# Curating Simulated Storyworlds — Chapter 10: Case Study: Bad News > 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 10 of 14 only: Case Study: Bad News (~22,000 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 Adam Summerville and I had developed a prototype for a videogame about character knowledge phenomena—*Talk of the Town* (2015), as I explained in the previous chapter—but it was not hitting the mark. Following a critique session in Noah Wardrip-Fruin's class [wardrip2015cmpm290j], Adam emailed me to propose a revised game design. In the email, he noted that our initial asymmetric multiplayer design was all about character knowledge propagation, but our eventual simulation engine (the subject of the previous chapter) produced additional interesting phenomena that our original design did not showcase: > I think an important part of [the class critique] was about the networks that people build up. We only surfaced that as knowledge propagation, but I think it's not just knowledge propagation. We need to get at friends and family, and enemies and rivals. [Note: Personal communication, March 13, 2015.] Adam's proposed redesign, aimed at capturing all of the social intrigue modeled in the simulation engine, was a single-player realtime experience inspired by *The Legend of Zelda: Majora's Mask* [nintendo2000majora]: > As I've mulled on this, I think that something more akin to Majora's mask is the way to go. [...] In MM you meet up with a gang of kids early on [...] who give you a notebook. The notebook is important for 2 things: 1) As you learn about people's daily habits it gets put into the notebook. 2) As you learn people's goals/dreams/aspirations it also gets put into the notebook. The main part of MM (that isn't the traditional going into dungeons and killing things and getting items) is all about helping these people out. But to do that you first have to figure out what they do around town, where they'll be, who they interact with, and why. I think that something like that would be a much more interesting use of the simulation than our current game. [...] It could be the same kind of idea. You could be a Mary Poppinsesque character who rides into a town filled with unfulfilled people, and you have to follow them around, see what makes them tick and make them happier [Note: Ibid.] In the same email, Adam also emphasized the issues with our overwhelming menu system for conversational interaction. I am extremely interested in the prospect of freeform conversational interaction in videogames, in the style of *Facade* [mateas2003integrating], and it seemed like the redesign could be an opportunity to explore this territory. At that time, Ian Horswill had recently published on *MKULTRA* [horswill2014game,horswill2014dialog], a videogame prototype featuring a clever *autocompletion*-driven dialogue interface that suggests completions to partial free-text inputs, thereby guiding players toward utterances the system can actually understand. [Note: *MKULTRA* also features *belief injection*, whereby players reprogram character beliefs through free-text input, and *AI sonification* [horswill2015sonification], by which character decision making (in an underlying Prolog engine) is cleverly sonified: “MKULTRA includes a sound synthesizer that is driven by a log of the activity of an NPC's problem solver, thereby allowing the player to ‘hear the character think' in real time” [horswill2015mkultra, p. 224].] In my response to Adam's proposal, which I was very excited about, I suggested that we could use this technology in our new design: > As for the dialogue redesign—the more I think about it, the more I think we need to just allow free-text input. Ian Horswill is working on a game (underpinned by an open-source Unity framework) in which the player character is controlled by free-text input, but crucially he has a bidirectional parser that suggests autocompletions (of partially composed commands) that are commands that the system can actually understand. Unlike Facade, it won't accept a command that it can't understand. With the autocompletions serving as scaffolding in the (otherwise open) choice space, I think it's a really nice middleground between a straight-up parser and a menu with hundreds of choices. If we wanted to go this route, I could email him about getting started with integrating this into our project [Note: Personal communication, March 14, 2015.] Adam was on board with the idea of using the *MKULTRA* technology, and so we emailed Ian Horswill, who was also excited about the prospect. Soon, this shift led to a major design decision: we would rewrite the *Talk of the Town* simulation engine using the *MKULTRA* framework. As I briefly mentioned in the last chapter, the integration of my Python simulation engine into the larger Unity project of our game proved to be a technical nightmare. Since Ian's technology was already fully integrated into a unified Unity project, it seemed advantageous to ditch the Python code and start from scratch using his code. Unfortunately, due to busy schedules all around, especially mine, our redesign and technical overhaul never got off the ground. It seemed that the project had fizzled out. [Note: Ian is a friend of the lab whom I already knew outside the context of this missed connection. We have kept in touch since, and he is a member of the this dissertation's reading committee.] Meanwhile, however, I had been gradually developing my own approach to conversational AI in games. Since arriving at UC Santa Cruz to work with the Expressive Intelligence Studio and Marilyn Walker's Natural Language and Dialogue Systems lab, my research agenda had primarily concerned the development of technologies that would enable freeform conversational interaction in games. Initially, this took the form of a project in which *Prom Week* [mccoy2013prom] dialogue exchanges were decomposed into individual lines, which were annotated and then procedurally recombined to form new dialogue exchanges that could express previously inexpressible storyworld states [ryan2014toward,ryan2014combinatorial]. Through a gradual design evolution, the interface for dialogue annotation became a full-fledged authoring tool—the idea was that dialogue could be annotated *at authoring time* to enable intelligent recombination as the primary generative mechanism. Under the influence of *Tracery* [compton2014tracery,compton2015tracery], a tool that had recently been created by my labmate Kate Compton, recombination was reformulated to work according to the operation of a generative grammar. As Adam and I were embarking on our game redesign, undergraduates Taylor Owen-Milner and Max Fisher produced a prototype of the tool, which I called *Expressionist* [ryan2015natural] (see Section 12.2). With a simulation engine already in place and an authoring tool for conversational AI developing, I had a wild idea that would incorporate the unique skill set of my labmate Ben Samuel. Here is an email that I sent to him in August 2015: > I hope you've had a great summer so far! I have a crazy idea [...] and I'm writing to solicit your help with it. [...] I want to run a computationally assisted live-action prototype, where players interact with the game by: freely speaking to me (game master) to give commands about things like navigation, etc.; freely speaking to game characters to inquire about whatever they would want to inquire about. After the player asks a game character something, I would probe that character's mind to see what knowledge they have regarding the inquiry, relay it to a live actor playing the character, and then the live actor would improvise a response that expresses the particular bit of character knowledge that I relayed. [...] So, as you might expect, I'm writing to ask you if you'd like to be the live actor (you would play every character). [Note: Personal communication, August 13, 2015.] Ben, who is now a professor at the University of New Orleans, is a unicorn: he is an AI researcher and a professional actor. Before I got to Santa Cruz, he had co-developed *Prom Week*, so he was no stranger to social simulation. As for acting, his performance in Hulu's first original scripted series, *Battleground* (2012) [walsh2012battleground], earned him praise from the *New York Times*: > Mr. Samuel, who in addition to acting is working toward a doctorate in computer science at the University of California, Santa Cruz, combines elements of the “Harry Potter” Daniel Radcliffe and Matthew Gray Gubler's Dr. Reid on “Criminal Minds,” and he's the best reason so far to watch “Battleground.” [hale2012shenanigans, p. C2] [Note: Incidentally, *Prom Week* and *Battleground* debuted on the same day in 2012.] This idea that I proposed to Ben had actually emerged out of discussions with Jonathan Lessard, a professor at Concordia University in Montreal, whom I had met at the Foundations of Digital Games conference in the beginning of summer. [Note: Jonathan Lessard is also a member of the reading committee for this dissertation.] At the conference, Jonathan presented a poster titled “Design Rationale for Natural Language Game Conversations” [lessard2015design], which reported on *A Tough Sell* (2014) and *SimProphet* (2015), two experimental videogames centered on freeform conversational interaction. [Note: Soon after, Jonathan released a third game in this vein, *SimHamlet* (2016), thereby culminating his *LabLabLab* trilogy [lessard2016designing]. I propose to call videogames like these—ones centered on conversational interaction—*talking simulators*.] I was thrilled to encounter this work, and Jonathan was intrigued to hear about the development of both *Expressionist*, which he would eventually utilize in a later project [lessard2017striving], and *Talk of the Town*, which dovetailed with his own interest in character simulation (e.g., see [lessard2016character]). We kept in touch over email after the conference to discuss the prospect of collaboration. Jonathan is a master of scope, which I am certainly not, and in one of our conversations about the *Talk of the Town* redesign—now conceived in large part as a testbed for my developing conversational technology—he suggested that I first explore the design space in a smaller project: > For example: Find John Smith, he was last seen at Awesome Jeans [...] Other small puzzles could be: Someone died in apt. X, find out who lived there and then find his closest relatives (to notify them). A casting agency is looking for someone with X and Y facial features, find someone fitting that description. [Note: Personal communication, July 29, 2015.] I loved Jonathan's ideas and decided to embark on a prototype that would follow the second premise: explore a *Talk of the Town* storyworld to find and notify the next of kin of an unidentified deceased character. With an early version of *Expressionist* now at my disposal, I was ready to start building a dialogue system for this videogame prototype. Immediately, however, I encountered a design problem, which I explained in this way in my email to Ben: > In setting out to build this game, I've come across a huge problem: I want dialogue to be the primary player interaction, but I want to be smart about writing content for the game, since the whole point is that it should be a small first exploration. More specifically, I want to write as little content as possible, but have that content pool closely align with the dialogue players would naturally want to say and hear in the game. [Note: Personal communication, August 13, 2015.] As such, the idea that I had proposed to Ben—running a computationally assisted live-action prototype—was specifically aimed at gathering a corpus that could guide the authoring process: > In the end, after audio-recording all player interactions, I'd have a corpus of natural/common character interactions in the domain of the gameplay experience, and I could then rely on this corpus during content creation. [Note: Ibid.] Needless to say, Ben was excited about this strange prospect and agreed to serve as actor. I reached out to Adam, as well, since the project was an offshoot of our work on *Talk of the Town* (2015), and he joined as a third codesigner (after writing back, “This is a crazy idea, but I kind of love it”). [Note: Personal communication, August 26, 2015.] We called the experiment *Bad News*, and within a few days we began preparing for its installation as a demo at the upcoming Experimental AI in Games workshop [ryan2015bad]. Improbably, within a year and a half of these humble beginnings, *Bad News* would be performed internationally at venues including IndieCade, Slamdance, and the San Francisco Museum of Modern Art, and featured in publications such as *Rolling Stone* [wright2017how] and even on a television morning show [mountain2017bad]. A collaboration with Ben Samuel and Adam Summerville, *Bad News* is an installation-based immersive experience for one or two players. [Note: These names are listed alphabetically. Though we take various roles in the context of a particular performance—guide, actor, wizard—with regard to the creation of this work we each take the title ‘codesigner'. As such, there is no meaning to any ordering of our names.] Through its utilization of the *Talk of the Town* simulation engine, described in the previous chapter, the storyworld for each performance is procedurally generated prior to the start of gameplay. During gameplay, a player explores this storyworld and converses with its denizens in an effort to find and notify the next of kin of an unidentified deceased character. The installation is driven by a *wizard-of-Oz* configuration [dahlback1993wizard] that incorporates a human experience manager and human actor, who collaborate to curate the storyworld's emergent material into narrative artifacts that are mounted into the experience through the mechanism of conversational storytelling. While earlier publications have reported on several aspects of the project [ryan2015bad,ryan2016bad,samuel2016bad,samuel2016crafting], in this chapter I will provide a general overview before focusing discussion on its instantiation of curationist emergent narrative, the subject of this dissertation. Though I will also describe how the project has evolved over time, this chapter concerns its final instantiation. ## Overview *Bad News* combines wizard-of-Oz techniques [dahlback1993wizard] and live improvisational acting into an emotionally charged one-on-one experience whose story and setting is uniquely generated, for each performance, by the *Talk of the Town* simulation engine. Each performance takes place in the summer of 1979, at which time a resident in a computer-generated American small town has died alone at home, and a mortician's assistant—the player—is tasked with tracking down and notifying the next of kin. To do this, the player navigates the simulated town to interact with its residents, who are each played live by a professional actor. Throughout gameplay, an unseen wizard listens in remotely to manage the unfolding experience via live coding and discreet communication with the actor. #### Antecedents and Related Pieces Before describing *Bad News* at length, I would like to mention some of the few related pieces in the peculiar area of *computationally assisted live performance*. First, *Saga II* (1960) [morse1960saga], the earliest known work of computational narrative, should probably be considered an example of this form. As I discussed at length in Section 4.1.2, in this collaboration between MIT and CBS-TV, seasoned Hollywood actors performed scripts generated by an early story generator. A more recent and better known example is *The Bus Station* (1993), a “live experiment” developed by Oz Project stalwarts Joe Bates and Peter Weyhrauch, in collaboration with playwright Margaret Kelso, as part of a deep exploration of drama management [kelso1993dramatic]. This piece placed interactors among improvisational actors in a tense scenario managed by a hidden director. Earlier, around 1983, Brenda Laurel and Eric A. Hulteen explored similar territory in the context of the “media room” at the Atari Sunnyvale Research Lab. In a company memo, Hulteen describes the space: > The media room itself is not so much a room in the traditional sense as a space or an environment. It has the capability to receive/understand/input all forms of human motor activity. Which is to say that it can sense any change that you can affect in the environment but that it can't read your thoughts. The room also has the capability to generate/output