# Curating Simulated Storyworlds — Chapter 2: Defining Emergent Narrative

> NOTE: THIS IS A MARKDOWN CONVERSION INTENDED FOR LLM CONSUMPTION.
> IF YOU ARE A HUMAN BEING, READ THE PDF INSTEAD:
> https://jamesryan.computer/papers/2018_ryan_curating_simulated_storyworlds.pdf
>
> The PDF is the canonical document, typeset with every figure and table. This file
> is a plain-text conversion of the LaTeX source, published HERE so that software agents
> can read the thesis without parsing a PDF. It is lossy and is not meant for humans.
>
> This file is chapter 2 of 14 only: Defining Emergent Narrative (~6,900 words).
> See below for links to markdown conversions of the individual chapters.
>
> **Structure.** Sections are `##`/`###`/`####` headings. Cross-references are resolved to
> their numbers ("Section 3.1.1", "Chapter 5"), so a heading can be searched for directly.
> Citations are inline bracketed keys such as `[ryan2017grimes]`; every key used in the text
> has an entry in the bibliography, formatted as `- **[key]** Author. Title. Venue, Year.`
> Footnotes appear inline as `[Note: ...]`.
>
> **Conversion limits.** Figures and most complex tables appear as `[Figure omitted]` or
> `[Table omitted]`. Code fences carry no language tag. A few diacritics in author names
> were simplified from the Latin-1 source.
>
> **All files.**
>
> Index (this list, plus the abstract; small): https://jamesryan.computer/thesis/index.md
> Whole thesis in one file (large, ~1.6 MB): https://jamesryan.computer/thesis/full.md
>
> By chapter:
> - 01. Introduction (~4,600 words, ~30 KB) — https://jamesryan.computer/thesis/01-introduction.md
> - 02. Defining Emergent Narrative (~6,900 words, ~47 KB) — https://jamesryan.computer/thesis/02-defining-emergent-narrative.md
> - 03. The Pleasure of Emergent Narrative (~21,000 words, ~135 KB) — https://jamesryan.computer/thesis/03-the-pleasure-of-emergent-narrative.md
> - 04. The Pain of Emergent Narrative (~41,000 words, ~268 KB) — https://jamesryan.computer/thesis/04-the-pain-of-emergent-narrative.md
> - 05. Refining Emergent Narrative (~11,000 words, ~73 KB) — https://jamesryan.computer/thesis/05-refining-emergent-narrative.md
> - 06. James Ryan Generator (~2,800 words, ~16 KB) — https://jamesryan.computer/thesis/06-james-ryan-generator.md
> - 07. World (~20,000 words, ~129 KB) — https://jamesryan.computer/thesis/07-world.md
> - 08. Case Study: Diol/Diel/Dial (~9,300 words, ~59 KB) — https://jamesryan.computer/thesis/08-case-study-diol-diel-dial.md
> - 09. Talk of the Town (~25,000 words, ~162 KB) — https://jamesryan.computer/thesis/09-talk-of-the-town.md
> - 10. Case Study: Bad News (~22,000 words, ~132 KB) — https://jamesryan.computer/thesis/10-case-study-bad-news.md
> - 11. Hennepin (~32,000 words, ~209 KB) — https://jamesryan.computer/thesis/11-hennepin.md
> - 12. Looking Ahead: Sheldon County (~11,000 words, ~75 KB) — https://jamesryan.computer/thesis/12-looking-ahead-sheldon-county.md
> - 13. Conclusion (~2,700 words, ~17 KB) — https://jamesryan.computer/thesis/13-conclusion.md
> - 14. Bibliography (~29,000 words, ~233 KB) — https://jamesryan.computer/thesis/14-bibliography.md

This dissertation is about emergent narrative: its pleasure, its pain, a framework for harnessing the former while simultaneously rectifying the latter, and a retrospective of projects that demonstrate that framework. Since it is, as such, our center of gravity, it would seem prudent to first explicitly define the term ‘emergent narrative' (as I treat it in this document). In this chapter, I develop a definition for the concept, explore the origins of the term ‘emergent narrative', and then conclude with a brief overview of academic work in this area. In the latter section, I give particular emphasis to the academic work of Ruth Aylett, Sandy Louchart, Mariet Theune, Ivo Swartjes, and their respective collaborators, since among scholars these research programs are particularly associated with the phrase ‘emergent narrative'.

## A Definition

At the level of intuition, ‘emergent narrative' means narrative that emerges out of the interaction of processes in a system. In some sense, however, all narrative satisfies this definition. For example, if we were together in person and I were tell you a story, the string of sounds constituting the telling would emerge from the physical articulatory system that enables me to produce speech. At a higher level, the structure of that story would emerge from an even more complex system: my brain, or perhaps a larger system in which my brain is embedded, such as my body or my being or the world or the universe, or the alpha and the omega and the interstices, et cetera, and so forth. [Note: Richard Walsh has essentially made this argument in describing narrative creation as a complex, dynamical process out of which artifacts emerge [walsh2007rhetoric, pp. 130–147]. Others have produced similar arguments with regard to reader cognition—due to the intersubjective nature of storytelling, a story *as mentally processed by a given interlocutor* may be viewed as a kind of idiosyncratic emergent narrative [bakhtin1981dialogic,duranti1986audience,goodwin1984notes]. To quote Elinor Ochs and Lisa Capps: “Narratives are tales that tellers and listeners map onto tellings of personal experience. In this sense, even the most silent of listeners is an author of an emergent narrative” [ochs1996narrating, p. 21].] By such an inclusive definition, all narrative might be considered emergent, but clearly this is too broad for the purposes of this investigation.

