26(3), 23-31 (2006). doi:10.1109/MCG.2006.56
D. Robert, S. Shenhav, Fundamental assumptions in narrative analysis: mapping the field. Qual.
Rep. 19, 1-17 (2014)
K.K. Ruthven, Feminist Literary Studies (Cambridge University Press, Cambridge, 1990)
D.H. Schram, G. Steen, The Psychology and Sociology ofLiterature (John Benjamins Publishing,
Amsterdam, 2001)
S. Simpkins, Literary Semiotics (Lexington Books, Lanham, 2001)
R.P. Sugg, Jungian Literary Criticism (Northwestern University Press, Evanston, 1992)
N. Szilas, U.M. Richle, Towards a computational model of dramatic tension. in Workshop on
Computational Models of Narrative 2013. (Hamburg, 2013). doi.10.4230/OASIcs.
CMN.2013.257
C.F. Vara, Creating Dreamlike Game Worlds Through Procedural Content Generation, in
Intelligent Narrative Technologies 7: Papers from the Workshop. Association for the
Advancement of Artificial Intelligence, (2014)
Y. Wilks, Form and content in semantics, in Computational Linguistics and Formal Semantics,
ed. by R. Johnson, M. Rosner (Cambridge University Press, Cambridge, 1992)
R.M. Young, M.O. Riedl, M. Branly, A. Jhala, R.J. Martin, C.J. Saretto, An architecture for
integrating plan-based behavior generation with interactive game environments. J. Game. Dev.
1(1), 51-70 (2004)