information to all of the human sense organs. It can provide things for you to see, hear, smell, feel, and taste. [hulteen1983media, n.p.] In a follow-up memo that was written a few months later, Laurel and Hulteen propose an in-house experiment called *The First Fischell–Turing Production*, to be held in the media room: [Note: Here, ‘Fischell' refers to Arthur “Artie” Fischell, an infamous fictional persona that Laurel and colleagues cultivated to the degree that Atari officials believed he was an actual employee of the company [fisher2018valley].] > The central goal [...] is to see how a first-person, interactive fantasy system implemented in the media room might work. We envision [...] a simulation of a single-user interaction with such a system. [...] The actual event will have the flavor of a planned improvisation, using operational technical elements where they exist and human improvisation for portions of the system that are still in the conceptual hacking stage. [laurel1983media, p. 2] Like *The Bus Station*, this experiment was intended to explore the prospect of doing drama management with a fully computational system: > The project will utilize a structure which is based on the interactive story idea, with a playwriting expert system (*Play-Right*) [...] which orchestrates system-controlled events and characters so as to move that collectively generated story forward in a dramatically interesting way. [...] The operational elements of the system will include some interactive devices, video, graphics, and animation, sound, and other other environmental effects. [laurel1983media, p. 2] Laurel and Hulteen's articulation for the merits of this approach strikingly anticipates not only *The Bus Station*—which is not very surprising, considering that Laurel was an ancillary member of the Oz Project and its intellectual forebear—but also the impetus for *Bad News* that I recounted in the opening of this chapter: > our purpose is to study the design of a system, major portions of which have yet to be implemented, by simulating its operation. The exercise will be improvisational in the sense that humans will enact roles that will, hopefully, later be automated [laurel1983media, p. 2] While I do not know for sure whether *The First Fischell–Turing Production* was ever performed, I have reason to believe it was. In a planning document that Oz figurehead Joe Bates drafted in preparation for the 1995 AAAI Spring Symposium on Interactive Story Systems—which I acquired from my advisor Michael Mateas, who photocopied it sduring his time as a graduate student on the Oz Project—a proposed agenda schedules video presentations on both Oz and Atari live experiments (presumably the ones I have mentioned): > Interactive Drama Experiments from CMU, Atari [...] Joe [Bates] shows two 10 minute videos of Oz live drama experiments and video comments by actors, director, playwright, users after experiments end. [...] Brenda [Laurel] shows/presents the live drama experiments done and planned at Atari in the early 80s In any event, a follow-up report by Laurel (also in 1983) laid out more extensive details on the planned implementation [laurel1983toward]. Eventually, her ideas about interactive fantasy would solidify in her 1986 PhD thesis [laurel1986toward] and the seminal 1991 monograph *Computers as Theatre* [laurel1991computers]. Later obscure projects in the area of computationally assisted performance include *Pleasure Island* (1999), a networked “online drama” pitting interactors against improvisational actors [chen2000pleasure], and *Terra Incognita* (2002), a live performance incorporating robot actors [sgouros2001generation,sgouros2002terra]. *AR Facade* (2006) was a gallery installation that mounted *Facade* [mateas2003faccade] in an *augmented reality* [azuma1997survey] environment, with human operators intervening live to guide its reactive-planning ecosystem [dow2006initial,dow2007ar,dow2008styles]. *Party Quirks*, by Brian Magerko and collaborators, is an agent-based mixed-initiative framework for playing a popular improv game [magerko2011digital,oneill2011knowledge]; generally, Magerko and his collaborators have been active in this area for the last decade [magerko2011employing,jacob2013viewpoints,hodhod2014computational,jacob2018creative]. In 2012, startup Alternate Reality produced a networked game that placed improvisational actors in precarious physical situations, which players were tasked with rectifying with the help of other in-game characters played by humans [kessler2012truman]. More recently, Rob Wittig and Mark C. Marino have developed the *netprov* form [marino2012netprov,burr2015bicycles], which they describe like this: > A typical netprov, if such can be described, involves a networked of platform (such as Twitter, a blog, or Facebook), a set of collaborators, a narrative premise, and a set of constraints. For the most part, netprovs have time limits, although that is not a necessary requirement. Fundamental to netprov is creative play on contemporary platforms. [wittig2017occupy, p. 4] While generally we found out about these projects later on, one antecedent directly inspired the design of *Bad News*: Dietrich “Squinky” Squinkifer's *Coffee: A Misunderstanding* [squinkifer2014coffee], which is a computationally assisted interactive play that Squinky has described in this way: > Each character in “Coffee” is portrayed by two audience volunteers. One, the “driver”, is given sets of branching dialogue options at key points in the story, which are selected using a menu displayed on a mobile device. The other volunteer plays the “puppet”, who recites the dialogue selected by the driver, the lines for which are also displayed on a mobile device. Meanwhile, the other audience members are able to see each character's dialogue options via a background projection, and are encouraged to heckle the performers and attempt to sway the drivers in their choices. [squinkifer2014coffee, p. 1] Squinky, a former UC Santa Cruz graduate student, ran a performance of this piece as a class project in Noah Wardrip-Fruin's *Playable Media* class [wardrip2014cmps290j], which I thoroughly enjoyed as an audience member. While Ben and Adam were not in the class, they were aware of the project, and it was almost certainly in the back of my head as the initial ideas about *Bad News*'s peculiar configuration began to take shape. Squinky has since produced additional games in this tradition with collaborator Jess Marcotte—together, they form the collective Handsome Foxes in Vests. Here is their description of *Most Sincere Greetings, Esteemed One* (2016): > Each pair of players is asked to perform a set of procedurally-generated instructions for greeting one another, then taken to a page where they receive procedurally-generated feedback on how well (or poorly) they executed the greeting. [squinky2016most, n.p.]. Caressing a cybernetic plant in *rustle your leaves to me softly* (2017) triggers a sensual procedural soundscape [squinky2017rustle, marcotte2017ggj]. Finally, *The Truly Terrific Traveling Troubleshooter* (2017) is a “radically soft physical/digital hybrid roleplaying game about emotional labour and otherness, which fits entirely inside of a carry-on suitcase” [squinky2017truly, n.p.], that being the titular Troubleshooter's toolkit [squinky2017truly2,marcotte2017truly]. I am aware of two projects in this growing area that have actually been inspired by *Bad News*. Ian Horswill's *Dear Leader's Happy Story Time* (2016) [horswill2016dear] leverages a story generator that operationalizes media tropes to produce kitschy plot outlines that are meant to be acted out by a group of friends as a party game. As Horswill explains, the project very intentionally leans into the clunky and mechanical aesthetics that characterize the outputs of story generation (a notion that I discussed in Section 3.1.4): > The approach taken in *Dear Leader* is to salvage bad stories through playful, ironic performance; calling attention to their artificiality and theatricality; and surfacing the mechanistic character of their generation. [horswill2016dear, p. 40] Horswill's project pushes further into the territory he explored in the earlier *Fiascomatic* (2015) [horswill2015fiascomatic], a scenario generater that produces storyworld representations that are meant to be used for Jason Morningstar's tabletop storytelling game *Fiasco* [morningstar2009fiasco]. In each of Horswill's projects, computational systems produce material that human interactors are meant to act out or perform in some way. The other project in this area that has been inspired by *Bad News* is *Hot Bread* (2017), a project by researchers Rossana Damiano, Vincenzo Lombardo, and Antonio Pizzo, and their students at the University of Turin [damiano2017thinning]. To create this piece, Lombardo used *Talk of the Town* to generate a storyworld whose emergent material was explored over the course of multiple weeks in a class on interactive storytelling. [Note: At this time, town seeding was not working properly in the simulation engine, which meant that Lombardo had to maintain the storyworld on his laptop for the duration of the experiment. Unfortunately, toward the end of the scheduled period for story sifting, he forgot about this and restarted his computer, and as a result the world disappeared forever.] Eventually, the students homed in on a romance between two characters who worked at rival bakeries in the town, and they used this material to build *Hot Bread*, a work of curationist emergent narrative that hooks into the researchers' *DoPPioGioco* framework for computationally assisted live performance that uses computer vision and emotional modeling [damiano2017doppiogioco]. The following year, I visited Italy to collaborate with this group on what became *Cattive Notizie*, a variant of *Bad News* that I describe in Section 10.2.5. Finally, I will note that there is also growing scholarly interest in the scholarly area of computationally assisted live performance, as a number of recent papers attest [martens2016towards,samuel2016computatrum,striner2017characterizing,azad2017scheduling]. ## Installation and Experience I will now provide a fairly extensive description of *Bad News*, but again my primary goal in this chapter is to provide a case study that considers the experience with regard to my framework for curationist emergent narrative. [Note: The overview provided in this section adapts the prose of a design document for the project titled “Bad News: Behind the Curtain” [ryan2017bad], which we have published on our project website.] For other reports on the project, see our earlier conference papers [ryan2016bad,samuel2016bad] or Ben Samuel's dissertation [samuel2016crafting]. Additionally, an earlier workshop abstract provides intellectual context on our initial conception of the project as a live-action prototype that would guide content creation for a digital game [ryan2015bad]. ### Premise It is the summer of 1979 and a resident of an American small town has died alone at home. The death has come to the attention of the office of the *county mortician*, though the event is not known in the town itself. The cause of death is unknown, but foul play is not suspected. [Note: *Bad News* is very intentionally not a murder mystery, but rather a meditation on the splendor of every life. While fortunately the death of a loved one is not an everyday matter in the context of our individual lives, it is an everyday matter in this world. This is what the piece explores.] In the town's jurisdiction, it is the responsibility of the county mortician's office to notify the next of kin of any death. Unfortunately, the county mortician is occupied by another matter in the county, which means his newly hired assistant (the player) is tasked with handling this particular case. Given a description of the body and the address at which it was discovered, the mortician's assistant must use this information to: 1) determine the identity of the deceased person, 2) determine the identity of the next of kin (closest familial relation in the town), and 3) track down and notify the next of kin of the death. [Note: More specifically, the next of kin is the closest living familial relation in the town, where closeness is defined according to the following hierarchy: spouse, parent, child, sibling, extended family. In many cases, there may be multiple next of kin, meaning multiple characters who are equally closely related to the deceased character. For example, if a character has several living siblings in town (and no spouse, parent, or child there), then each of the siblings is a next of kin. In such cases, notifying any single one of them will suffice.] Again, there are no indications of foul play, and indeed it is not the player's responsibility to determine the cause of death; she must simply find and notify the next of kin. ### Personnel [Figure omitted] There are five distinct roles that are central to a *Bad News* performance: - **Player.** The *player* is the person who is playing *Bad News* for a given performance. In some cases, we allow for multiple players, but generally it is intended as an intimate immersive experience for one player. - **Guide.** The *guide* is a team member who introduces the player to the narrative premise and our gameplay interactions in a brief session preceding a performance. Additionally, in an exhibition setting, the guide serves as a public-facing team member who is free to discuss the project with passersby, as Figure 10.1 shows. This role is fulfilled by codesigner Adam Summerville. [Note: In a few installations, Tyler Brothers has filled in for Adam. Alternative names for his role that Adam has nominated include ‘carnival barker' and ‘den mother', the latter capturing his proclivity for making sure Ben and I are fed and hydrated in what can be eight or ten hour sessions of back-to-back performances.] - **Actor.** The *actor* is a team member who acts out every character that appears in a given performance. Additionally, the actor plays the mortician in an opening sequence (more on that below). This role is fulfilled by codesigner Ben Samuel. - **Wizard.** The *wizard* is a team member who listens in on gameplay via a microphone to manage the experience in real time. His primary duties are executing player commands and discreet communication with the actor. As I will explain below, much of the wizard–actor communication concerns the delivery of narratively potent material sifted out from the simulation—as such, a critical activity of the wizard is *story sifting*. In exhibition contexts, the wizard may perform his duties in a public-facing area, as a kind of performative *livecoding* [collins2011live]. I fulfill the wizard role in *Bad News*. - **Simulation.** The *simulation* is an instance of the *Talk of the Town* framework, which was the subject of the previous chapter. This contains the unique procedurally generated town that will serve as the storyworld for the performance. The town is generated just prior to a performance and then kept in memory, so that the wizard may explore it by constructing the appropriate Python queries. Additionally, aspects of the storyworld are expressed in the actor and player interfaces, as I discuss in the next section. ### Installation [Figure omitted] *Bad News* is an installation-based experience that is driven by a *wizard-of-Oz* [dahlback1993wizard] configuration that integrates physical and technological elements. Figure 10.2 shows a performance at an installation that was built for the Slamdance DIG showcase, held in Los Angeles in December 2016. Our installations have varied from exhibition to exhibition, but there are several critical components that must be in place in order for the piece to be performed: - **Gameplay area.** A *gameplay area* houses a table, two chairs (on opposite sides), and our custom *model theatre*, which is the centerpiece of the installation. On the actor's side of the table is a computer, hidden from the player's view by a fixed lower curtain on the model theatre; this computer displays the *actor interface*, which I will discuss below. On the player's side of the table is a notebook and pen (for taking notes), as well as a tablet computer; the latter displays the *player interface*, which I will also explain shortly. Additionally, a microphone is placed on the table—the actor's computer captures live audio that is conveyed to the wizard's computer over a local area network. - **Wizard command center.