#### The Importance of Computation

Let us say, then, that in this document ‘emergent narrative' refers to narrative that emerges out of the interaction of processes in a *digital* system. Still, this is too inclusive: when a work of narrative cinema is played on a DVD player, the sequence of frames that compose it are in a sense emerging out of the digital system constituted on the player, but that is not what I mean by ‘emergent narrative'. How about we swap out ‘digital' for ‘computational'? This is better, but even DVD players use computation, and movies that use computer-generated imagery do so through the utilization of computational systems that enable the authoring and rendering of such imagery (at production time). More problematically, all videogame narrative seems to fall under this definition—even cutscenes require some computation, let alone configurations that incorporate player interactivity.

So, what is missing so far? How is emergent narrative different than other methods for composing procedural narrative, such as the kind of hand-authored (yet still procedural) approach that is typical of narrative crafting in videogames? While an authored videogame narrative may require the interaction of computational processes to be *presented*, the very constitution of the narrative—comprising its characters, plot, the telling, and so forth—does not emerge from such a system. Emergent narrative, however, does not just require a computational system to be presented, but moreover requires a computational system to be *composed*. [Note: Game designer Marc “MAHK” LeBlanc has made a distinction between emergent narrative and what he terms ‘embedded narrative' [leblanc2000formal], which is characterized by preconstructed storylines that may be embedded in a gameplay experience (and enacted through gameplay). Later work by Katie Salen and Eric Zimmerman [salen2004rules], Henry Jenkins [jenkins2004game], and Ernest W. Adams [adams2005interactive] revisited and expanded upon this notion.] In this way, the process of composing emergent narrative is fundamentally removed from human authorship, or more precisely, it is the result of a different mode of human authorship: rather than crafting a narrative artifact, a human crafts a system out of which narrative artifacts may emerge. [Note: As committee member Ian Horswill noted in feedback on an earlier draft of this thesis, some writers have described the ideation process in fiction writing as working by a kind of mental simulation. For instance, Lajos Egri has expounded an approach to dramatic composition whereby plot emerges from the more primary matter of character behavior [egri1960art]. While the metaphor may be compelling, and the craft advice handy, this phenomenon is quite distinct from that of building a computer simulation out of which narrative may emerge. In the latter, character activity is concretized (reified in computational material) and the authoring procedure is more latent—instead of composing a story, one crafts a system that models storyworlds. In this sense, developing a character simulation works similarly to the writing method called *worldbuilding* [roine2016imaginative,pacat2017fantasy]. In Section 3.1.3, I will discuss this connection at length. Noah Wardrip-Fruin, in his own feedback on my earlier draft, asked, “Do you want your definition of emergent narrative to require a digital system, or a computational one? For example, if the processes were implemented as a board game, should that fall within the definition?” (personal communication, July 1, 2018). While it could certainly be fruitful to frame the concept of emergent narrative in terms of board games—Thijs Alofs has approached this with the *Interactive Storyteller* project [alofs2011tabletop,alofs2012interactive,alofs2015tabletop]—that would take us outside the purview of this thesis, as the next section clarifies.]

#### Requirement: A Simulated World

We are now very close to the definition employed in this document, but still we are thinking too broadly. To hone further, let us consider a system called *Say Anything* [swanson2008say,swanson2012say], which was developed by Reid Swanson as part of his doctoral work under the tutelage of Andrew S. Gordon. In *Say Anything*, a human interactor and an AI system collaborate to craft a story together by a novel interaction pattern. To start, the interactor provides the first sentence of the story to the system by typing it in free text. Next, the system processes this sentence in an attempt to find the most related sentence to it that appears in a corpus of one million narrative blog posts compiled by Gordon and Swanson [gordon2009identifying]. After isolating the most related sentence in the blog corpus, the system retrieves the sentence that immediately *succeeds* that sentence in the blog post in which it appeared. The idea here is that the successor sentence in the original blog post may serve as a good response to what the interactor provided, since a human writer decided to place the successor sentence after a sentence that is related to what the interactor provided. [Note: In artificial intelligence, this is called *case-based reasoning* [kolodner2014case].] After retrieving the successor sentence, the system adapts it to better match the prompt provided by the interactor (e.g., by replacing character names), and then appends it to the unfolding story. From here, the interactor provides another sentence, the system forms a response to that next sentence, and so forth until the story has concluded.

