Intermittent Mechanism

Note: This blog post assumes some familiarity with the 2005 adventure game Façade, and how it works. For a background on it, please check out my presentation here: https://docs.google.com/presentation/d/1kDkpDEd–1-67mpshn3-0NO8ISIG6rgdZfsTQvsa3_c/edit?usp=sharing

Façade was the first game I ever downloaded on my PC. This was back in middle school, when Let’s Plays showing the game struggling to keep up with a barrage of hilarious and vulgar inputs from YouTubers were all the craze. In turn, when I played it for myself, I did the very same thing – push the game’s boundaries with bizarre dialogue and laughed as Grace and Trip looked at me in wide eyed surprise. I never ended up finishing it though, and now, after playing it through for the first time, I regret that.

Even back then, however, I was fascinated with the potential of a game like Façade. In-between bouts of laughter, I marveled at how Trip and Grace were still giving relatively human responses to the ridiculous things I was saying, and wondered how such a thing was ever made – especially in 2005.

Years later, I now have played Façade, and have only become more fascinated by how it works. Additionally, I have since began researching computer science and looking into concepts such as artificial intelligence and game design. Digital experiences like Façade, and the future that lies before them, is the place where all these interests intersect.

I touched on this in my presentation, but it is impossible to deny there is a significant demand for a follow-up to Façade. Everywhere I looked in the process of researching for my presentation and blog, there were commenters clamoring for a follow-up or remake. To their point, a follow-up was in one point in development, and by Façade’s original creators, to boot. Titled The Party, developer Matthew Mateas pitched it as such in an interview:

“where Façade had two computer-generated characters, The Party will have ten, a far more complicated proposition, but dramatically richer… The game will last about forty minutes, rather than twenty. It will support more physical action, allowing the player to do things like rendezvous with characters in a private room, lock doors, carry things around, and fire a weapon.”

Matthew Mateas

To me, and other fans of Façade, this concept sounds nothing short of perfect, and represents the ultimate evolution of what Mateas’ original project sought out to do. In turn, the way he capped off his interview was rather unfortunate. When asked if such a project was feasible, Mateas responds by saying a simple prototype would be “doable within twenty years” (Mateas 2006).

Twenty years? Surely that has to be some mistake. With how viral Façade went on the internet (even despite the fact that it got popular for its unintentional comedic moments), surely the market must be saturated with AI-driven verbal sandboxes just like it. Something like The Party must be in development as we speak!

Sadly, none of these hopes have any evidence to back them up. Releases of this type of game, which I call “Artificial Intelligence (AI) Adventures”, have been scarce over the years since Façade’s release. Only two notable examples come to mind: Event[0] and the AI Dungeon series, and both will be focused on later in this blog post, all in the context of Façade and the future of this genre.

But first: what is an “AI Adventure”? I see AI Adventures as an evolution of the “Text Adventure” game, a genre popularized in the 70s and marked by the usage of text commands to influence the game state – being utilized in controlling one’s own character, influencing the environment, and defeating enemies all by utilizing a bank of commands (similar to a computer terminal). However, these types of games are limited in terms of what specific commands they can process – if you don’t type in something that is explicitly planned for, all the game can do is give you a generic response and wait for you to tell it something it can parse. This is where AI Adventure titles differentiate themselves. In games of this type, artificial intelligence and language processing are used to make software attempt to respond intelligently to any input, no matter if a hard-baked response to it is present. This allows for a near-infinite amount of versatility and gameplay options, but also justifies why AI Adventure titles might be so rare. Making AI models powerful enough to reasonably respond to players both poses a huge challenge technologically, and also doesn’t make for a marketable framework – games like this, by design, are often absent of non-stop action, cinematic set pieces, or riveting multiplayer, all of which are big selling points of today’s industry.