** In a separate area, the *wizard command center* is set up. For some performances, this has been a separate room, or even a separate city: using the internet (instead of a local network), we carried out multiple performances at the Slamdance DIG showcase in which I managed Los Angeles gameplay from my living room in Santa Cruz. [Note: In fact, I piped the audio through a sound system, and my wife and I enjoyed each performance in the style of a radio play.] In other performances, this area is public-facing or even on stage before an audience, turning the wizard's activities into a kind of metaperformance based on performative livecoding (and live commentary from both the wizard and the guide); I will discuss the latter configuration at more length below. In any event, this area houses the wizard's computer, and potentially an array of additional screens to display the player and actor interfaces that I manage. - **Partition.** A *partition* separates the gameplay area from public areas. If the gameplay area does not have a whole room to itself, we install some kind of partition that may separate it from a public area with attendee traffic. We do this to maintain the intimacy of the experience—a core design goal—and to ensure that the player does not feel like she is performing for an audience. In some installations, such as the one pictured in Figure 10.2, this partition has worked to both isolate the gameplay area and to separate the gameplay and wizard areas. (Though it is not visible in the image, a curtain from the right edge of the facade to the wall behind the camera could be closed to cordon off the gameplay area.) Additionally, it can be used as a public-facing facade that lends visual intrigue to the installation. ### Experience The *Bad News* experience proceeds in three distinct phases: lead-in, gameplay, post-mortem. In this section, I will provide an overview of what occurs (both in gameplay and behind the scenes) during each of these phases. While each performance tends to total about an hour, core gameplay (in which the player is exploring the town to speak with residents) generally lasts around forty minutes. #### Lead-In The *lead-in* phase begins when a player arrives at the installation for her scheduled performance. At this time, the guide takes her aside to explain what will happen, while the wizard and actor begin to prepare. This all typically takes about five minutes. Specifically, this phase entails the following procedures: [Figure omitted] - **Guide preliminaries.** The guide meets with the player to introduce her to the narrative premise and the game's mechanics. The latter are also captured in a reference sheet, which he provides to her; this sheet contains a list of actions that the player can take during gameplay by simply speaking them aloud (the wizard will hear the command and livecode to execute it). - **Town generation.** The wizard starts up the *Talk of the Town* simulation engine and uses it to generate a storyworld (by the procedures explained at depth in the previous chapter). Once a specific date in August of 1979 has been reached, the procedure generates the *narrative premise* by automatically selecting a character in the town who will be the deceased character in this performance, and then having that character die at that time. To be precise, this selection (and death) occurs at the level of the *Bad News* code layer, which is driven by, but decoupled from, the *Talk of the Town* engine. Here, instead of random selection, the system employs some light reasoning, as it attempts to target a character whom most of the town knows, and makes sure that the selected character is home alone and has at least one relative in the town. [Note: In early performances, the selection was random, which led to a case of a town hermit whom no one actually knew and whose next of kin was an estranged cousin.] Additionally, this set-up procedure identifies the *next of kin* of the deceased, given the definition of it that is operational in *Bad News*. The entire procedure of creating a town and preparing the narrative premise typically takes between two to three minutes. As I noted in the last chapter, it is not feasible to generate many towns at the beginning of an exhibition day, since each takes up several gigabytes of memory. - **Initial story sifting.** Once the town has been generated, the wizard (me) begins to explore the generated simulation material in search of nuggets of dramatic intrigue—this, of course, is the curationist task of *story sifting* (see Section 5.3). In doing this initial story sifting, I isolate a set of *emergent scenarios* of the kind discussed in Section 9.3. Again, these are narratively potent situations that happened to have emerged in the particular storyworld in which gameplay is about to take place. In the case of *Bad News* story sifting, my aim is to uncover scenarios that involve the deceased character in some way, since that person's life is the structural scaffold that will undergird the performance. Just as the randomly selected character Yvette anchored an exploration of a generated town at the Workshop on Tutorials in Intelligent Narrative Technology, as Figure 9.6 illustrates, so too does the deceased character in this case. As such, in my initial story sifting I may encounter a love triangle involving the deceased character, or a secret affair or family drama, or any of the types of scenarios outlined in Section 9.3. Typically, during this period I identify between three and five scenarios, which I deliver to the actor—via our chat, explained next—who then adopts experience-level goals such that these emergent scenarios become conversational stories that he delivers to the player. That is, the actor will attempt to naturalistically reveal these scenarios to the player through conversation during gameplay. Beyond gathering a set of potent scenarios, I will more generally explore the deceased's person life history—family history, work history, social networks, friends, enemies, romantic history, love interests, characters in love with them, and so forth—to get a better sense of who this person was. Again, such excavated material may be delivered to the actor. - **Actor priming.** In the period preceding gameplay, the actor is already stationed behind the model theatre, which has its curtains drawn. As I have noted, the wizard and actor begin communicating during this period via a live web chat (see Figure 10.3). During the process of initial story sifting, the wizard feeds the actor information on the newly created storyworld, as the two prepare for the next performance. This is how the actor finds out about the deceased character, next of kin, town, and anything else that is of importance to the playthrough. Critically, this is where the wizard and actor collaborate to isolate a set of emergent scenarios that the actor will work to naturalistically reveal through conversational storytelling. - **Entering the gameplay area.** Once the player is ready, the guide escorts her into the gameplay area, where she takes a seat on her side of the table. Here, the guide hands her a notebook, pen, and the tablet computer—as he explains, on the latter a *player interface* will display information about her location as she explores the town during gameplay, as Figure 10.4 illustrates. To start, the interface shows an initial prompt that reiterates the guide's explanation about how to take action and asks for the player to speak aloud when she is ready to begin. At any point, the wizard may display arbitrary text onto the screen, either as a means of communicating directly to the player or of describing some aspect of the storyworld that is not typically rendered in the interface text. [Note: A favorite usage of this mechanism transpired in the course of researcher Jessica Hammer's playthrough at the 2016 ACM CHI conference. Upon observing a potential next of kin at a deli, Jessica returned to the home of the deceased character to search for photographs that would show him with someone matching the appearance of the prospective next of kin. While the simulation does not model home interiors—and certainly not artifacts like photographs—this clever move seemed worthy of reward, and so I extemporaneously composed a textual description of a photograph, hanging on the wall, that depicted the deceased character arm in arm with a person having the appearance under question (since Jessica's suspicions about the next of kin were correct). This was risky, since it required salting my hastily composed message with function calls (to insert a physical description of the next of kin), and any typo in this code would have caused the player interface to screw up at an intense moment in the experience. Remarkably, what I typed computed, and the moment worked beautifully. This was perhaps my proudest moment as wizard.] #### Gameplay [Figure omitted] At this point, the player is sitting on her side of the curtain and the actor sits across from her, though she has not seen him yet due to the curtain being closed. The guide has by now left the gameplay area and is likely interacting with exhibition passersby, and the wizard remains out of sight at his command center, where he awaits the player's indication that she is ready to start. From here, the experience proceeds as follows: [Note: For an outline of an example playthrough, see our conference paper on the project [samuel2016bad, pp. 117-118] or Ben Samuel's dissertation [samuel2016bad, pp. 375–378].] - **The death scene.** Once the player says that she is ready to begin, the wizard updates her interface to show a new prompt indicating that she is alone at the *death scene*. Critically, this prompt provides information about the deceased person's appearance and home address—the player will need to use this initial information to determine the identity of this person. Figure 10.4 shows an example prompt. - **The mortician scene.** A few moments after the death-scene prompt is first displayed, the actor is cued by the wizard to begin the *mortician scene*. At this point, he opens the curtain and reveals himself as the county mortician (recall that the player's diegetic role is assistant to the county mortician). The mortician apologizes for arriving late and informs his assistant that he must leave to handle a matter elsewhere in the county, which means that the player must track down and inform the next of kin on her own. Gently, and in detail, he outlines what this will entail, and encourages her to take notes (see Figure 10.10 for an example of player notes taken during gameplay). To respect the privacy of the family of the deceased, the mortician suggests that he and the player collaborate to come up with a *cover story* for her, since it would be tactless to openly parade around the town as a mortician's assistant. [Note: We've found that having a cover story—on top of the narrative framing of being the mortician's assistant—serves as a sort of scaffolding that helps players who are uncomfortable with role-playing to improvise more readily. Example cover stories include a county historian, genealogist, private investigator, lottery representative, and more.] Additionally, the mortician asks the player to participate in a brief imaginary conversation with a town resident, as practice. For this, he draws the curtain of the model theatre, pretends to be a person hearing a knock on his door, and then opens the curtain to engage in conversation with the player. Finally, the mortician fields any last questions before closing the curtain to exit the scene; from here, core gameplay begins. The mortician sequence is the only scene in *Bad News* that has a scripted structure, and it is intended as a sort of diegetic tutorial that onboards the player for our peculiar mode of gameplay (and particularly the kind of role-playing and improvisation that it entails). It can take anywhere between five and fifteen minutes, depending on the initial comfort level of the player—we extend it until the player appears to be ready to proceed—or other various concerns. [Note: In a recent personal communication (June 14, 2018) in which I asked about his favorite *Bad News* experiences, Ben Samuel recalled a notable mortician scene: “Another memorable playthrough involved Tarn Adams of Dwarf Fortress fame, whose playthrough was perhaps most notable for having a thirty minute long mortician scene, when the typical duration averages around five minutes. In addition to the usual conversation pieces centered around introducing the player to the nature of speaking with an improviser and helping them to develop their cover story, this mortician sequence included philosophical musings on the nature of life and death, and the unique role our profession has in bridging these two disparate states of being.”] - **Gameplay in three acts.** I loosely conceive of *Bad News*'s core gameplay sequence (in which the player explores the town to speak with its residents) as having an experience arc with a three-act structure: - **Act I: Learning the name.** Players typically spend the first five or ten minutes of core gameplay working to learn the deceased person's name. [Note: Note that the wizard and actor can coordinate to provide this information more easily if a player is having a hard time finding the name.] Here, there are a number of strategies that players have taken: do a reverse look-up in the city residential directory, using the person's address; [Note: This strategy was first employed by Ben Spalding, who at the time was affiliated with the Expressive Intelligence Studio as a student in UC Santa Cruz's Digital Arts and New Media MFA program. Ben was our most frequent early playtester and provided invaluable insights that guided an initial series of design iterations. This particular strategy was innovated as part of Ben's individual exploration of what *high-level play* [treanor2011burgertime] would mean in the unlikely context of *Bad News*. From our standpoint, his *metagaming* [boluk2017metagaming] led to the identification and development of a series of levers with which we could exert control over the experience.] inquire with a neighbor (using the cover story); head to a hot spot in town, such as a bar; and many more. To take an action, such as viewing the city directory or navigating to a new location, the player simply says what she'd like to do out loud—the wizard is always listening in, and it is his job to promptly update the player interface after each command. Fig. Figure 10.4 shows the player interface after a player has navigated to a bar. Wherever she goes, the player is free to engage nearby characters in conversation; again, this is done by speaking aloud. Whenever she does this, the *actor interface*, shown in Figure 10.5, updates to show information about the character with whom she is starting a conversation. At this point, the actor scans the interface to learn that character's personality, history, and life data, and then spends a few moments getting into character (consciously adjusting his stance, mannerisms, intonation, and so forth accordingly). Once prepared, the actor opens the curtain to play that character live. Throughout a given portrayal, the actor must constrain himself so as not to contradict the character's personality, beliefs, or life history—this requires glancing at the actor interface and the web chat with the wizard as needed, which Ben does with expert discreetness. [Note: In fact, several players have been surprised to discover, after gameplay, that Ben even had a hidden screen on his side of the theatre.] The actor is allowed, however, to *augment* the simulational material with additional detail. [Note: For instance, in one generated town, the player encountered a character drinking alone at a bar—a house painter in his mid-twenties with few friends in the town. Though the simulation does not model character ambitions, the actor augmented the character's base simulational material to improvise a scene about him dreaming of leaving his small town to become an artist in New York City. When it comes to this kind of narrativization, our motto is *augment but do not contradict*, as I have already noted. I will discuss this example at more depth below.] When a conversation ends, the actor closes the curtain of the model theatre and the wizard waits for the next player command. In his own dissertation, Ben writes at length about the experience of acting in *Bad News* [samuel2016crafting, pp. 369–375]. - **Act II: Tracking down the next of kin.