While narrative emerges bottom-up in *Say Anything* through the interaction of processes in a system that is partly computational, it is not an example of emergent narrative as the notion is typically conceived. This is due primarily to the absence in this system of a fundamental component of emergent-narrative systems: a *simulated storyworld*. In emergent narrative, a computer simulation models a storyworld, and narrative emerges as the byproduct of running the simulation over a series of timesteps. [Note: In his feedback on an earlier draft of this thesis, committee member Jonathan Lessard noted that I do not define or even discuss my usage of the term ‘simulation', which is curious given its centrality to the concerns herein. My lack of consideration in this regard is in line with conventional discourse in the academic area of videogames, as Jonathan noted in pointing me toward a fantastic paper by Veli-Matti Karhulahti titled “Do Videogames Simulate? Virtuality and Imitation in the Philosophy of Simulation” [karhulahti2015videogames]. Karhulahti explains that, while participants in the *ludology vs. narratology debate* [frasca2003ludologists,murray2005last,mateas2005build] took one another to task vis-à-vis the appropriation of narratological concepts in the nascent scholarly area of game studies, statements such as “All computer games are simulations” [parker2013simulation, p. 1] have been curiously uncontested. Writing about discourses on the philosophy of simulation and in simulation research—where, for instance, the terms ‘computer simulation' and ‘computerized simulation' have distinct meanings [karhulahti2015videogames, p. 3]—Karhulahti argues that the essential prerequisite for simulation inasmuch as those fields are concerned, that a reference system is imitated, is generally violated in the case of videogame “simulations”. While he concedes that a peculiar, but now conventional, sense of the term seems to have evolved in computational media, Karhulahti worries that this challenges cross-scientific communication: “I recommend videogame scholars not to pass the word simulation through duty free interdisciplinarity, but to reserve the word simulation for those already-numerous, well-established purposes that have been recognized above” [karhulahti2015videogames, p. 12]. Instead, he recommends the term ‘virtualization' (and, in turn, ‘virtual universe') for cases where no reference system is modeled. In this dissertation, and in my life, I am guilty of Karhulahti's charges. The historically conventional sense of the term—by which I mean something like ‘an artificial system that imitates a natural one'—strikes me as curiously narrow. My introduction to the phenomenon of simulation was through videogames as a small child, so it never occurred to me that a simulation should imitate some reference system. To me, the term suggests an artifact that is synthetic, processual, and has the quality of being its own enclosed world. In saying this, I am rearticulating the unexamined sense of the term that Karhulahti attributes to videogame scholars, but what I mean to express is that I personally developed this (mis)understanding through my experience in the world: in my exposure to the term, ‘simulation' has been almost exclusively associated with videogames. When the world changes, words change, and I personally do not feel beholden to an etymologic or scientific basis for the “true” meaning of ‘simulation'. While I am thoroughly impressed by Karhulahti's argument—his paper is fabulous—I will continue to employ the (arguably) degenerate sense of the term.] Typically, or perhaps universally, the storyworld features characters who to some degree act autonomously, in which case the emergent narrative may be more concretely conceived as being constituted in the accumulation of character activity over the course of simulated story time. [Note: I can imagine an emergent-narrative system that would not model characters, but it would simply be one whose stories are not about characters. Still, in order to produce narrative, the system would have to model *something* that could serve as the subject(s) of its emergent stories, and I think the term ‘character' is a close enough approximation to ‘subject', and it is certainly a more natural general term.]  If the system is made interactive—for instance, by allowing an interactor to decide how one or more of the characters will act—then the interactor is to some degree responsible for the narrative that emerges over the course of her interaction with the system. While this is also the case with *Say Anything*, such a configuration resides more firmly in the emergent-narrative tradition because the resulting narrative emerges not in a general sense, but more specifically *emerges out of a character simulation*—this is the defining quality of emergent narrative.

Critically, the degree to which a system is situated in the emergent-narrative paradigm depends on what might be referred to as the *integrity* of its character simulation. [Note: Specifically, what someone *with my artistic inclination* might refer to as ‘integrity'. Again, this document is in part a manifesto.] In the area of *narrative planning* [young1999notes,riedl2010narrative], for instance, systems commonly modulate character behavior (or other simulated phenomena) according to the activity of a plot-level agent that is removed from the storyworld in which the characters are diegetically situated. When the simulation is tampered with in this way, stories no longer emerge, but instead spawn according to the policies of the interventionist mechanism—in the wake of such tactics, a work of emergent narrative is transformed into something else entirely. [Note: Generally, simulation integrity is compromised in an effort to guarantee story structure. While such tactics have been demonstrated to work well, they undermine the distinct pleasures of emergent narrative, which I discuss in the next chapter. This dissertation revolves around a novel approach to emergent narrative that guarantees story structure without intervening in the storyworld, thereby maintaining the pleasures of the form.] Here, I mean to speak primarily in aesthetic terms (as cued by the value-laden language of ‘tampering' with a simulation), since emergent narrative is really just an artistic genre. While the form is associated with a set of architectural features and technical strategies, I view these as ultimately constituting a poetics that is in the service of a specific pleasure: emergent stories are a joy to behold because it feels like they actually happen. In the next chapter, I unpack this idea further.

#### Converging on a Definition

Let us now converge on a concise definition that captures all the features outlined so far. First, though, I have a final subtlety to note: ‘emergent narrative' refers to the approach to procedural narrative that I have described in this section, but it can also refer to actual narrative material generated by a system taking that approach. So, for the purposes of this document, our definition of ‘emergent narrative' is as follows: *narrative that emerges out of computer simulation of character activity, or the methodology of generating narrative in that way*. Critically, this definition does not stipulate that a work of emergent narrative must be interactive. While academic work in the area has typically treated emergent narrative as (specifically) an approach to interactive storytelling [aylett1999narrative,aylett1999virtual,louchart2004emergent,louchart2004theoretical,swartjes2006fabula,swartjes2010whose,walsh2011emergent,weallans2012distributed,truesdale2013suitability], I take a more ecumenical stance by which three distinct traditions are treated as instantiating the same essential form: the simulationist school of noninteractive story generation, as exemplified by systems like *Saga II* [morse1960saga], *Tale-Spin* [meehan1976metanovel], and Sheldon Klein's murder-mystery generator [klein1971program,klein1973automatic]; videogames driven by emergent character simulation, such as *Dwarf Fortress* [dwarf] and *The Sims* [maxis2000sims]; and the bottom-up approach to academic interactive storytelling that is exemplified by the ‘emergent narrative' research programs of Ruth Aylett, Sandy Louchart, Mariet Theune, Ivo Swartjes, and their respective collaborators.