But this makes the AI Adventure games that did end up coming out there all the more fascinating. I would consider the 2016 title, Event[0], a perfect example of the genre. At first glance, the text-input system of this game looks more like a traditional adventure game than an AI one, as the player’s main interaction is with a literal computer terminal – a system which is notoriously resistant to any out-of-the-box input handed to it. However, I should note that this is not just any terminal, but in fact one controlled by an in-universe sentient AI, and one capable of driving the game’s entire story forward by procedurally generating over two million lines of dialogue. In fact, the AI behind the computer terminal is so complex that it is able to shape a personality in accordance with the player’s input and even unintentionally made a new ending to the game through a bug in how it worked. AI-driven actions like this, that are both profound story-wise and technologically, are what make me beyond excited for the future of AI Adventures, and stand as the true evolution of Façade.

Speaking of the future, I would be remiss not to mention the AI Dungeon series when discussing this new genre. Originally created in 2019, the game is at first glance a text adventure game not unlike Zork and other titles of old – its interface, after all, is just simple text atop a black screen, narrating one’s journey paragraph-by-paragraph as they go along. It doesn’t take one long, however, to see where AI Dungeon makes itself unique (it’s in the name, after all!). In truth, every line of text within the series of games is written by an AI, building off an input consisting solely of a training set of known stories and the player’s own instructions, rather than a bank of keywords and actions. All you have to do is choose a setting (or make your own) and give it a sentence to start out with, and a wholly custom text adventure will form around your successive inputs. Furthermore, the AI used is not just powerful, but one of the most robust models in the entire world: as of the latest version, AI Dungeon is powered by OpenAI’s Generative Pre-trained Transformer 3, or GPT-3 for short. To give you an idea of how awe-inspiring and versatile this AI is, here is a set of images that GPT-3 made from scratch, just from someone typing a text prompt to it.

To me, these images are shocking – the illustrations are so well done that I fear for the clip art industry’s future, and the picture mockups are indistinguishable from real photos. Now, imagine this type of power, but brought to storytelling.

If you have played AI Dungeon, you may see this as a bit of an exaggeration. And I’ll admit – stories within the game can often get derailed or sound like gibberish. But even still, the ability of this game to improvise around the player’s input dwarfs that of Façade. I have seen the program respond reasonably to full paragraphs of input, I have seen it dig up old characters and utilize them flawlessly, and I have seen it build an understandable story complete with a beginning, middle, and end. AI Dungeon is nothing short of awe-inspiring and is the type of experience that I’d recommend to anyone besides the faint of heart (as capable as the model is, it has a tendency to create content that veers into the explicit, which led to developers making the controversial decision to try and wipe such stories from the platform).

This all is not to say that AI Dungeon has proved Façade redundant in the modern age. I thoroughly have enjoyed playing both, and each experience offers something completely different. Rather, I mean to say that the future of AI Adventures could represent a wonderful melding of these two experiences. What if Mateas’ Party could finally be realized, with the power of GPT-3 at the helm? What if someone could form 3-D explorable worlds around the infinitely expansive creations of AI Dungeon?

These are games that I would stop at nothing to get to play in my lifetime, but such a sentiment requires a lot of hope. In the above-mentioned interview, it was hinted that Façade’s follow-up has likely not be realized due to lack of publisher interest. Event[0], despite good reviews, hasn’t necessarily clawed into the mainstream. AI Dungeon’s first version floundered to pay for its own bandwidth. But, with its newest GPT-3-powered edition attracting $3.3 million in seed funding, I firmly believe the future is bright for AI Adventures. With news like this, Mateas’ 20-year estimate for The Party could be delightfully proven wrong.

Or not. Whatever happens, though, I’ll be waiting for it all the same.

Works Cited

Adventure Gamers. AdventureGamers.com, cdn.nivoli.com/adventuregamers/images/screenshots/15792/3997.jpg.

AI and Games. “The Story of Facade: The AI-Powered Interactive Drama | AI and Games.” YouTube, 22 Apr. 2020, http://www.youtube.com/watch?v=POv1cOX8xUM&ab_channel=AIandGames.