** Once the player has determined the deceased person's name, the remaining work is to determine who the next of kin is and where that person is currently located. This is the phase in which the core experience of *Bad News* begins to open up. During this period, the player must ask around the town about the deceased character and the life that person lived—as such, it is primarily during these conversations that the actor seeks to reveal the emergent scenarios that were isolated as narratively potent during the wizard's initial story sifting. Here, the tone of the piece tends to become more intimate, as the townspeople begin to open up more to the player (and the player begins to open up more in her own improvisation). Through these conversations, the player not only discovers the identity of the next of kin, but more critically she learns about the life that the deceased person lived. Often times, information uncovered during this phase will raise the emotional stakes of the impending notification. [Note: In one playthrough, the next of a kin was an older women who twice prior had married an older man who died of natural causes. After the death of her second husband, she instead married a young man in his twenties—tragically, this man was the deceased character for the playthrough. In mentally preparing for the notification scene, the player spoke aloud to himself about the difficult prospect of delivering bad news to a woman who had already heard it on two prior occasions.] This ‘act' typically lasts between twenty and thirty minutes, and it culminates once the player has determined the location of the next kin and traveled to that spot. - **Act III: The notification scene.** This scene may occur at either a home or a public place—it all depends on where the next of kin happens to be located in the simulation. Often, a player takes a few moments to collect herself before engaging in conversation with the next of kin. Preceding this scene, Act II can vary considerably in tone, depending on the player's improvisation and the particular characters and storylines that she encounters; it often features comedic moments. Act III, however, is always emotionally intense. Players frequently ask the next of kin to sit down (even though the actor is already sitting down in physical reality), or request to take the interaction to somewhere more private in the storyworld. In several cases, the actor and/or player have begun to weep. [Note: In one of the design team's favorite moments, Noah Wardrip-Fruin reached his hand across the model theatre as a gesture of consolation, which the next of kin accepted in a tender scene.] Once the notification has been processed by the next of kin—the culmination of which depends on Ben's performance—the actor closes the curtain and the player interface updates to express that the player has successfully completed her task: “You have successfully notified the next of kin. You have delivered the bad news.” [Note: I should note that it is entirely possible for a player to notify the wrong person. According to the rules of the experience, the game ends when the player tells *anyone* about the death—that is, any divulging is treated as the notification (because otherwise the family's privacy is not respected). While the next of kin has been successfully notified in nearly every single playthrough, there have been a few cases of the wrong person being informed. In an unfortunate testament to *Talk of the Town*'s emergent serendipity, one player duo notified the next of kin of a living character who improbably shared all of the peculiar physical attributes of the deceased, including a noticeable tattoo, birthmark, and scar. A stranger incident concerned a player who successfully tracked down the next of kin but then elected to leave that scene to instead notify the janitor at the deceased character's apartment complex.] [Figure omitted] #### Post-Mortem While core gameplay ends when the curtain closes for the final time, the experience is not yet over. As part of our *contract of care*—a practice in immersive-experience design that entails the development of scaffolding to ensure greater experiencer comfort [machon2013immersive, p. 150][striner2017characterizing,crouch2017mediated]—we carry out a gentle off-boarding process through which the player may be eased back into the real world after her emotionally intense visit to a *Talk of the Town* town. [Note: For more on the *Bad News* contract of care, see Ben Samuel's dissertation [samuel2016crafting, pp. 366–368].] Additionally, in this closing period, the experience converges back onto the simulation procedure that enabled it in the first place: in an *epilogue* scene we use *Talk of the Town* to simulate ahead in time to see what became of all the performance's principal characters. What follows are the final phases of the *Bad News* experience: - **Decompression.** After the curtain has closed for the final time, we allow the player a few moments to collect herself and properly exit the storyworld. This means waiting to enter the gameplay area until the player's body language expresses that she is ready to exit the storyworld and speak to the guide, wizard, and actor. Typically the actor himself will likewise require a few moments to collect himself and return to the role of Ben Samuel. - **Post-mortem.** Once it feels appropriate, the guide and wizard enter the gameplay area and thank the player for playing; the actor emerges at this time to do the same. From here, we ask the player for her thoughts, and broadly discuss the experience, as a kind of informal postmortem. - **Epilogue.** Finally, we escort the player to the wizard command center for a brief *epilogue* scene. Just as *Talk of the Town* can simulate the history of a town up until the summer of 1979, it may also simulate the *future* of a town from that point on. This simply means running the simulation procedure of Section 9.2 again, but this time with a later termination date. [Note: It was not until after fifty or so performances of *Bad News* that I realized the feasibility of doing an epilogue.] Specifically, I simulate up to the summer of 2009—thirty years after the time in which the performance took place—and then explore the emergent material to determine what happened to the principal characters and locales that the player encountered during her visit to the town. Often, many of the principal characters will have themselves died—the simulation will say on which date—but updates can be positive as well (such as a deceased character's business still thriving, or young children in 1979 becoming prominent town figures by 2009). Sometimes, the epilogue is poignant: in one playthrough, the deceased character's brother was the next of kin, and also the town realtor—here, the simulation revealed that after receiving the death notification he had to carry out the sale of his deceased sister's now-vacant home. Unfortunately, however, nothing that the player does during gameplay alters the course of the future of the town. This is because gameplay occurs at a much higher level of detail than *Talk of the Town*'s simulation, which makes it infeasible to impute what the player has done back into the actual storyworld. Nonetheless, however, apophenia is always at work in human interpretation (see Section 9.3), which means players are likely to mistakenly discern ways in which the storyworld was changed by their actions. I will discuss a powerful example of this kind of apophenia in Section 10.4. - **Goodbyes.** Once the epilogue concludes, the player's *Bad News* experience is over. No one had ever been to the town that she visited, and no one will go there again. I close the console window and it disappears forever. ### Variants [Figure omitted] While our installation and core gameplay experience are not variable, in a few cases we have mounted *Bad News* in ways that deviate from some of the above description. In this section, I will describe these variants. #### Augmented Radio Play In a few cases, we have mounted *Bad News* itself into a larger media experience that features a live audience. We call this configuration an *augmented radio play*, and it tends to work as follows. First off, the gameplay area is secluded in a private room somewhere, but the wizard command center is housed on a stage in an auditorium area. In this area, all of the various interfaces are displayed on a large screen on stage: wizard console, actor–wizard chat, actor interface, and player interface. Figure 10.6 shows an example of how this might look. Before we start the performance, codesigner and guide Adam Summerville selects a willing player from the audience, and the guide preliminaries then occur on stage (so that the audience also learns how the experience works). Next, the player is taken to a secluded area that is removed from both the auditorium and the gameplay area. At this point, the wizard generates a new town and the wizard and the guide explain the purpose of story sifting, which is then carried out on stage through livecoding. [Note: In this case, the wizard's livecoding is a kind of performance of its own, a phenomenon about which Nick Collins has written [collins2011live].] As part of this process, we solicit audience members to call out ideas about interesting story material that the wizard may attempt to excavate; in many cases, these audience call-outs have significantly altered the shape of the performance. Finally, the player is led into the gameplay area and the mortician scene begins. Critically, however, the audio being captured by the microphone there is piped into the auditorium, so the wizard and guide and audience members can hear what is happening. In this way, the audience experience resembles that of hearing a radio play, though in this case it is augmented by a look at how the experience works behind the scenes. #### *Cattive Notizie * [Figure omitted] Additionally, there has been another more radical reformulation of *Bad News*, but one that takes a different name. In the course of a visit to the University of Turin in Italy, in October 2017, I collaborated with researchers Vincenzo Lombardo, Antonio Pizzi, and Rossana Damiano to design and mount a *Bad News* variant that we called *Cattive Notizie* (Italian for ‘bad news'). This was the group who had produced *Hot Bread* [damiano2017thinning], the *Talk of the Town*-driven work of curationist emergent narrative that I discussed above. In this experience, the gameplay area was reconstituted as an entire room whose four walls were covered in mounted screens—as such, the result worked similarly to a *CAVE* [cruz1992cave,cruz1995virtual] experience. Instead of a single actor, we enlisted several dozen undergraduates who would each potentially act over the course of a playthrough. To enter a new location in the town, the player carried out the physical action of entering the *CAVE*-like room, where an array of actors stood on numbered circles corresponding to the numbering of nearby characters shown on the player interface, as Figure 10.7 shows. On one of the walls facing the player, we displayed the player interface. To speak to a particular character, the player indicated that she would like to converse with the person of the corresponding number, and the two positioned themselves on markers specifying where exactly to stand for the conversation. From the vantages afforded by these positions, the actor could see the actor interface displayed on one wall, while the player could not (though she could still see her player interface). Aside from these alterations, the experience worked in the same way, except that I modified aspects of the *Talk of the Town* simulation instance so that each storyworld would be rigged to be an Italian–American town. While the project was very experimental and the actors did not have enough time to develop the peculiar set of skills that such a performance requires, the results were highly intriguing. Generally, I am very excited about the prospect of a *Bad News*-like piece that distributes the experience across a more extensive physical space with multiple actors and multiple players, perhaps even employing multiple wizards. #### A New Troupe and the Home Version Finally, I would like to note the ongoing efforts of Jonathan Lessard and performer Noah Drew to mount *Bad News* on their own, with Lessard as wizard and Drew as actor. This project is still in a preliminary phase, as its team takes on the challenges of learning how to both interact adeptly with someone else's simulation engine and also develop a rapport in the peculiar area of computationally assisted performance. All of the *Bad News* codesigners are very excited about the idea of *Bad News* being performed by people who are not us, and indeed this project relates to our long-held ideas about a so-called *home version* of the experience that would enable anyone to put it on informally with their friends as a kind of computationally driven *tabletop storygame* [reed2017changeful,white2018tabletop]. In his dissertation, Ben Samuel briefly describes our ideas for a home version [samuel2016crafting, p. 383], which Adam Summervile is now planning to pursue as part of his early research agenda as a professor at Cal Poly Pomona, a position he has just started this fall. ## A Curation Example The following example of *Bad News* curation transpired in the course of a performance at the Fort Mason Chapel in San Francisco on September 30th, 2018, where *Bad News* was installed as part of *Come Out & Play San Francisco*, a festival of public games. In this installation, the gameplay area was a private room closed off from the rest of the chapel, and so I carried out my wizard duties from an entirely separate area; like in all our performances, I could still listen in via a microphone in the gameplay area. For this particular playthrough, we recorded audio and also saved copies of the actor–wizard web chat and wizard terminal session log; the following quotations originate from those sources. None of them are edited in any way, though of course they are excerpts taken from the larger experience duration. In these quotations, the speaker of each utterance is indicated by a prefix and excerpts from my wizard livecoding are formatted as Python code listings; in the listings, commands are preceded by ‘>{>}>' and the lines following a command show the returned values. #### Initial Story Sifting As I explained in Section 10.2.4, a *Bad News* performance begins with the wizard (me) carrying out a process of initial *story sifting*, whereby a set of narratively potent emergent scenarios may be identified for delivery to the actor. Before I can do this, however, I need to find out who the deceased character and next of kin are, and then relay that information to Ben. At the start of a new performance, neither of us know anything about the storyworld, because it did not even exist until that moment. In this performance, a young deli cashier named Ronald Zeise has died, and there are a record *twelve* next of kin: [Note: As I noted above, in *Bad News* the next of kin is the closest living familial relation in town, and if there are several characters related to the deceased character by the same degree, then each is treated as a next of kin. Notifying any single one of these individuals suffices. In the code examples in this section, the variable ‘d‘ is bound to the deceased character, while ‘nok‘ is bound to a list containing all of the next-of-kin characters.] ``` >>> print d Ronald Zeise, 1954-None >>> print d.occupations [, ] >>> print d.occupations[0] Cook at Zilencio 1973-1975 >>> print d.occupations[1] Cashier at Steele Creek Avenue Delicatessen since 1978 >>> >>> >>> print nok [, , , , , , , , , , , ] >>> len(nok) 12 >>> for n in nok: ... print d.relation_to_me(n) ... cousin cousin cousin aunt cousin cousin cousin cousin cousin cousin aunt uncle ``` This was a remarkable number, and in fact it was the first case of a deceased character with extended family as the next of kin—more typically, this group comprises fewer and more immediate family members. Having sifted out this pertinent information, I hurriedly relay it to Ben via our actor–wizard web chat: > Wizard: d is Ronald Zeise, 1954- Wizard: Cook at Zilencio 1973-1975 Wizard: >{>}> print d.occupations[1] Wizard: Cashier at Steele Creek Avenue Delicatessen since 1978 Wizard: WHOA Wizard: 12 NOK Actor: WHAT At this point, the player is still in a preliminary session with the guide, who is explaining how the experience works. Upon further hasty exploration, I encounter an interesting nugget—the deceased's