 *the methodology characterized by computational systems in which narrative emerges bottom-up from the interaction of processes in underlying simulations that typically feature autonomous characters (or, alternatively, the actual narrative material produced by this method)*.

#### ‘Narrative' Confusion

Note, finally, that while I have taken great pains to define the particular meaning of ‘emergent' that is operational in the term ‘emergent narrative', I have not spent any effort on ‘narrative'. This is because the issue of what exactly ‘narrative' means is murky, particularly in procedural narrative [horswill2013science, pp. 82–83], and especially in emergent narrative, as I will discuss at length later on. First, at a general level, the word itself has multiple conflicting usages: ‘narrative' can be used in the singular to denote a concrete story (and in the plural to denote multiple stories), or it can be used as a mass noun to denote the abstract notion of narrative itself. Moreover, the words ‘narrative' and ‘story' are sometimes used as terms of art that correspond to the notions of *fabula* (the sequence of events recounted in a story) and *syuzhet* (a story as it is told), respectively. More confusingly still, these particular usages of ‘narrative' and ‘story' are sometimes eschewed in favor of the respective alternatives ‘story' and ‘discourse' [chatman1980story]. This terminological variation is couched in a century-long tradition of structuralist work in narratology, as Michael Scheffel recounts concisely (citations inserted for clarity):

> Various binary oppositions have been put forward, such as *fabula*/*sujet* (e.g. Toma¨evskij 1925 [toma1925teoria]), *histoire*/*discours* (e.g. Todorov 1966 [todorov1980categories]; story/discourse), and story/plot (e.g. Forster 1927 [forster1972aspects]), as have multileveled models such as *Geschehen*/*Geschichte*/*Text der Geschichte* (Stierle 1971 [stierle1973geschehen]; happenings/story/text of the story), *histoire*/*recit*/*narration* (Genette 1972 [genette1980narrative], 1983 [genette1988narrative]; story/narrative/narrating), and *Geschehen*/*Geschichte*/*Erzahlung*/*Prasentation der Erzahlung* (Schmid 1982 [schmid1982narrativen]; happenings/story/narration/presentation of the narration). [scheffel2010narrative, para. 4] [Note: See Scheffel's full paper for an excellent brief history of this narratological tradition [scheffel2010narrative].]

Let me emphasize this situation: some scholars use ‘story' to mean ‘fabula', while others use it to mean the counterpart, ‘syuzhet'—thus, with regard to this binary classification, ‘story' can mean one thing or the complete opposite. [Note: The issue is here has certainly been muddled by challenges rooted in translation. For more on this confusion, see Richard Walsh's account [walsh2007rhetoric, pp. 53–56].] For this reason, and because ‘story' and ‘narrative' already have extensive informal usage, I prefer to use the terms ‘fabula' and ‘syuzhet' in cases where the formal narratological concepts need to be referenced. In composing this document, however, I have not found these concepts to be of much use, primarily for the reason that they do not apply particularly well to the case of emergent narrative.

#### A Note on Terminology

As I will argue at length in Section 3.1.1, emergent narrative actually works more like nonfiction than fiction, which means that formalisms that have been developed specifically for the latter—such as the distinction between fabula and syuzhet—do not apply particularly well to this domain. In fiction writing, we may, for instance, distinguish between the order of story events as they transpired (in a storyworld), as opposed to the order in which they are recounted. While this is also true of narrative nonfiction, in that genre there is the even more primary matter of all the occurring events (in the real world) that are not recounted or even suggested in the surface account—in fiction, this base layer cannot obtain, because events that are not even suggested cannot be known to have occurred.

As such, we find that something like a three-tiered typology serves better: all the events of a world, the events that are recounted, and those events *as* recounted. Indeed, this is the essence of some of the models included in the quotation of Scheffel above, most famously that of Gérard Genette. This is the one preferred by Jonathan Lessard, a member of the reading committee for this thesis, who suggested, in his comments on an earlier draft, that I consistently use Genette's terminology (specifically, the common translations ‘narrative' and ‘story' to refer to the latter two concepts that Genette called ‘recit' and ‘narration'). This is certainly good advice, but as I mentioned above, I do not like to use ‘story' and ‘narrative' formally, on account of their well-established informal senses that denote roughly equivalent concepts. As such, in this document, I use these terms informally (and interchangeably) to mean something like ‘concrete narrative artifact'. Additionally, I will employ phrases like ‘plot' or ‘raw narrative material' to refer to specific senses of ‘fabula', and phrases like ‘presentation', ‘telling', or ‘surface text' to refer to specific senses of ‘syuzhet'.

While in those cases I choose my words informally, in this thesis I *do* adopt one particular typology whose terms I use consistently: Hayden White's model of historiography [white2014metahistory], and in turn, the terms ‘historical field', ‘chronicle', ‘story', and ‘emplotment'. Indeed, it is in White's model, as opposed to any narratological formalism targeted at fiction, that I find the strongest application to emergent narrative. This is because I believe that the latter works more like nonfiction than fiction—as I have already stated and will argue at length below—which makes the task of crafting emergent narrative strikingly akin to the practice of history writing. In the next chapter, I dive thoroughly into these concerns, but for now I would to emphasize again that in this document I do not formally employ any of the narratological terms of art quoted above (because I do not think that they apply particularly well to the concerns of emergent narrative).