“AI Dungeon-maker Latitude Raises $3.3M to Build Games with ‘infinite’ Story Possibilities – TechCrunch.” TechCrunch, 4 Feb. 2021, techcrunch.com/2021/02/04/latitude-seed-funding/?guccounter=1&guce_referrer=aHR0cHM6Ly9lbi53aWtpcGVkaWEub3JnLw&guce_referrer_sig=AQAAAAZ6u5hPIa7VBrZTc6BQBNL81pqT_xjcdsiu4ntoYsIL2Vx-oqID-_5GlsAbJYsXzitWCaX8IWOEQtGD-WytQv9sNJen5QpzcXlbVdrWTGvaQpZtARACtnSxLINFIg3UjLd-NSxxYpxQr8zs5pYjyWxuG2t-5So2A-7Fqyp20rHO.

“Event[0] is 2001 Meets Firewatch, Due This September.” Eurogamer.net, 13 July 2016, http://www.eurogamer.net/articles/2016-07-13-event-0-is-2001-meets-firewatch-due-this-september.

Event[0]. Directed by Ocelot Society, 2016.

Facade. Directed by Andrew Stern, and Michael Mateas, 2005.

Harris, John. “Creating the Ever-improvising Text Adventures of AI Dungeon 2.” Game Developer, 9 Jan. 2020, http://www.gamedeveloper.com/design/creating-the-ever-improvising-text-adventures-of-i-ai-dungeon-2-i-.

Latitude. AI Dungeon, play.aidungeon.io/.

Mateas, M., and A. Stern. “A behavior language for story-based believable agents.” IEEE Intelligent Systems, vol. 17, no. 4, 2002, pp. 39-47.

Ocelot Society. “Event[0] by Ocelot Society.” Itch.io, ocelotsociety.itch.io/event0.

OpenAI, 18 June 2021, openai.com/.

Quach, Katyanna. “AI Game Bans Players for NSFW Stories It Generated Itself.” The Register: Enterprise Technology News and Analysis, 8 Oct. 2021, http://www.theregister.com/2021/10/08/ai_game_abuse/#:~:text=AI%20Dungeon%20was%20predisposed%20to,and%20characters%20into%20their%20stories.&text=The%20document%20contained%20a%20dump,the%20website%20Choose%20Your%20Story.

Rauch, Jonathan. “Sex, Lies, and Videogames.” The Atlantic, 1 Nov. 2006, http://www.theatlantic.com/magazine/archive/2006/11/sex-lies-and-videogames/305293/.

“This AI-Powered Choose-Your-Own-Adventure Text Game Is Super Fun and Makes No Sense.” Gizmodo, 3 Aug. 2020, gizmodo.com/this-ai-powered-choose-your-own-adventure-text-game-is-1844593111.

“Virtual Friends: AI-Powered Chatbots Help Against Self-Isolation.” MedicalExpo E-Magazine, 7 July 2021, emag.medicalexpo.com/ai-powered-chatbots-to-help-against-self-isolation-during-covid-19/.

Wawro, Alex. “Event[0] Has an Ending So Secret Even the Dev Team Didn’t Know.” Game Developer, 12 July 2017, http://www.gamedeveloper.com/design/-i-event-0-i-has-an-ending-so-secret-even-the-dev-team-didn-t-know-about-it.

Harrison Scott

Colossal Cave Adventure: A Modern Critique

My Initial Expectations

Heading into my first playthrough, my expectations were for Adventure to be a procedurally-generated cave game with a set few dozen rooms generated on the fly by the program. Starting at the end of the road, I quickly found myself in the forest; with the idea that my player was indeed moving in space, I continued in one direction for a solid five minutes before giving up; I promptly “turned” my character around (in my mind’s eye, anyway), and buckled up for what I thought was going to be another five minutes of backtracking to my starting point.

Only travelling back didn’t take five minutes– it was instant. I realized the game did not work based on a traditional 2-D basis, but rather via a finite number of ‘cells’ interlinked with one another; I found there were two forest cells, one of which was “near both a valley and a road,” and the other of which runs along “a deep valley to one side.” As a result, I learned space could be non-euclidian in this video game world; I could walk through the forest for hours, turn around, and realize I hadn’t moved far at all. 