great-great-grandfather was a founder of the town who had worked at night as the cemetery groundskeeper for nearly fifty years: ``` >>> print d.father Edward Zeise, left city in 1973 >>> print d.father.father Robert Zeise, 1885-1954 >>> print d.father.father.father None >>> print d.mother Carol Zeise, left city in 1973 >>> print d.mother.father None >>> print d.father.mother Hattie Horsman, 1892-1961 >>> print d.father.mother.father Henry Horsman, 1840-1909 >>> print d.father.mother.father.father Walter Horsman, 1817-1886 >>> print d.father.mother.father.father.occupations [] >>> print d.father.mother.father.father.occupations[0] Groundskeeper at Mosses Cemetery 1839-1883 >>> print d.father.mother.father.father.occupations[0].shift night ``` As I am looking this up, Ben inquires about the deceased character's family and in my response I communicate that this emergent scenario could make for an interesting story for Ben to deliver conversationally at some point: > Actor: what happened to parents Actor: clearly no spouse Actor: no siblings I gues? Wizard: they left town in 1973 Wizard: siblings left too Wizard: they were younger Wizard: they are all descendents of Walter Horsman, 1817-1886 Wizard: founder Actor: nice Wizard: groundskeeper of the cemetery 1839-1883 Wizard: i like a story about him being a creepy guy? Actor: coooool Actor: yeah Wizard: it was night shift Actor: perfect Wizard: old man Horsman Wizard: lol Actor: founded the town Actor: just so he could be a gravekeeper Wizard: hahaha yes Wizard: love it Shortly after this exchange, the player is led into the gameplay area and the mortician's scene commences. This means Ben is not available to continue chatting, but this scene does not require maintenance of the player and actor interfaces, which allows me to continue story sifting without hindrance. In the course of searching more through the emergent material pertaining to the deceased's ancestors, I encounter a remarkable case of a restaurant named Zilencio that had operated for 131 years but recently had shut down (recall that the year is 1979): ``` >>> print list(d.ancestors)[2].occupations [, ] >>> print list(d.ancestors)[2].occupations[0] Waiter at Zilencio 1850-1875 >>> print list(d.ancestors)[2].occupations[1] Farmer at Thomas Dowey's farm 1875-1894 >>> print list(d.ancestors)[2].occupations[0].company Zilencio, 317 Steele Creek Avenue (1844-1975) >>> zil = list(d.ancestors)[2].occupations[0].company >>> print zil Zilencio, 317 Steele Creek Avenue (1844-1975) ``` I now become curious about the circumstances of Zilencio's recent closure, so I sift further to see who owned it last: ``` >>> print zil Zilencio, 317 Steele Creek Avenue (1844-1975) >>> print zil.owner Proprietor at Zilencio 1973-1975 >>> print zil.owner.person Douglas Salva, 31 years old ``` Apparently a character named Douglas Salva was Zilencio's final owner, and he had only been in charge for two years before its remarkable closure. Digging deeper, I learn that Salva had originally been a waiter at the restaurant, but was now a stocker at a department store: ``` >>> doug = zil.owner.person >>> print doug Douglas Salva, 31 years old >>> print doug.occupations [, , ] >>> print doug.occupations[0] Waiter at Zilencio 1966-1973 >>> print doug.occupations[1] Proprietor at Zilencio 1973-1975 >>> print doug.occupations[2] Stocker at Flom Department Store since 1977 ``` I wondered why a waiter would become the owner of a restaurant, since I know that in-house promotion in *Talk of the Town* tends to elevate employees with a occupational status immediately below the open position (in this case, a restaurant manager). As I explained in Section 9.2.7, however, in addition to favoring current employees for open positions, characters also prefer to hire family members, and so candidates meeting both criteria score highly in the utility procedure that drives the hiring process. As such, I now suspect nepotism, and my sifting confirms this: ``` >>> print doug.occupations[1].preceded_by Proprietor at Zilencio 1964-1973 >>> print doug.occupations[1].preceded_by.person Henry Ereth, 69 years old >>> henry = doug.occupations[1].preceded_by.person >>> print henry.relation_to_me(doug) nephew ``` More intriguingly still, I find a connection to the deceased character. Apparently the late Ronald Zeise was a cashier at a deli that now stands on the former site of Zilencio, and moreover he had actually been a cook at the latter during the brief period of Salva's ownership: ``` >>> print zil.lot A lot at 317 Steele Creek Avenue on which Steele Creek Avenue Delicatessen has been erected >>> print d.occupation Cashier at Steele Creek Avenue Delicatessen since 1978 >>> print d Ronald Zeise, 1954-None >>> >>> >>> print d.occupation Cashier at Steele Creek Avenue Delicatessen since 1978 >>> print d.occupations [, ] >>> print d.occupations[0] Cook at Zilencio 1973-1975 ``` This seems like a potent emergent scenario, especially given the connection to the deceased character, and so I relay it to Ben with the suggested narrative spin that the town is angry at Salva for killing a beloved institution: > Wizard: there was a town instutition Wizard: Zilencio, 317 Steele Creek Avenue (1844-1975) Wizard: that closed after 130 years Wizard: b/c of this failure Douglas Salva, 31 years old Wizard: who now works at dept store Wizard: as stocker Actor: Who was Salva before? Wizard: whole town is mad Wizard: owned Zil #### Curation at Work By now, the mortician scene has ended and the player leaves the death scene and moves into the lobby of the deceased character's apartment complex. Here, he initiates a conversation with a janitor there named Clarence and explains that he is a restaurateur who is working to open a dinner theatre in the town. After a few minutes, the player asks if the janitor knows any town lore: > Player: One of the things that really puts a town on the map for us is a sense of local lore. Local legend. We had a really successful restaurant for a while, in another state. One of the reasons it was so great is because the town had its stories. Actor: Yeah, uh-huh. Player: You know, it wasn't all pretty. We did have the, you know, we had the feud between, you know, the different families. And I was wondering if you felt like this was a candidate town for that and, uh, if you could tell me anything. This is a perfect opportunity for Ben to impart the two emergent scenarios that I have uncovered—the creepy cemetery groundskeeper Walter Horsman and the late town institution Zilencio—and so I quickly remind him in the chat by hastily copying and pasting information from my Python terminal: > Wizard: Zilencio, 317 Steele Creek Avenue (1844-1975) Wizard: Douglas Salva, 31 years old Wizard: >{>}> print zil.owner Wizard: Proprietor at Zilencio 1973-1975 Wizard: Walter Horsman, 1817-1886 Wizard: Groundskeeper at Mosses Cemetery 1839-1883 Ben then launches into a brilliant turn of conversational storytelling that (as I will explain shortly) nicely demonstrates our narrativization motto, *augment but do not contradict*: > Actor: Yeah, we got a couple stories. I got two stories for you. Player: Okay. Actor: I've got an old story and I've got a new story. Player: Great. Actor: Old story first. Player: Okay, let's do old. Actor: Okay. The town was founded, I don't know, hundred years ago—hundred fifty years ago—something like that. Guy who founded it, rumor has it, the reason why he wanted to found the town in the first place—because he just wanted a place where he could be a grave keeper. Player: Huh. Actor: He founded the town. Player: A grave keeper. Actor: He could have any job he wanted. Next thing, very next thing, starts up a cemetery, becomes the undertaker there. Player: Can't imagine he'd have a whole lot of business to begin with. Actor: [laughs] Player: I guess if you build it they will come. Actor: I suppose so. Or maybe he had some more sinister plans in mind. Maybe he wanted to accelerate the cemetery a little faster. Maybe he built the town to attract flies into his trap. Player: That certainly is a chilling thought, Clarence. Actor: [laughs] Ah, I mean, it sounds crazy, right? No way. No way that can be true. [laughs] But we still, we still talk about him! Walter Horsman is his name. Player: Walter Horsman? Actor: Old Man Horsman. I mean he's long dead. [laughs] His body in the very cemetery that he started. Player: Has the cemetery become quite liked by, uh— Actor: Yeah, yeah, it's doing okay. Yeah. [sighs] Old Man Horsman. Player: I have to say that I sort of understand the desire to devote one's life to the next world. Actor: What? What do you mean? Player: I just find it interesting that, you know, someone can be alive at one minute and the next thing you know, they're gone. Like you just told me. Actor: I feel like you just summed it up. Player: Yeah. Actor: I feel like that's about all there is. Player: That's about—that's life! I agree. Through the delivery of this story about the creepy gravedigger, Old Man Horsman, a procedure of curation has been consummated. Critically, this procedure has worked according to the curationist architecture introduced in Section 5.3. First, the history of a *simulated storyworld* transpired. This was a *Talk of the Town* town named Mosses, founded in 1839, and one of its settlers was a man named Walter Horsman who came there to work in its newly established cemetery as a night-shift groundskeeper. Horsman then proceeded to live out a life in the town: he worked the same job for the next forty-four years, married and had children, and then passed away in 1886. (Incidentally, the restaurant Zilencio was opened in 1844, which means he likely visited it many times over the course of his life.) Over the next century, his children had children who had children and so forth, and one of his great-great-grandchildren was a man named Ronald Zeise who died suddenly in the summer of 1979. All of this information was recorded by *Talk of the Town* in the form of stored data that could be queried in an interactive Python session—in curationist terms, this stored data is a *chronicle*. In my capacity as wizard, I then began working as a *story sifter* to excavate material capturing all the information that I have just outlined. I encountered this particular material by working backward through the deceased character Ronald Zeise's family tree, and I decided to sift it out because the notion of a night-shift gravedigger is evocative, especially in the context of *Bad News*. Next, I packaged up this sifted material and passed it on to the actor, Ben Samuel, who worked as a *narrativizer* to turn the raw material into an actual narrative artifact that is delivered through conversational storytelling. Through this delivery, the narrative artifact was *mounted* in a full-fledged *media experience* that was then encountered by an *experiencer*. In this case, this experience was *Bad News* and its experiencer was the player, who engaged with the mounted narrative artifact through conversational interaction in the context of the experience. #### Augment But Do Not Contradict In Section 10.6, I will again outline the curationist configuration that underpins *Bad News*, but for now I would like to include a few more examples of curated emergent stories that were also mounted in the course of this particular performance. The following exchange begins right where the last one left off, as part of the same conversation with the janitor in the lobby: [Figure omitted] > Player: So what about this new story? Actor: Ah, okay, okay. Well, the tension is still a little high around town about this. So you might want to be a little careful about who you talk to. Ah, have you heard of Zilencio? Player: Zilencio? No, I haven't. Sounds promising. Actor: Uh, yeah. It was an institution. Player: Oh. Actor: It's been a part of our town since as long as I can remember. I remember my parents and my grandparents telling me about it. It's gone. Player: Really? What kind of institution was that, Clarence? Is it an insane— Actor: Oh, no, no, no. By ‘institution', I mean a—by ‘institution' I mean a—what's the word? Landmark. Player: Oh, landmark, oh. A place that people knew. Actor: Yes. Yeah, yeah, yeah, sorry about that. Player: A destination, if you will. What was it? Was it a business? Actor: Oh, restaurant. It was a restaurant. So you—you, doing your research—are gonna wanna hear about this. And specifically, you're not gonna wanna hire Salva. Player: Salva? Is Salva one one of the staff there? Actor: Douglas Salva. Player: Okay. Actor: Waiter there for ten years. Became a proprietor of it—*the* proprietor of it—and within two years this destination, if you will, closed down. Player: Really? Actor: Gone. Player: Why, what happened? Just poor management? I know that feeling. Actor: Yeah! Yeah, you know what? You know, Douglas might want to talk to you. You know, he might be able to learn a thing or two from you. I don't, I don't know anything. And frankly, I don't want to talk to the guy—everyone's pretty mad at him. Again, the same procedure as above has been carried out: the raw transpiring of a storyworld has been captured in a chronicle, which I sifted to extract material that was delivered to Ben, who narrativized that material to build a story, which he then mounted into the experience through conversational storytelling. Note that while Ben's story *augments* the raw source material—data expressing that there was a restaurant that operated for 131 years until it was closed shortly after being taken over by a character named Douglas Salva—it does not *contradict* it. This distinction is critical with regard to the artistic orientation of the curationist framework that is the subject of this dissertation. As I have argued at numerous points in this dissertation, following the ideas of Hayden White and Arthur Danto and Espen Aarseth and others, stories themselves cannot manifest in records or in the raw transpiring of a simulation. Rather, emergent stories come to actually exist through a procedure of curation. Because this procedure is constructive, it is therefore also augmenting. In fact, augmentation is unavoidable in the narrativization step, if only for the simple reason that narrativization requires the construction of what Arthur Danto has called *narrative sentences* [danto1962narrative]. As I explained in Section 3.1.1, a narrative sentence is one that packs descriptions of multiple events (and their interrelations) into a single sentence, such as the canonical “The Thirty Years War began in 1618” [danto1962narrative, p. 155]. Even the construction of a narrative sentence is an augmenting act, since crafting one requires the assembly and aggregation of multiple chronicle entries into a single unit of content. This means that narrativization is inherently steeped in a process of augmentation, and as such, augmentation is not treated as an offense in the curationist framework. In emergent narrative, a simulated storyworld will always be impoverished in its representation (due to the ontology of simulation), and so if a curated story is to feature the level of detail that is typical of narration, augmentation likely must occur. The key, however, is to not *contradict*: when the facts of a chronicle are contradicted, the history of the simulated storyworld as it actually transpired is polluted, and narrative intervention has occurred thereby. As I argued in Section 5.1, intervention kills the aesthetics of emergent narrative, and so contradiction should be avoided in the narrativization process. [Note: To be clear, I think it could be fruitful to deliberately model a *subjective narrativization* that *does* contradict the storyworld, but in ways that are specifically targeted to a subjective storyteller. While an example of this does not appear in the transcripts printed in this section, a *Bad News* character may embellish in a story, or even tell outright lies, if doing so makes sense given her relation to the people and events being recounted. This entails narrativization that contradicts the simulation, but in ways that deliberately target a subjective telling. Moreover, such subjectivity will be readable to the player in that she is likely to assume that not all the stories she hears are true, as in the real world. To name another example, my original dissertation topic concerned characters who would generate self-serving conversational stories recounting past events. But in a project like *Sheldon County*, a generative podcast about the emergent events of a uniquely simulated storyworld, I would not