#### Preview: Narrativism

Finally, to add to the terminological confusion that I have just outlined, the shape of the narrative artifact produced in emergent narrative is often nebulous. As I will explain at length in Chapter 4, in many projects the total history of everything that has been simulated is treated as the emergent narrative—but this is peculiar, because such accumulations of simulated material lack the features that are associated with narratological or informal definitions of story. Even stories that recount lived experience have far more narrative structure. By analogy, if such stories worked like emergent narrative tends to work, then the lived experience *itself* would be the narrative, rather than a story constructed to impart the lived experience. But few would call lived experience itself a story: no one says ‘You won't believe the story that happened to me today.' [Note: Instead, one says something like ‘You won't believe the story *of* what happened to me today.' Here, ‘of' is critical: the story is constructed *out of* the raw material of lived experience, which is not in and of itself a story. To claim otherwise is to conflate experience with narration, and in emergent narrative we find a related misconception: simulation is confused with narration.]

This confabulation is what scholar Espen Aarseth has termed ‘narrativism' [aarseth2004genre]. As I discuss more thoroughly in Chapter 4 (in extending critiques by Aarseth and others), I find the treatment of accumulated simulated material as narrative to be problematic, and I think it is at the root of some of emergent narrative's failings to date. The core subject of this dissertation, as Chapter 5 hammers home, is a refined approach to emergent narrative that entails the *curation* of raw simulated material into an actual narrative artifact, which is itself mounted in a full-fledged media experience. Again, I call this *curationist emergent narrative*.

## Origins of the Term

In this section, I will briefly outline the provenance of the term ‘emergent narrative'. To my knowledge, this history has not been discussed (to this extent) in any prior publication.

#### Precursor Terms

The actual approach denoted by the term ‘emergent narrative' precedes its 1990s coinage (about which more soon) by decades. As I explain in Section 4.1.2, the earliest known effort in story generation, *Saga II* (1960) [morse1960saga], is a work of emergent narrative: in this project, stories about Wild West altercations are generated by simulating the activity of two autonomous characters. Other more famous early systems, namely Sheldon Klein's murder-mystery generator (1967–1973) [klein1971program,klein1973automatic] and Jim Meehan's *Tale-Spin* (1975–1977) [meehan1976metanovel], are also associated with the approach. [Note: Both of these systems are discussed at length later on. These are not the date ranges that have been associated with those two systems in our field, but they are attested in documents that I have recently excavated. In a 1969 note on his “automatic novel writing” project, Sheldon Klein lists a project start date of September 1967 [klein1969automatic, p. 418]. Likewise, while *Tale-Spin*'s first reporting is conventionally understood to have been in Meehan's 1976 thesis [meehan1976metanovel], by then he had already published a now obscure article on the project [meehan1975using] in a 1975 issue of the *American Journal of Computational Linguistics*.] Thus, we may look at how these projects were described to identify precursor terms such as “world simulation based story generation” [dehn1981story, p. 16] or “the simulative approach” to story generation [ryan1987heuristics, p. 513].

#### Environmental Storytelling as ‘Emergent Narrative'

While the collocation ‘emergent narrative' appears incidentally (with alternative connotations) in prior works, [Note: One such paper even has the tantalizing title “Interactive Storytelling”, but it uses that phrase in the sense of interactive conversational storytelling, a practice that “provides a helpful structure for a young child's emergent narrative ability” [trousdale1990interactive, p. 169]. Another paper actually describes a kind of emergent narrative constituted in the dynamics of an annual car race and its host city [peace1991city], but this is still not what we mean by the term here.] the earliest usage that I have found with reference to procedural narrative appeared in *The Observer* in November 1994. In an article about the advent of “interactive movie games” (videogames that integrate full-motion video footage [lessard2009fahrenheit]), Jim McClellan reviews the newly released Philips CD-i game *Burn:Cycle* [tripmedia1994burn]. As McClellan notes, new mediums (such as interactive movies) tend to beget new turns of phrases: “I suppose when you're scratching together a new hybrid media form, you get to not only write your own rules but come up with your own terminology” [mcclellan1994hooray, p. 84]. One of the new phrases coined by *Burn:Cycle* developers Dave Collier and Olaf Wendt, who McClellan interviewed for the piece, was ‘emergent narrative':

>

The idea is to explore, pick up clues and uncover the narrative that has been plotted out in advance. “I think what we're doing is giving people a lot of movie components, like beads, and saying: ‘You put the thread through,'{”} says Collier. “It's what we call emergent narrative.” [mcclellan1994hooray, p. 84]