Perhaps motivated by Murray’s four essential properties of digital environments, I realized my hopes for procedural generation weren’t met; the program, though, is still quite procedural in the way it computes gameplay. I began to wonder whether Adventure qualified as a digital environment in Murray’s sense, thinking through the other three essential properties:

1. Is Adventure participatory?
   1. Most certainly. As alluded to by Aarseth, the player must participate in non-trivial activity to continue their playthrough– entering text in a field. 
2. Is Adventure spatial?
   1. In a non-euclidian sense, yes, though the number of cells (40) actually seems quite limited after a few hours of gameplay; the game makes up for this, in my opinion, with some great description and puzzling (alongside unjustified hurdles I critique later on in this blog post).
      1. If each of the 40 rooms were 130x130x130ft (40x40x40m) on average (a major overestimate), one could still fit up to 3,600,000,000,000 colossal caves in a single Minecraft world. Although Minecraft is significantly larger in scale, Adventure’s narration and design makes the world feel every bit as dense and complex as any Minecraft cave system.  
3. Is Adventure encyclopedic?
   1. Most certainly, for there are myriad items the player can pick up. However, compared (admittedly, unfairly) to modern-day titles like World of Warcraft (housing over 116,000 unique items), Adventure’s implementation is light.

Critical Overview

Tutorial: Outside the Cave

Perhaps an unavoidable consequence of using a basic NLP system, I found the introduction sequence unbearably frustrating. Aside from the realization that space could be non-euclidian in this game world (spurned by minutes spent wandering aimlessly through the forest), the game’s occasionally inconsistent vocabulary was cause for frustration. For example, my first command upon loading up the game for the first time was “enter building;” after all, the directions specify our character stands before a wellhouse. To my surprise, this command worked– after picking up the loose items, I went to leave the building. However, entering “leave building” produced this output:

But you aren’t in the well house. 

I then tried many combinations of the same command:

>exit house

But you aren’t in the well house.

>flee house

That’s not a verb I recognize.

>run away

You can’t go that way.

At a loss, I finally “caved” and read the game’s instructions– I was able to leave the house by simply “walking” west. 

After contemplation, I thought this blatant contradiction– “enter building” working as expected and “exit building” doing the opposite– could be a design choice aimed at acclimating the player to the cardinal direction movement system. Perhaps Crowther/Woods thought allowing the player to enter but not exit would provide the player with enough curiosity and incentive to actually read the “HELP” text. A better way to motivate the player in this way, in my opinion, would be to script the following after the player tries to leave the building:

You start to leave the building, but spot a sign above the entryway. ‘ONLY TRAVEL NWSE, OR RISK GETTING LOST.’ You stop before the entryway.

This would more obviously call the player’s attention to the NWSE movement scheme, as opposed to the irritating:

But you aren’t in the well house. 

On the other hand, one aspect the tutorial sequence gets right is world building. For instance, although the first few areas beyond the end of the road– the valley, the slit in the streambed– are devoid of any interesting action aside from aesthetic descriptions, the lack of activity in both of these places serves the game’s overall narrative by making the above-ground seem boring and trite. When we combine this trite environment with details alluding to passage below– a gully flowing beyond the wellhouse, water flowing into a slit in the streambed, the dry streambed itself leading into a damp depression surrounding the grate– the player’s curiosity is piqued. Much like the flowing water, I felt drawn to the depths below. 

Depth Level 1: Hall of Mists

Beginning with my critiques of this section, I found it unnecessarily counterintuitive at times. For example, the black rod was a major cause for confusion– thinking it was some sort of weapon, I picked it up as soon as I could to prepare myself for the depths below. Once I found myself in the Hall of the Mountain King faced by a giant snake, I realized the rod was completely useless; recalling the bird in the canyon and the wicker cage I had picked up earlier, I quickly backtracked to try and capture the bird– only to be met with:

>capture bird

The bird was unafraid when you entered, but as you approach it becomes disturbed and you cannot catch it. 

I soon returned back to the snake, only to be attacked by a dwarf; even the dwarf’s axe, however, wouldn’t make a mark on the snake. 

Although the bird’s fear of the rod is mentioned in the game’s “HELP” text, I feel this fear works against the player’s progression in an unjustified and unnecessary way. For instance, if the rod were revealed to have belonged to a wizard that killed birds, this complication would feel more justified; as it stands, the bird’s fear of the rod seems to be nothing more than the developers trying to work against the player. 