target a subjective telling because its stories are told from the vantage of an omniscient narrator. If its underlying system, *Sheldon*, were to contradict the storyworld through its narrativization, the project itself would be undermined—what is the point of simulating a storyworld if the telling will rely on artifice and invention, rather than the actual material generated by that world? Indeed, when contradiction is permitted, there is no point to taking the approach of emergent narrative, and in fact, the result will not even be a work of emergent narrative. To be clear, this does not mean that such a method is inferior, but rather that it is something entirely different. As I have already explained, this thesis is intended primarily as an art manifesto that promulgates a fiercely emergentist approach to procedural narrative. My value-laden language is meant to instantiate this artistic orientation, not to cast judgement on intellectual persuasions or technical methods that differ from mine.] Thus, while Ben's narrativization augments the sifted source material to inject flavor and detail, his process is still fully in line with the curationist approach. In fact, his method exemplifies the very kind of narrativization that I now seek to do automatically, starting with *Sheldon County*. #### The Story Retold Following his conversation with the janitor, the player inquires with the landlord of the apartment complex and asks for directions to the Steele Creek Avenue Delicatessen. As I noted above, the late deceased character worked at this deli, which was incidentally on the site of the former town institution Zilencio. Here is an excerpt from this conversation (note the parenthetical wizard chat messages that I have injected at approximately the points at which Ben would have received them on his hidden actor interface): > Actor: I think you might be able to find your way there. It's, uh, well it used to be one of the most popular places in town. [laughs] Player: I heard it closed recently. Are you talking about Zilencio? Actor: Yeah. Steel Creek has yet to prove itself. Give it another 130 years! Player: Wow, well I haven't heard all the stories, but I sure would love to hear that story. Actor: Have you heard about Doug Salva? Player: Douglas Salva? I heard he was the proprietor. Actor: Yeah, yeah. If you could call it that. Player: Not that I'd want to probably hire him for a restaurant that I'd create, would I? But I've heard there's some, uh, controversy about him. About how he ran the place. Actor: Don't think there's a lot of controversy. Uh, the place existed and now it doesn't. Pretty, uh, open-and-shut case. Player: Ah, I see. Well, do you think I could talk to Douglas? I sure would love to include his story in, you know, a piece I'm putting together. {[}Wizard: Stocker now] {[}Wizard: at Flom Dept Store] Actor: [laughs] Yeah, yeah. Uh, you can probably find him— {[}Wizard: LOL] {[}Wizard: he lives there!] {[}Wizard: unit #7] Actor: Well, there's two places you might be able to find him. The Flom Department Store had the decency to take him in after, you know, he suddenly found himself in need of a job. Uh, but the sad sack actually lives right here. Unit #7. You might be able to find him. Player: Oh, really? Actor: Yeah. {[}Wizard: he's at home] {[}Wizard: works nights] Player: Wow, Douglas Salva lives right here? Actor: Yeah. Don't remind me. I have people picketing the place more often than I care to count. Here, we find a case of Ben curating the same sifted material into an entirely new narrative artifact, which is likewise mounted into the experience through conversational storytelling. This is achieved by retargeting the narrativization procedure so that the result is a conversational story delivered from the perspective of, and in the style of, a different character than before (the landlord, as opposed to the janitor). While he does not contradict the storyworld—in fact, he references the additional detail that Salva now works at the Flom Department Store—there has been a retargeting to the variant perspective of the landlord character. Relative to the janitor, this character's personality profile led to a more brusque performance, and the telling differs accordingly. Also in this example, we see what might be called *just-in-time story sifting*: a primary task of the wizard is to provide the actor information that is not expressed on the actor interface but is needed in the conversation. Ben is very good at parsing these messages quickly and naturalistically—many players have been surprised to learn that he relied on a hidden screen—but I still try to keep these messages to a minimum. #### The Man Himself As his next move, the player buzzes Douglas Salva's apartment. Before this critical scene starts, I send Ben some additional sifted material to guide his performance (and his narrativization): > Wizard: okay you're doug Wizard: very open Wizard: very neurotic! Wizard: you were a great waiter from 66-73 Actor: gonna invite him in ok Wizard: okay Wizard: Ronnie is a neighbor Wizard: very high charge Wizard: uncle gave you job Wizard: after you were greater waiter for decade Wizard: Henry Ereth, 69 years old Wizard: your great uncle owned it before that Wizard: for 30 years Wizard: henry ereth Wizard: before that great great uncle owned Wizard: it's a family business Wizard: uncle Henry never had kids, left it to you Salva answers the buzz, but does not say anything, and so the player speaks up to ask if anyone is there, and a conversation transpires: > Player: Uh, Mr. Douglas Salva? This is Winston— Actor: Just—I don't, I don't wanna talk to anyone, okay? Leave me alone. Player: Oh Mr. Salva? I, um, I must have misrepresented myself. I'm a restaurateur, um, looking to start up a restaurant. I actually wanted to make a proposal for you. Actor: You don't, you don't want me near your restaurant. Is this some kind of joke? Player: No, I cannot be serious enough when I tell you that this is no joke whatsoever, Mr. Salva. Um, do you think I could have the opportunity to, uh, to introduce you to an idea. If you don't like it, you can send me on my way. This is not a joke and I am not here to humiliate you. Actor: [sighs] Okay. Yeah. Yeah, why not. Uh, uh, come in. Come on in. Salva buzzes the player into the apartment (signaled by Ben drawing open the curtain of the model theatre) and Ben proceeds to curate the emergent Zilencio material once more, but this time from the dejected perspective of Salva himself (and using the additional material that I sent to him over the chat): > Player: Is this Mr. Salva? Actor: Call me, call me, call me Doug. Player: Winston Delacroix. Actor: Hi. Player: Hi. Hi. I'm from Newtown. And I would like to make a proposal to you. I do appreciate you letting me in. I hear there's, um, quite a lot has gone under the bridge in town recently. Um, but, as my pappy used to say, there's nothing like learning from one's failures to get up and, um, and do a great job. Um, one of my business school professors once told me, if I haven't failed before—if I haven't messed up—then he's not going to listen to me, because it's only through mistakes and failure that we actually learn how to proceed. Actor: Yeah, look. Look. Thank you. Thank you. That means a lot to me. It does. It really does. Um, I feel your, your teacher who told you that in a, in a nice classroom—like, that sounds really inspirational. That sounds, that sounds really great. Um, imagine, if you will, that the failure is at a restaurant that you've been a waiter at for ten years. The best waiter they've ever seen. A waiter so good they made you the boss. They made you the boss. [sobbing] Everyone loves you. Everyone loves the restaurant. [bawling] And I did my best. I'm sorry. I'm sorry. I'm, I'm—you're the first person that hasn't yelled at me in so long. I'm so sorry. I'm so sorry. Player: Mr. Salva, you need not apologize for one instant. Someone with your mettle and your experience is exactly the kind of person we're looking for. Someone who can really feel and understand what can happen. Actor: [sobbing] Yeah. Player: Now, this might sound bizarre to you. This might sound strange. But we are looking for someone exactly like you to help us with the new restaurant we want to start. And I know—I know that's gonna sound strange—but one thing we need to start with is honesty. So I need you to tell me in about a sentence or two what went wrong. And then when we can meet on that level, then we can take it to the next step. Because I believe it's only by admitting—it's only by putting it out there what happened—that we can have a true bond. Would you do that with me, Mr. Salva? Actor: [sobbing] Yeah. Player: Okay, excellent. Tell me what happened. Actor: It's easy. I, uh, I rose above my station. Simple as that. I thought that like, you know, I just knew it all inside and out. I thought, I thought that I could make a difference. I thought that I could take this great place that I loved so much and just make it even better. But I was naive. I was overly optimistic. I lacked any formal business training. I clearly had no intuition for it. I convinced the powers that be—I convinced my uncle—to give me the job. And I wasn't ready. I didn't know what I was doing. Player: Mr. Salva, so I do follow along: who is your uncle? Actor: Huh? Player: Who is your uncle? Actor: Um, um, Henry. Player: Henry, okay. Is that—okay, I'm sorry, keep going. I didn't mean to slow you down. Actor: I thought that I could—you know, he meant so much to me, the restaurant meant so much to me—I thought that I could follow in his footsteps. Keep it in the family. I just didn't know what I was doing. Um, I wish, I wish that wasn't the truth. I wish that I could tell you that, you know, people didn't believe in me, and because of that they weren't working hard for me. I wish I could tell you that there was some problem with the economy and that all of the restaurants were suffering. No, it's nothing. Everyone believed in me. The economy was fine. It was me. Player: So what happened? Actor: I just, I spent too much. I didn't—I spent money on—I spent money on I don't know what you would call it. See, I don't know what you would call it. Marketing, I guess. Advertisements. I tried, I tried to change the feel of the restaurant. I wanted—people already had such high respect for it—I wanted to make it classier. I wanted to make it highfalutin. Player: Understood. I understand now. Actor: I had, I had ice sculptures. I had these giant statues. I imported pillars. Player: Zilencio was the name of the restaurant, as I understand it. Actor: Yeah. Player: You did not hold back, did you? And what did your uncle think of this, when you were spending all this money? Actor: He never, uh, he never told me. He just said, uh, you know, I trust you. Player: Do you have any other family that were impacted by this? Actor: I mean, I think, in general, the town has been understanding enough that it's my fault—and my fault alone—that it happened. I don't think my family has been under fire, as it were. Thank goodness. It could be worse. From here, the player proceeds to concoct a proposal for a new restaurant in town that would actually be themed around the redemption and resurrection of Douglas Salva: “I should tell you the restaurant name we're thinking of, or playing with, is ‘The Rise of the Phoenix”'. [Note: Of course, this was not a real plan, since the player character was in actuality a mortician's assistant, not a restaurateur. In the epilogue following gameplay, we discovered what became of Douglas Salva in the following decades, as I will explain in the next section.] Through discussion of whom to invite to a press event, the player cleverly discovers the identity of the next of kin. #### The Importance of Curation In this way, the story material that I sifted out of the chronicle prior to the start of the performance was curated into a story that not only was mounted in the experience, but moreover came to form its bedrock. The Zilencio story became the structural scaffold of the entire performance, one whose centrality actually made the deceased character into something more like a side character. Recall, however, that this material was only encountered through an improvisational exploration of the storyworld's history. As the terminal excerpt above showed, I was serendipitously exploring the work history of one of the deceased's characters ancestors, when I encountered Zilencio and its remarkable duration of operation. This was only one of the storyworld's emergent scenarios, though, and I could just as well have sifted out material pertaining to another, which in turn would have utterly changed the course of the performance. [Note: As I noted above, in our so-called *augmented radio play* configuration, we encourage audience members to call out for ideas as to what to look for in story sifting. In this way, the audience becomes a collaborator in the curation procedure, and in some cases audience suggestions have significantly altered the experience.] As such, this is a clear expression of the importance of curation in *Bad News*, as well as the tendency of the underlying *Talk of the Town* simulation engine to produce narrative cornucopias, as I explained in the last chapter. One consideration that we have often discussed, and that players have likewise brought up, is all of the myriad emergent scenarios that we never even come to know about in the course of a performance. Who knows what other intrigue lied hidden in the town of Mosses? Indeed, this is part of the appeal of automatic curation: a computerized story sifter could sift through *all* of the material stored in a storyworld chronicle to identify not only the most promising nuggets, but also the inscrutable connections that serendipitously bind those nuggets. In Chapter 12, I return to the notion of automatic curation, which I am now exploring in *Sheldon County*. ## The Pleasure of *Bad News * What is the pleasure of *Bad News*? In this section I will attempt to answer this question, partly through the utilization of the intellectual framework that I developed in Chapter 3 (to appraise the very form of emergent narrative). Additionally, I will argue for the pleasure of this work in a more essential way, by appealing to its reportable successes, including awards and favorable media attention. In the case of my appraisal of *Diol/Diel/Dial*, this was not possible—since it appears that nobody but me has experienced that piece, which should itself be viewed as a damning appraisal—but in my view this is the strongest form of evaluation for projects in the area of computational media. I think it is worthwhile to make purely technical contributions, but my personal aim is to develop compelling media experiences, and the only way to evaluate such experiences is to release them into the world to see what others think. #### Evidence of Success Let us start with the success of *Bad News*, which suggests that it does indeed give pleasure (that I will in turn attempt to unpack in the remainder of the section). First, *Bad News* has been installed internationally, at a number of venues, and in that course it has won awards. One of our first installations was at the 2016 ACM CHI conference, where we took home the Innovative Game Design award in the student game competition. Later that year, the project was accepted into IndieCade, the foremost international festival for independent videogames, where *Bad News* won the Audience Choice award. Additionally, we have installed the piece at the Slamdance Film Festival, where it was included in the DIG exhibition of digital works, and the San Francisco Museum of Modern Art, as part of an exhibition of *mixed reality* [ohta2014mixed] works. *Bad News* has also been performed outside the United States, following invitations from universities in Montreal, Canada, and Turin, Italy. [Note: Specifically, it was installed at Concordia University, Universite de Montreal, and the University of Turin, though in the latter case this was specifically the experience variant *Cattive Notizie*, as I explained above.] The project has earned favorable critical reception, as well, both in the media and among renowned game designers. Writing for *Rolling Stone*, Steven T. Wright proclaimed, “this marvel of procedural performance can only be played by a lucky few, and that's a crying shame” [wright2017how, n.p.]