 Since interactive movies represent arguably the antithesis of emergent narrative, at least in terms of procedural approaches, it is not surprising that this connotation does not coincide with the contemporary sense of the term. Interestingly, however, Collier and Wendt's ‘emergent narrative' seems to almost perfectly denote what today would be called *environmental storytelling* [carson2000environmental1,carson2000environmental2,pearce2007narrative]. [Note: A concept borrowed from theme-park design, *environmental storytelling* is a technique in videogame design by which a game's narrative is relayed primarily through the mechanism of its spatial enviroment (and often a series artifacts that are distributed across that environment). I prefer Clara Fernández-Vara's refinement, *indexical storytelling*, which more deeply considers the role of the player: “indexical storytelling is actually more story-building, both on the part of the designer and the player—the designer creates the elements of the story and integrates them in the world, the player has to interpret them and piece them together” [fernandez2011game, p. 6]. Moreover, Fernández-Vara's framework also considers how a player may leave such traces in a world, thereby inscribing her own story onto it, as another kind of indexical storytelling. Though conventional environmental storytelling, such as that found in so-called *walking simulators* [carbo2016walking], may be antithetical to emergent narrative, I see indexical storytelling as actually being quite compatible. For one, world simulations are apt at incorporating player behavior, which makes player-driven indexical storytelling a strong prospect in works of emergent narrative. I am also interested in the prospect of *generative* environmental storytelling. As I note at several points later on, this is arguably an unavoidable feature of world generation, especially when artifacts are modeled: the result of such a procedure is a storyworld scattered with indices that encode and suggest the history of that world. What I would personally like to see is exploration of game designs that are specifically built on this idea. In a recent paper, Daniel Livingstone and collaborators have promoted this phenomenon in terms of what they call *archaeological storytelling* [livingstone2016archaeological]. Relatedly, scholars in the larger area of *archaeogaming*, which is about “archaeology in and of videogames” [reinhard2018archaeogaming,reinhard2017video], promote an approach to gameplay that treats simulated storyworlds as archaeological sites (for instance, the worlds of *No Man's Sky* [flick2017exploring]). More broadly, Tanya X. Short has called for simulationist game designs that encourage player investigation of simulated material [short2017designing, p. 114-115].]

#### Galyean's Forgotten Coinage

When it comes to the current sense of ‘emergent narrative', we find a first usage in the work of Tinsley “Tag” Galyean, who at the time was a PhD student at MIT. In a 1995 paper presented at the AAAI Spring Symposium on Interactive Story Systems: Plot and Story, [Note: This symposium was an important early meeting in the academic field of interactive storytelling, which is today represented primarily by the workshop on Intelligent Narrative Technologies and the International Conference on Interactive Digital Storytelling (ICIDS). These were preceded by the AAAI Fall Symposium on
Narrative Intelligence (1999), as well as the International Conference on Virtual Storytelling (2001–2007) and the International Conference on Technologies for Interactive Digital Storytelling and Entertainment (2003–2006), which merged to form ICIDS. Sadly, while each of the other eight AAAI Spring Symposia of 1995 published proceedings, Interactive Story Systems: Plot and Story did not (I confirmed this with AAAI). However, it appears that working notes for the symposium (comprising the papers) were distributed to attendees, so there is hope that this (otherwise lost) pioneering historical work is extant. Among its papers, for example, is one by Janet Murray apparently titled “Dr. Quinn on the Holodeck, or blueprint for an electronic storyland” [murray1995quinn] (as cited later by Clark Elliot [elliott1998story, p. 188])—this would have preceded her famous *Hamlet on the Holodeck* [murray1997hamlet] by two years. To my knowledge, however, the *first* academic meeting in the area was the 1990 AAAI Workshop on Interactive Fiction and Synthetic Realities, which was organized by Joe Bates and Abbe Don (and for which a proceedings was likewise not published) [bates1990computational]. In a 1991 paper, Bates stated that the workshop was at that point “the only place that workers in the relevant fields have come together to seriously discuss broad research toward inhabited, dramatic virtual worlds” [bates1992virtual, p. 3]. Elsewhere, in a newsgroup post, Bates explained that the field was “starting
to become acceptable as a ‘serious' AI research area” [bates1990how, n.p.]. Among researchers writing in this century, only Federico Peinado has referenced this meeting [peinado2006interactive, p. 43].] Galyean articulates an interventionist approach to interactive storytelling akin to drama management, which he calls “narrative guidance” [galyean1995narrative]. [Note: Curiously, this paper does not cite anything else, let alone the highly related work on drama management carried out by the symposium's co-organizer, Joe Bates [kelso1993dramatic]. To be fair, Galyean takes a distinctly cinematic perspective and is particularly interested in how, for example, camera control may be used as a technique for dramatic intervention, as his dissertation elaborates [galyean1995thesis].] Critically, Galyean positions this approach in opposition to what he calls ‘emergent narrative', defined as follows:

> We all construct narratives out of our daily activities to help us remember, understand, categorize and share experiences. It is this skill that many interactive systems exploit. They give us environments to explore. We, by combining the elements of these spaces with our goals (the user's goals), allow a narrative to emerge. If any narrative structure (or story) emerges it is a product of our interactions and goals as we navigate the experience. I call this “Emergent Narrative.” This approach has provided a number of successful interactive experiences such as flight simulators, games (i.e. DOOM), and narrative puzzles like MIST [Myst]. [galyean1995narrative, p. 1]

 While this definition differs from mine, it is quite a bit closer than the sense of the phrase discussed in the previous section. In a paper presented a few weeks later at another symposium, Galyean describes the notion again:

> Architectural walk throughs for example, scientific visualization, and even games like DOOM place us in alternative worlds while giving us methods for navigating these virtual spaces. These methods allow smooth and continuous interaction that can immediately influence the constantly changing presentation, but they rely on the user's actions and thoughts to bring structure to the experience. If any narrative structure (or story) emerges it is a product of our interactions and goals as we navigate the experience. I call this emergent narrative. [galyean1995guided, p. 103]