On the other hand, I found Depth Level 1’s aesthetic design especially compelling. Following white mist meant magic was afoot; whether it was drifting upward from a cavern below or lining the walls of the corridor my player was lighting up with their lantern, the use of this white mist added a sense of ambience and atmosphere I really enjoyed. Furthermore, some of the cells were especially interesting to imagine; for instance, the “window” cell caught my attention due to its mystery:

“You’re at a low window overlooking a huge pit, which extends up out of sight. A floor is indistinctly visible over 50 feet below. Traces of white mist cover the floor of the pit, becoming thicker to the right. Marks in the dust around the window would seem to indicate that someone has been here recently. Directly across the pit from you and 25 feet away there is a similar window looking into a lighted room. A shadowy figure can be seen there peering back at you.”

Is this pit some kind of stadium? Are these windows intended to be for two spectators? Why just two? Why can’t we see the floor? Why is the mist thicker to the right side of the room? Who was at this window earlier, the dust preserving their presence? Who lies across the pit? So many questions, zero answers– the fact that the player feels motivated to ask them is an indication that the cave’s worldbuilding is working wonders. 

Depth Level 2: Dirty Room

Perhaps a combination of being accustomed to the control scheme and becoming more interested in completing the game, I have nothing but praise for Depth Levels 2 and 3 of the game. Depth Level 2 stood out especially due to the unnatural way in which its main chambers are reached; before them, we must descend to Depth Level 3 and find an entrance upward back into Depth Level 2. As we pass through Depth Level 2 for the first time, the player feels there’s more to explore here and searches Depth Level 3 for an alternate entrance to the level above. 

Another design choice I found rather charming was the troll’s demand for treasure. Each time the player crosses the troll’s bridge, they must pay with at least one form of treasure; to free Depth Level 2’s bear and acquire its gold chain, the player must come prepared to sacrifice two pieces of treasure to cross the bridge twice. One issue: it is possible for the player to be stuck on the bear side of the troll bridge without any treasure. If this is the case, the player is soft-locked from continuing their playthrough; they must restart, as they can no longer progress to areas on the other side of the troll’s bridge. 

Depth Level 3: The Bedquilt

Depth Level 3, home to a few secret levels, was rather easy to navigate by this point in my playthrough. One aspect I could pinpoint as enjoying, however, was the bedquilt itself. This cell is completely unique to other cells found in Adventure in that escaping it is 100% dependent on random number generation. There is a chance that walking any of 4 directions will allow the player to escape the cell– progressing beyond the bedquilt is simply a matter of brute force. 

Overall Impression

Although the game isn’t the procedurally-generated cave crawler I was expecting, I definitely understand and appreciate the cult appeal. Aside from a few annoyances (not being able to “exit” the well house, the intricacies of the infamous black rod, potentially soft-locking a playthrough by either killing the bird or running out of treasure on the other side of the troll bridge), the game is a fount of entertainment. 

by Sydney Glenn

! There are spoilers for Disappearing Rain in this blog post !
To avoid spoilers, look for the (SPOILER) tag bracketing certain lines.

Upside down underwater amidst a turbulent storm– that’s what my first two hours engaging with Deena Larsen’s hypertext fiction Disappearing Rain felt like. Confused and blinded by the darkness of the water, I flailed against the torrents, waves hitting me from all angles as I struggled to discern up from down. Yet despite the turmoil, I was already absorbed in the mystery of the narrative– the disappearance of Anna Mizunami. As I fell down a rabbit hole of hyperlinks and haiku, like Mat Anderson and Patricia Tomaszek, contributors to the Electronic Literature Directory, muse in their eponymous article about Disappearing Rain, “Larsen … orchestrate[d] [my] own disappearance in the virtual reality of the Internet,” one that, as I would soon realize, mirrored the disappearance of Anna herself (Anderson and Tomaszek).