. Game designer Mattie Brice listed it among her top ten games of 2016 [brice2016top], and designer and scholar Tracy Fullerton called it “a game that shook my world”. [Note: Fullerton said this in the course of a keynote presentation at the 2016 International Conference on Interactive Digital Storytelling, where Ben and I happened to be in the audience.] While performing *Bad News* is gratifying in and of itself—and while we believe it makes technical and intellectual contributions—its various successes in terms of invited and juried installations, favorable reviews, and festival awards are the accomplishments that have made us most proud. In turn, I contend that such accolades provide the strongest form of positive evaluation that can be reported in the field of computational media. #### Actual Worlds and Personal Connections Now that I have established that the project has been successful, I will now attempt to unpack the pleasure of *Bad News* by referring back to the ideas that were introduced in Chapter 3, particularly the *aesthetics of emergent narrative* that I identified there. A critical consideration here is that the aesthetics that are activated in this piece are rooted in features of the underlying simulation, which are in turn exposed to the player through a procedure of curation. As such, the pleasure of this experience is perhaps most fundamentally rooted in its utilization of curation, but I will save that discussion for Section 10.6. First, analogies to both nonfiction (see Section 3.1.1) and lived experience (Section 3.1.2) obtain in this work, which activates the *aesthetics of the actual* and the *aesthetics of the personal*. As I have alluded to above, *Bad News* is about the splendor of everyday life. It is a game about death, but it does not handle death like games tend to, opting instead for a solemn treatment. While ostensibly there is a puzzle at the core of the experience, the notification task is really about examining a life that someone has led and is now leaving behind. As codesigner Adam Summerville noted in a filmed performance at the Slamdance Film Festival, “It's more about examining this person's life and figuring out who they were and who they affected and who is affected by their passing”. [Note: Footage from this performance, including this quotation, is captured in a project spotlight video that was produced by Marie Jamora and Jason McLagan for Slamdance TV [jamora2017spotlight].] This grounding in everyday life (and everyday death) is further emphasized in the naturalistic conversation that undergirds the experience, as opposed to the stylized dialogue of fictional media. In total, *Bad News* has the flavor of nonfiction and, as such, the aesthetics of the actual are clearly at work at the level of its gameplay. Moreover, the underlying *Talk of the Town* character simulation—which we augment but do not contradict—both supports and reinforces these aesthetics. An analogy to lived experience, then, obtains through the player's engagement with the apparently actual world of the town she visits. In his recent recollection of highlight moments, Ben Samuel cited the case of a player who deeply engaged the experience in an extended playthrough: > I love it when the player recognizes the true weight of the responsibility they bear, and allow themselves to be swept up in the emotion of the final reveal. [...] One player, whose playthrough lasted for nearly two hours, made meaningful connections with essentially every person she met, culminating in an emotional final exchange where both player and next of kin were reduced to tears. [Note: Personal communication, June 14, 2018.] Indeed, through interactions with the actor, whose improvisational skill and style of delivery maintain the verisimilitude, a player comes to actually *know* a set of *Talk of the Town* characters; by navigating one of its towns, she truly visits the homes and businesses there. I am claiming this by an appeal to reason, but the experience actually bakes in a mechanism by which this claim may be tested: the *epilogue*. Whereas a procedural town history (explored by livecoding with me) may intrigue an experiencer for being both synthetic and veridical—the aesthetics of the actual—the generated epilogue hits players on a more personal level. In both cases, the encounter is with raw *Talk of the Town* data displayed in a console window (see Figure 9.3), but in an epilogue the interactor has come to *know* the little chunks of data. One player, the researcher Mike Cook, was especially affected by the epilogue of his playthrough, and I wrote to him recently to ask why. Here is what he said: > Bad News' epilogue has stayed with me to this day as one of the most impactful generative experiences I've ever had. Most simulation-based generative systems are impressive in a sort of Sublime-esque way, a feeling of awe derived from the scale of what is happening—that's the first hundred years of Bad News' simulation. The town is formed, businesses are founded, dynasties are formed, it's incredible to know it's happening but there's little at stake. But as I sat down to see the epilogue, in the brief flash as three more decades passed by, this wasn't a random jumble of names and events any more. Caught up in the whirlpool were people I had met, relationships I was invested in, unfinished stories that begged for an ending—it all felt very fragile, and terrifying. I've never been more invested in a generative system than I was at that moment. [Note: Personal communication, June 20, 2018.] In another epilogue—the one following the playthrough that was featured in the last section—a player expressed deep remorse about something he had done in the town. To recapitulate, in this storyworld, a character had become a pariah after tanking a family restaurant that was a beloved town institution. When the player eventually met with this person, who had become a recluse, they engaged in deep conversation and eventually the player proposed a plan to repair the character's damaged reputation and life. While I think the player was sincere in wanting to help the character, the plan was in fact a lie that pertained to the player's cover story—he was really attempting to gain information about the next of kin. In the epilogue, the player and I discovered that the character, Douglas Salva, had remained a recluse who, thirty years later, still worked as a stocker on a department store's night shift. Even though this was illustrated to the player through sanitary computer printout in a console window, Douglas Salva had become real to him, and he expressed a remarkably deep remorse at having caused this entity—*a mere bundle of data*—pain. #### Inexhaustible Narrative Fields I have already discussed at length (in Section 9.3) how *Talk of the Town* exemplifies what I have called the cornucopian approach to emergent narrative, through its *overgeneration* of emergent scenarios that charge each generated town with a considerable narrative potential. When a storyworld feels like an inexhaustible narrative field—as it may for a *Bad News* player who can go anywhere in a town and talk to any of its denizens, each of whom have lived out entire simulated lives—the *aesthetics of the vast* are at work. Furthermore, the *aesthetics of a larger context* are also yielded in that each *Talk of the Town* emergent scenario stands against a backdrop of other ones, and moreover the monotony of everyday existence that constitutes the far majority of a town's history. In *Bad News*, the deceased character works as a kind of narrative center of gravity, and so the emergent scenarios pertaining to that person's life are particularly imbued with connection to a larger context. #### Shared Authorship Distilling insights that we gleaned from discussions with players after gameplay, Ben Samuel writes in his own dissertation about how *Talk of the Town*'s simulation undergirds a sense of agency (citation inserted for clarity): > many players expressed that the experience felt very unique—both by virtue of the fact that this is a gameplay experience unlike most others, but also because the player has free reign to explore a town with hundreds of characters in any way that they choose. Thus players report feeling high senses of agency over the shape of their gameplay session. Players have said that they felt transported to the world [green2004understanding], and were able to readily visualize the people that they spoke with and the places that they visited. Many players found their towns so vibrant that they were shocked to learn during the postmortem that the towns were not designed by hand. [samuel2016bad, p. 380] In this way, and especially through her conversational interactions with the town denizens, the player *coauthors* an emergent story that is built atop the emergent narrative material generated by the underlying simulation. Another recent recollection by Ben provides a nice illustration: > Many of my favorite memories hinge upon the addition of information provided by the player that complemented and augmented the baseline histories of the town's citizens provided by the simulation. At the 2016 CHI conference, a player struck up a conversation with a young man who painted houses, but had aspirations to move to a studio in New York and become a fully fledged artist. This player gave him very practical advice, and showed him how he could begin to pursue his artistic dreams without even needing to leave his hometown. Another player, when he first learned the name of the town, asked the mortician if he had heard about a recent hullabaloo involving a trolley car and some dogs. The mortician hadn't heard of it, and the player seemed disappointed. However, as the player walked around town, many of the folks he spoke with excitedly asked him if he had heard of the recent trolley trouble, to which he always enthusiastically said “yes!” As the player searched the town and found the next of kin, he slowly learned more and more about the story behind the trolley, none of which was actually modeled within the simulation, but that didn't contradict anything established by it either. [Note: Personal communication, June 14, 2018.] As I have mentioned several times by now, Ben Samuel's dissertation presents his framework for *shared authorship*, which concerns experiences in which system and human collaborate to craft a narrative artifact that neither could have developed alone. In that context, Ben uses *Bad News* as the subject of a case study exploring these ideas [samuel2016crafting, pp. 384–405]—for the purposes of this dissertation, that writing may be consulted for an extensive explanation of how we cultivated an *aesthetics of the coauthored* in this project. #### Finding the Story A related phenomenon to shared authorship undergirds the *aesthetics of the uncovered*: when through endeavor one encounters an interesting story that would have otherwise gone unnoticed, a distinct pleasure obtains. In *Bad News*, stories are uncovered in a few different ways. Ultimately, they emerge from the simulation by virtue of some amount of story sifting, which means that I do a lot of the uncovering, since I am the wizard. When I extract potent material to pass off to Ben, it is extremely rewarding to hear it become realized, later on, in a conversational story. In these cases, the aesthetics of the uncovered are activated for me personally, which expresses the degree to which *Bad News*'s personnel are also consumers of the experience. Through discussion with him, I know that this pleasure also holds for Ben, even though he is the one delivering the conversational stories. In a recent paper [samuel2016computatrum], collaborators and I wrote about the degree to which creators of experiences in the area of computationally assisted performance may also be consumers of them. The aesthetics of the uncovered also prevails with regard to the experience of *Bad News* players. As I noted above, the storyworlds that drive this project are so vast that players can only plumb the surfaces of their narrative depths, which means that most of the latent interesting stories in a town go unnoticed by the player. When a player steers the conversation toward the revelation of a particular narrative, she uncovers something that would have otherwise remained hidden. This is especially true of cases in which player inquiries have lead me, as wizard, to extract narrative material that I had not already encountered. Finally, in the case of our *augmented radio play* performances, audience members have frequently called out ideas that have guided my story sifting toward potent narrative outcomes. When this happens, the live audience directly experiences the aesthetics of the uncovered. #### Small Miracles I have already explained, in Section 9.3, how many of *Talk of the Town*'s evocative emergent scenarios result from simple probabilistic procedures. Many of these, such as the restaurant that operated for over a century, seem to be intriguing simply because they are improbable. As Ben Samuel explains in his recollection of the playthrough involving that restaurant, in *Bad News* a kind of higher-order improbability may also obtain according to unlikely connections between the player's improvisation and the underlying simulation: > One of my favorite recurring themes is the serendipity that transpires when information concocted by the player dovetails so perfectly with the specifics of the simulation that it feels like it must have been planned and scripted from the start. One such example occurred during a playthrough that took place during San Francisco's 2016 “Come Out And Play” Festival. The player of this particular run had conceived of an ingenious cover story, that no previous player had ever presented: they were a restaurateur hoping to open a dinner theatre in town, searching to cast someone to play a character that matched the description of the deceased to a T. [...] Eventually the player was directed towards someone who would be able to help him, but who had some tragic family history of his own. This would-be informant came from a long line of restaurateurs himself—his great-great-uncle (for some number of greats) established a restaurant over a hundred years prior. This restaurant had been a staple of the town, and was passed down from father to son for many generations, until finally it was bequeathed to an owner who would never have any children of his own. When this owner chose to retire, he passed the restaurant on to his nephew, who had been a successful server at this restaurant all his life. However, less than a year later, the nephew managed to drive the restaurant into the ground; a century's worth of family history and town pride forever lost. [...] A town that has a restaurant that lasts nearly a hundred years is a near impossible occurrence in Bad News. Players who utilized the specifics of their cover story to create an emotional connection with the denizens of the town were almost just as rare. For both to transpire in a single playthrough, and to be so deeply interwoven into each other, was a small miracle, and will forever remain one of my favorite Bad Newsian experiences. [Note: Personal communication, June 14, 2018.] Intrigue rooted in improbability is what yields the *aesthetics of the improbable*, and when player improvisation serendipitously connects to remarkable underlying material, the aesthetics are doubly activated. #### Embracing the Uncanny In *Bad News*, we consciously design toward the *aesthetics of the uncanny* by acknowledging or even emphasizing *Talk of the Town*'s peculiarities at the surface of gameplay. Earlier, in Section 9.2.5, I discussed a bug in the engine's social simulation that produces romantic connections between characters with massive age differences. Like I mentioned in that section, through playtesting we discovered that this uncanny element actually yielded considerable intrigue—players were compelled to inquire about these bizarre relationships—and so we kept the ‘bug' and embraced the uncanny. As another example, in one of my favorite *Bad News* moments, Noah Wardrip-Fruin visited the home of a next of kin only to find that a four-year old child was home alone there. This curious situation was produced by the *Talk of the Town* subsystem that handles character routines, outlined in Section 9.2.4. Specifically, when a parent goes out on an errand, the routine system neglects to model that any young children who are at home should probably be brought along on the trip. In this case, the child's mother had gone off to a jewelry store, but Noah's player interface expressed that someone was inside the house making a small noise. He