[Figure omitted]

#### Aylett's Academic Usage

Galyean's term did not catch on, perhaps because his contritubitions that introduced the idea were actually about an oppositional interventionist approach (his ‘narrative guidance'). In any event, it was a few years later, at the 1999 AAAI Fall Symposium on Narrative Intelligence [mateas2003narrative], that the phrase ‘emergent narrative' would come into wider usage in the field. Here, Ruth Aylett presented a (now seminal) paper titled “Narrative in Virtual Environments—Towards Emergent Narrative”, which introduced the bottom-up character-centric approach that she had been exploring in her (yet unpublished) *Virtual Teletubbies* project [aylett1999virtual] (see Figure 2.1). Aylett does not explicitly define ‘emergent narrative' in the paper, but instead treats it as an intuitive concept that can be explained as the bottom-up mechanism by which narrative obtains in example domains like soccer matches and improv scenes [aylett1999narrative, pp. 84–85]. [Note: In several earlier papers by Barbara Hayes-Roth and collaborators (beginning in 1995), the patterns of improv acting are invoked to articulate a new paradigm for procedural narrative [hayes1995multiagent,hayes1995agents,hayes1995directed,hayes1995gdc,hayes1996improvisational,hayes1997story]. Ken Perlin and Athomas Goldberg also adopted this perspective in an antecedent project [perlin1996improv]. However, none of these publications use the language of ‘emergence' to describe their systems. A striking 1994 article by John Barger (the inventor of the term ‘blog') speaks of authoring toward “the stories you want to see ‘emerge'{”} in a blue-sky system for interactive storytelling [barger1994confabulating, n.p]. With uncanny prescience, Barger almost exactly describes *The Sims*, even using the moniker “SimNormal”: “I like to imagine a sort of SimCity-like environment where a community of little NPC's go through their daily routines of life, in a completely rote fashion [...] In honor of Normal, Illinois, I call this world “SimNormal”. [...] So the game might open with a person in their apartment, waking up in the morning, and if they go outdoors they may be presented with a series of neighbors whose dramas they can become involved in [...] If the player chooses not to follow up any given opening, that's no problem... there will be many other options to explore” [barger1994confabulating, n.p.]. Alternatively, a fascinating 1998 paper by Peter Bøgh Andersen considers the physical sense of ‘emergence'  in describing *The Eye of Wodon*, whose storyworld is represented as a dynamical system that both the player and the system perturb [andersen1998multimedia]. For a deep dive into the connection between dynamical systems and narrative, see Jo Alyson Parker's monograph *Narrative Form and Chaos Theory in Sterne, Proust, Woolf, and Faulker* [parker2007narrative].] From here, Aylett's research program took off, producing a series of papers on the topic (discussed in the next section), and the term ‘emergent narrative' became a common token in the academic field of interactive storytelling (though one sometimes specifically associated with Aylett's program).

#### LeBlanc's Game-Design Usage

Around this same time, the term ‘emergent narrative' also entered the parlance of game design, both in academia and industry, through the apparently independent coinage of game designer Marc “MAHK” LeBlanc. At the Game Developers Conference in 2000, LeBlanc delivered a presentation titled “Formal Design Tools: Emergent Complexity, Emergent Narrative”. Video does not appear to survive (though LeBlanc's slide deck is available online [leblanc2000formal]), so I will quote from Ernest W. Adams's summary of the talk (included in a 2005 lecture):

> At the 2000 Game Developers' Conference, Marc LeBlanc gave a lecture called “Emergent Complexity, Emergent Narrative.” He introduced the idea that narrative can emerge from complex automated systems rather than from pre-written blocks of material. He made a distinction between what he called “embedded” and “emergent” narrative. Embedded narrative is pre-constructed, and the player encounters and experiences it in the course of gameplay. Emergent narrative arises out of the process of playing. [adams2005interactive, n.p.]

 As Adams notes, LeBlanc's ideas were not new—in a 1995 lecture on interactive movies, Adams himself commanded: “Your job is not to tell stories; your job is to build worlds in which stories can happen” [adams1995challenge, n.p.]—but nonetheless his explicit formalization, and especially the concise term ‘emergent narrative', were helpful additions to the game designer's conceptual library (and lexicon). [Note: Confusingly, LeBlanc's other coinage, ‘embedded narrative', coincides with Marie-Laure Ryan's term for “any story-like representation produced in the mind of a character and reproduced in the mind of a reader” [ryan1986embedded, p. 320]. As Ryan acknowledges, the term has also been used more loosely by narratologists to refer to any narrative material embedded in other narrative material, such as in the case of a frame story [bal1981notes].]

#### A Commonplace Term

Shortly after LeBlanc's presentation, game designer Eric Zimmerman referenced his concept of emergent narrative in a piece titled “Against Hypertext” (2000) [zimmerman2000against]. In 2004, Zimmerman co-authored (along with Katie Salen) a popular book on game design, *Rules of Play: Game Design Fundamentals*, which includes several segments on emergent narrative [salen2004rules, pp. 383–385, 402, 418, 579–580]. As the authors acknowledge, they adapt the concept from LeBlanc's 2000 presentation, [Note: Though Salen and Zimmerman accidentally cite LeBlanc's Game Developers Conference talk of 1999, which described feedback systems in games (and was built upon by his 2000 presentation on emergent narrative) [leblanc1999formal].] and from here the term (and concrete idea) became firmly entrenched in the (conceptual) lexicon of game design. Indeed, it is likely primarily through this book that the term has found widespread usage in lay circles. [Note: To be clear, others were also writing at this time about the more general phenomenon of emergence in games. Here, a seminal example is Jesper Juul's 2002 paper “The Open and the Closed: Games of Emergence and Games of Progression” [juul2002open].]