Disappearing Rain follows a cast of six family members caught up in the sudden disappearance of Anna, one of the twin daughters. Players can follow multiple characters’ story paths, including Amy, Anna’s twin sister, as she delves into the depths of Anna’s computer looking for clues, Kit and Richard, the twins’ parents, as they fight their increasing desperation over Anna’s whereabouts, and Sophie and Yuki, the grandmother and great-grandmother, in their own stories of credit card fraud, navigating memories, and the preserving the souls of their ancestors.

The story of Disappearing Rain is scattered across 144 different webpages connected with over 1,000 hyperlinks, forming an intricate spiderweb of plot points. While the prospect of piecing together so many story fragments may intimidate first-time players, Larsen has already taken this into account; she provides multiple diving spots into the story via the home page. The most visually striking examples include two links attached to kanji characters “mizu” and “kawa,” meaning “water” and “river” respectively. Accompanying the characters are the titles “Part 1: Water Leavings” and “Part 2: River Journeys,” which, along with Larsen’s description of the work, inform the player that the story content divides into two thematic halves. The description also holds four more links associated with the storylines of specific characters or character pairs as well as the title itslef. Finally, two unassuming lines, “Come on in” and “The water’s fine,” are the last remaining hyperlinks.

For my opening dive, I clicked on “Water Leavings,” and my eyes were immediately inundated with tons of visual stimuli. The “Water Leavings” and “River Journeys” pages contain a series of haiku– one central, bolded haiku reading from top to bottom, and a series of smaller, unbolded haiku reading in lines from left to right that branch off the individual lines of the bolded haiku. Clicking a link on an individual line or segment of a haiku links to another page with a short section from the story, each typically containing multiple linked words to another webpage’s story. And at the bottom of each page lies two haikus oriented horizontally– the bolded haiku (associated with part 1 and part 2), and the haiku subsection linked to the particular line of the bolded haiku. Thus, the haiku form an organizational system resembling a table of contents– the story divides into two thematic parts, represented with a bolded haiku; each line of the bolded haiku links to a sub-haiku, each line of which contains a page of the story. Thus, each line of the bolded haiku functions like the chapter of a book, containing a series of pages as represented figuratively by the individual parts of the sub-haiku and literally by the contents within these parts’ webpages.

Upon my first entry into Disappearing Rain, I was not able to glean this information about the “Water Leavings” page, so I simply clicked the first word that appealed to me. And thus, for my next two hours of engagement, I followed the flow of the haiku, often clicking page-embedded links on a whim, or following each individual haiku to its end. The experience was gripping to say the least– sometimes I would read a particularly cryptic linked phrase from a webpage and click it, only to be delivered to a seemingly-unrelated anecdote. Other times, the connection would be more apparent, or occasionally, contain an entirely new piece of the story from a much later or earlier point in time. The latter type of link never ceased to shock me as it felt like I was stumbling upon plot spoilers; since my experience was nonlinear and chaotic, I had little context for the plot or characters, and the sudden skips in time only exacerbated my confusion, albeit magnifying my excitement for the mystery tenfold in the process.

Only after examining the constitution of the home page and discovering the organizational structure of the haiku did I perceive a straight path through the narrative. Despite it appearing obvious in hindsight, my discovery that the horizontal haiku of “Water Leavings” page were entirely chronological– both within the sub-haiku and between the sub-haiku –I was thrilled and proud of myself at the discovery, as I also was when I investigated the haiku on each character page and discovered them to be chronological selections of haiku lines from both “Water Leavings” and “River Journeys.” These character haiku extracted chunks of the story from within both parts and arranged them linearly; they can feel a bit disjointed to read without context of the full plot, but they glean insight at what moments Larsen thinks is most impactful in the journeys of the characters.

To properly reveal the ingenuity of Disappearing Rain, I will circle back to the quote I mentioned at the beginning. Larsen attempts to replicate the experience of disappearing into the Internet for the player by dropping them into a pool of stories without any imposed pathway through it. While she does offer structured paths in the forms of the Water Leavings and River Journeys haiku and the character-associated haiku, the novelty of the hyperlink fiction medium to most players will likely invite them to jump in headfirst, attracted by the appeal of clicking on the links, and the embedded links within the pages will form a disjointed path through the story. Unbeknownst, or possibly partially realized by them, they will mirror the experience of Amy, jumping from clue to clue without any particular logic– and if they were like me, spend hours engaged, with strained eyes, an actualized Internet binge.