knocked, but no one answered—this is because a procedure for deciding who comes to the door specifies that small children never do. Noah then spoke aloud that he wished to look through the window, and so I extemporaneously typed a description of a small child watching television and displayed it on the player interface. Next, Noah did something that remarkably no other player has ever done—he broke into the house. This forced Ben to play the role of a four-year-old boy, which he did masterfully, and through considerable effort Noah learned from the boy that the mother had gone to the jewelry store. [Note: I have already mentioned that, in Tarn Adams's playthrough, Ben was tasked with playing a 111-year-old woman who was the next of kin. Considered together, these roles illustrate the breadth of characters that Ben may have to play at any moment. That he does this so well is impressive, especially considering that he does not know his next role until moments before a scene, and moreover none of the characters that he plays even exist until the hour of performance. The kid has chops!] This scene was bizarre in a good way—the aesthetics of the uncanny prevailed. #### Disposable Worlds Finally, the *aesthetics of the ephemeral*: no player's town has been visited before, and no one will go there again. To conclude the epilogue, I terminate the *Talk of the Town* simulation instance and the player and I watch as her little world disappears back to nothing—it was only an electric wind. ## The Pain of *Bad News * Owing to the success of this project, there is no need for me to take on the third person, as I did for the brutal critique of *Diol/Diel/Dial* that was provided in Section 8.4. While a lack of curation was identified as a fatal flaw of that experience, *Bad News* is a full-fledged curationist work that instantiates the entire pipeline of the curationist architecture. This allows raw storyworld material to ultimately be expressed in constructed narrative artifacts that are mounted in the media experience, in this case through conversational storytelling. #### Toward Automatic Curation Of course, curation in this project is human-powered, and this could be viewed as inflicting a distinct kind of pain. As I have expressed at several points in this document, the ultimate vision of curationist emergent narrative is one of fully automatic curation, but while this work is curationist, it does not come very close to meeting that vision. Indeed, in his own dissertation, my collaborator Ben Samuel also identifies human reliance as a potential weakness of this project: > The most damning criticism of Bad News in a dissertation about creating computational works that enable shared authorship is that it relies significantly on humans to achieve its goals. The presence of humans affords wonderful advantages, [...] but sadly without humans at play there are no computational mechanisms to fall back on. [samuel2016crafting, p. 400] In fact, Ben even proceeds to pinpoint a lack of automatic curation of simulated material, like I am doing in this section: > For example, I have discussed how all of the narrative in the system is emergent. The actor and the wizard are capable of leveraging their natural human proclivity for recognizing narrative to unearth the narratively interesting sequences of events from the town's entire simulated history. That is to say, though these interesting sequences were created by the simulation, the simulation itself has no recognition of their significance. [samuel2016crafting, p. 400] To be clear, I do think that computationally assisted live performance—specifically, hybrid works that incorporate human performers—is an exciting nascent area that we should not seek to computationalize. I know that Ben agrees with this too. What I am attempting to articulate here is that the prospect of fully automatic curation is itself alluring, in its own way, and in this regard *Bad News* tantalizes with its suggestion of how that might feel, because though it incorporates computation, its curation is in no way automatic. #### The Cover-Up Another ramification of the heavy utilization of humans in this project is related to a distinct appeal of that morphology: when humans are in the loop, the failings of underlying computational systems can be covered up quite easily. Much of the pain of *Diol/Diel/Dial* was constituted in blunders at the level of *World*'s simulation, as I expressed in Section 8.4, but in *Bad News* the errors of *Talk of the Town* do not really make it to the surface. Where the simulation lacks, the wizard and especially the actor augment, filling in its modeling gaps and injecting causal links that are not there and so forth. As I have stated several times now, I believe that computational techniques that are intended to support media experiences cannot be truly appraised except through actual implemented experiences of the kind that they are meant to support. This is not an original idea, but rather it follows earlier articulations by several other researchers [mateas2001expressive,stern2008embracing,horswill2013science,koenitz2014five,wardrip2014envisioning]. As an expression of this sentiment in this document, I do not appraise my simulation engines outside of the case studies that dissect actual media experiences built atop them. In the case of this case, however, we have a media experience that, through human power, is adept at leveraging a simulation engine's strong points while simultaneously hiding its weaknesses (and even potentially its errors). As such, this makes genuine appraisal of the underlying simulation difficult, but incidentally such an evaluation will in fact appear in this document, though elsewhere. In the story that opens the next chapter, I discuss an exploratory project in which I attempted to implement automatic curation of *Talk of the Town*'s storyworlds. My specific aim in this effort was to automatically sift for interesting emergent scenarios, with the idea that such material could then be passed off to a narrativizer that I would build in turn. Essentially, this would mean loosely operationalizing both the wizard and actor strategies that are employed in *Bad News*. What I found, however, was that this was essentially impossible due to particular deficits in the *Talk of the Town* simulation. As such, in telling this story, I will ultimately provide a more detailed appraisal of that engine. ## Curationism and *Bad News * In this project, the wizard and actor collaborate to construct stories out of the emergent material of a *Talk of the Town* storyworld, which are then mounted into the *Bad News* experience through the modality of conversational storytelling. As I have already shown in discussing the extensive example of curation in Section 10.3, this instantiates the curationist architecture outlined in Section 5.3—specifically the *feedforward curation* variant—though in a way that is largely dependent on human power. As such, *Bad News* represents an intermediate stage in my curationist practice, one that spans between *Diol/Diel/Dial*'s lack of curation and the fully automatic curation of *Sheldon County*. In this section, I will briefly summarize how *Bad News* instantiates the framework for curationist emergent narrative that I presented in Chapter 5. As I did in my earlier case study of *Diol/Diel/Dial* (see Section 8.5), I will now proceed to answer the following question: for each curationist architectural component, who or what in this project acts as that component? Figure 10.9 illustrates these concerns with regard to *Bad News*, which I will also outline in prose, component by component: - **Experiencer.** *Bad News*'s *experiencer* is the player who sits in the gameplay area of our physical installation and explores the generated storyworld to speak with its residents. As I have noted, in special cases we allow for multiple players, in which case each one is an experiencer. In some special cases, there may be other experiencers who are not players. For example, in our so-called *augmented radio play* variant, described in Section 10.2.5, audience members may also experience the piece, but not in the same way that the player does. For the sake of simplicity and comprehensibility, Figure 10.9 does not capture these considerations. - **Simulated storyworld.** In each *Bad News* performance, a *Talk of the Town* simulation instance is the *simulated storyworld*. It is the procedurally generated American small town that the player explores to find and notify the next of kin of one of its residents who has died. - **Chronicler.** As with *Diol/Diel/Dial*, there is no distinct *chronicler* module at work in *Bad News*, but rather the associated functionality is built into the *Talk of the Town* code. To simplify, we can say that *Talk of the Town* is its own chronicler. - **Chronicle.** A *chronicle* in this project comprises all of the various data that is queryable in the wizard console, which is a Python interpreter wrapped around an active simulation instance that is stored in working memory. This is all of the data that is tracked by the *Talk of the Town* simulation engine, and it represents a stored record of much of what has transpired in the history of the town at hand. [Figure omitted] - **Story sifter.** Unlike *Diol/Diel/Dial*, this project actually employs a *story sifter*. As I explained in great detail in Section 10.3, in *Bad News* the human wizard does story sifting. Specifically, I do livecoding to sift through the accumulated data representing the history of the town, my aim being to uncover interesting emergent scenarios (of the kind outlined in Section 9.3), especially ones that connect to the deceased character in some way. It could also make sense to say that the actor interface also sifts this material, since it dynamically updates (as the experience proceeds) to display storyworld information to the actor. However, the data displayed on this interface tends to lack narrative intrigue; rather than the makings of storylines, it expresses the basic material that Ben needs to perform. As such, I think it is probably not worth considering this notion in detail, and Figure 10.9 does not capture this interpretation. - **Material.** In this project, the sifted *material* is textual data that the wizard delivers to the actor via a chat interface (the actor–wizard web chat). Figure 10.3 shows an example of such marerial, as seen in a screenshot of the chat taken during a performance. - **Narrativizer.** Again *Bad News* prevails over *Diol/Diel/Dial*, in this case because the experience itself, rather than a human interactor, carries out narrativization. As I discussed at length in Section 4.2.1, a fundamental pitfall in emergent narrative is simulations that do not understand their own storyworlds and instead offload the burden of narrativization onto their human experiencers. In the case of this project, the simulation does not understand its own storyworlds, but a human duo works to curate them so that actual narrative artifacts may be constructed and mounted in the experience. In this configuration, the actor is the *narrativizer*. As I showed in detail in Section 10.3, one of Ben's major tasks as actor is to narrativize the excavated material that I deliver to him to build narrative artifacts that are mounted into the *Bad News* experience through conversational storytelling. [Figure omitted] - **Story.** In *Bad News*, a *story* is told by the actor in the course of a conversation with the player. In the form of performance transcripts included in Section 10.3, I provided several examples of such stories. Like I have noted several times above, a constructed story may be delivered in any modality that supports narrative. For example, in Section 3.1.1, I provided a case study of curationist emergent narrative that centered on Tim Denee's *Oilfurnace*, which is a comic that recounts the emergent intrigue of a *Dwarf Fortress* storyworld. Clearly, curationism utilizes an ecumenical notion of ‘story', and these two examples express the wide range of modalities that may be targeted in works following this framework. With *Sheldon County*, I am again targeting the modality of *oral narrative* [bauman1986story], but in that case the delivery is monologic instead of conversational. - **Media experience.** In curationism, the term ‘media experience' specifically denotes the experience that curated narrative artifacts (each being a ‘story') are mounted into for human encounter—thus, in this project, the *media experience* is *Bad News* itself. Because this experience is not the simulated storyworld, but rather a decoupled installation-based work, the project uses the architectural variant of *feedforward curation*. By this variant, described in Section 5.3.2, stories that are constructed out of the generated material of a storyworld are mounted into a decoupled *second-order* experience. I realize that the matter may be a bit confusing in this case, since a *Bad News* player *does* explore the simulated storyworld recounted by the stories that are imparted to her. The distinction here is subtle and it arises from the peculiar human-powered nature of the project. In *feedback curation*, constructed stories are fed back into the storyworld itself, since that is the mounting target. In *Bad News*, curation is not cybernetic because the stories that Ben constructs are not actually fed back into the storyworld—when he delivers a story to the player, I have no means with which to actually inject that story into the world through the mechanism of *Talk of the Town* simulation. This is there is no representational scheme by which I could encode a conversational story in a way that the system would actually be able to process. In feedback curation, the story that is fed back into the storyworld is actually understood by the simulation that models that world. An illustrative case of this would be a videogame where non-player characters tell stories to one another about what has happened in the world, and critically in such a way that the recipients of those stories actually process them in a meaningful way. [Note: While I am not aware of any examples of such a thing—though it was the basis for my original dissertation topic and I would like to return to the idea at some point—the example of feedback curation that I gave in Section 5.3.2 was *Eve Online* player-constructed propaganda—about which Marcus Carter has written [carter2015emitexts]—being broadcasted in-game (thus, in the actual storyworld). In this case, the simulation does not understand the stories, but the human players do, and they may respond by taking action that actually changes the world. In the case of *Bad News*, Ben may play a character who tells a story and is affected by the telling, but the experience has no means by which the underlying representation of that character, in *Talk of the Town*, may be altered accordingly. As such, the *Talk of the Town* storyworld does not change when stories are constructed out of its generated material. Generally, while *Bad News* takes place in those storyworlds, at a technical and ontological level, it more precisely takes place in a kind of liminal projection of that world onto the high-fidelity space of our world (and onto the modality of face-to-face conversation). It would be interesting to explore how Ben's curated stories might somehow be injected back into the storyworld, perhaps after gameplay. While the epilogue does not incorporate anything that happens during a performance, as I explained in Section 10.2.4, I really wished it did.] #### Summary To recapitulate, *Bad News* is a full-fledged work of curationist emergent narrative, but one that relies heavily on human power. Specifically, this project that employs the *feedforward curation* variant of the curationist architecture (see Section 5.3.2), since it mounts stories curated out of *Talk of the Town* material into a second-order experience, *Bad News*. In its instantiation of this pattern, a human wizard works a story sifter to deliver narratively potent raw material to a human actor, who narrativizes that material to build narrative artifacts that are then mounted in the experience through conversational storytelling. As a successful proof of concept of curationist emergent narrative, *Bad News* begs the question: what would such a project look like if this kind of curation was carried out without human help? Later, in Chapter 12, I outline my ongoing exploration of fully automatic curation in the generative podcast *Sheldon County*.