Finally, it is worth noting two other early usages of the term in influential papers, both of which appeared in the 2004 volume *First Person: New Media As Story, Performance, and Game*, edited by Pat Harrigan and Noah-Wardrip-Fruin. In the seminal “Game Design as Narrative Architecture” [jenkins2004game], Henry Jenkins utilizes the terms ‘embedded narrative' and ‘emergent narrative' with LeBlanc's same senses for them, but does so without citation (nor is Aylett or Galyean cited). A few pages later, in “Towards a Game Theory of Game” [pearce2004towards], Celia Pearce uses ‘emergent narrative' as a commonplace term (i.e., undefined, without citation). [Note: In his book *Half-Real*, published the following year, Jesper Juul critiques the concept by referencing both Jenkins's and Pearce's articles as its primary articulations [juul2011half, pp. 157–159].] Interestingly, these scholars would likely have been exposed to both Aylett's and LeBlanc's ideas. In any event, it is safe to say that the term ‘emergent narrative' was in common use—“in the zeitgeist” [boring1955dual]—by 2004.

## Brief Overview

As I noted in the last section, the emergentist approach to procedural narrative is as old as the field itself, originating in some of the earliest known story generators: *Saga II*, *Tale-Spin*, and Sheldon Klein's automatic novel writer. Many more projects have followed in recent decades, including academic systems that demonstrate AI research, like *Virtual Storyteller* [theune2003virtual], *FearNot!* [aylett2005fearnot], and *Prom Week* [mccoy2013prom], as well as commercial games like *The Sims* [maxis2000sims], *Dwarf Fortress* [dwarf], *RimWorld* [rimworld], and *Caves of Qud* [grinblat2010caves]. Other character-based systems, such as the *Petz* series of digital games [frank1997socially], the interactive-story system of Marc Cavazza and others [cavazza2002character], and Chris Crawford's *Storytron* [crawford2012chris], could also be called works of emergent narrative.

The *emergent narrative* research programs cultivated by Ruth Aylett, Sandy Louchart, Mariet Theune, Ivo Swartjes, and others have typically concerned only interactive systems, most often research systems like *FearNot!* [aylett2005fearnot] and *Virtual Storyteller* [theune2003virtual]. As I have stated already, in this dissertation, I do not limit my concerns to interactive systems. A basic appeal of emergent narrative is of course the very *emergence* that it yields, producing stories that even a system's designer might not have anticipated. Indeed, games with rich underlying simulations have often been lauded for their emergent qualities [juul2002open,alexander2013spector,whitehead2013systems]—this is what got *Dwarf Fortress* into the Museum of Modern Art [winslow2013simcity]. The fundamental rhetoric in academic work advocating this bottom-up approach is that it defeats the sticky issue in top-down interactive narrative of accommodating player actions [riedl2006linear,aylett1999narrative]. If the affordances given to a player are a subset of (or are coextensive with) the actions that non-player characters (NPCs) may themselves take in an underlying simulation, then an emergent-narrative system will be inherently reactive to player inputs. As such, the stories that emerge in these systems naturally incorporate player actions. In Aylett's early formulation, emergent narrative was articulated as a solution to this problem of conciliating interactivity and narrativity [aylett1999narrative], which has been called the *narrative paradox* [louchart2004theoretical,aylett2005fearnot] and the *interaction dilemma* [peinado2004transferring]. I am not aware of any championing of the form outside of the context of interactive experiences, though several noninteractive story generators have taken the approach. Generally, the simulationist approach to story generation has been lambasted, as I will explain at length in Chapter 4.

Papers in the area of emergent narrative have typically introduced new systems [aylett2005fearnot,brom2009emohawk,alvarez2011socially] or more generally concerned the engineering of systems [louchart2004emergent,aylett2006unscripted,weallans2012distributed,truesdale2013suitability,bevensee2013conceptualizing], though others have discussed emergent narrative through the lens of narrative theory [louchart2004narrative,louchart2004theoretical], creativity [kriegel2008emergent,rank12012creativity,aylett2013contain], and improvisation [swartjes2007emergent,swartjes2010whose]. Finally, another class of contributions to the discipline has concerned the particular *design practices* that are employed in the domain [louchart2008purposeful,louchart2008authoring,suttie2013theoretical,kriegel2007authoring,samuel2014introducing].

In this dissertation, I provide what is to my knowledge the most extensive *apologetics* for emergent narrative, with particular attention to the hidden merits of dismissed noninteractive systems, such as Sheldon Klein's murder-mystery generator [klein1971program,klein1973automatic]. In turn, I also provide what I believe to be the most extensive *polemic* for the form to have appeared so far. Here, my approach is to paradoxically take the most devastating takedowns of emergent narrative as an intellectual *starting point* for a refined instantiation of the form, *curationist emergent narrative*. Throughout the discussion, I will weave intellectual, technical, design, and historical threads into a holistic account of what has happened so far and where we might go next.