I simply wanted to piece the story together, but my current approach was doing me no favors. So I returned to the home page, and after a critical examination of the Water Leavings page structure, found a chronological way through the study. I restrained from clicking any of the mid-text links to prevent myself from confusing my understanding of the plot. This experience was rewarding in the sense I achieved a fairly clear knowledge of what was going on with each character and where Anna actually disappeared to. It also brought me smaller revelations, such as, (SPOILER WARNING) the realization that the home page linked lines “Come on in. The water’s fine” were what Amy says to us, the readers, to invite us into the Internet with all the other disappeared souls, as well as what Sophie said to Amy at the river where she contemplated suicide (SPOILER END). However, while uncovering the story was indeed satisfying, something about concluding my experience with Disappearing Rain there felt wrong.

In hindsight, I think this disappointment stemmed from the conflict between my hardwired need to perceive Disappearing Rain as a story versus a recreation of the imagined experience of disappearing online, which I believe is closer to what Larsen intended. Viewing it as a story, the only way to have a traditionally-satisfying reading experience is to read Disappearing Rain from beginning to end. However, if this true, there is no need for it to be written as hypertext fiction– if the goal is purely to understand the story, a linear novel would suffice. Thus I’d claim that Larsen envisioned a concrete purpose for her choice of hypertext as a medium for the story, and I’d argue that it intends to simulate the experiences of Amy, Anna, and potentially Yuki as well.

The connection to Amy is fairly obvious– in scouring Anna’s search history for clues, bouncing from topic to topic without context, similar to how we experience the sudden skips between events in the story. However, I feel like the act of skipping through time is better explained by the descriptions Yuki’s (SPOILER WARNING) battle with Alzheimer’s. She hallucinates Anna’s ghost (as indicated by the “Hallucinations” link in “snow falls” linking to a description of Yuki with Anna’s spirit in “crystal edges”) and is brought to tears only to forget moments later; and when she dies, her memories begin to congeal in one pocket in time– in “murky puddles,” Larsen describes Yuki’s memories “shift[ing] like the water bubbling from the back of the fountain,” no longer fixed linearly but overlapping on top of each other. Yuki struggles with remembering individual memories, which Larsen describes as “push[ing] the water of her memories into the forefront of her mind.” When she tries to select one, the multitude of them “[cascade] out again like the waterfall in the fountain.” But perhaps most important is Yuki’s specific inability to discern days and faces from each other– this directly mirrors the new player’s experience should they choose to stumble from link to link before experiencing Disappearing Rain chronologically. Entering the story with no knowledge, like Yuki into the space of her memories, they stumble upon a random moment in time with no conception of when it exists in relation to the others, only to be redirected to a completely different one at a different location in time, with only small, abstract hyperlinked-threads connecting the two. (SPOILER END)

I can’t help but feel that the nonlinearity of the experience might connect to Anna’s (SPOILER WARNING) and Amy’s experience inside the Internet, yet due to the story’s evasive descriptions of the Internet-space, I can’t conclude anything decisive. But I still wonder, how does Larsen envision time working in the online space where Anna and Amy reside? Do they skip from future to past to present, like us readers? Or do they experience the flow of time linearly, occupying the same space in time as their family? How does this work with the letters they write to their family and contacts on the outside world? (SPOILER END). Perhaps the intentional ambiguity reflects Larsen’s and the world’s collective uncertainty with the Internet and the mystical viewpoint many adopted towards it– an interesting phenomenon to observe 22 years later when the Internet is as pervasive and ordinary as ever.

For some readers, piecing together a story nonlinearly by navigating hyperlinks can feel uncomfortable or stressful; clicking on a link could be a source of anxiety if the abstract, seemingly random connections between pages instill a fear of experimentation, of clicking a “wrong link” and being shunted off to irrelevant plot point and tainting one’s experience of the narrative. I believe Larsen’s choice to include multiple entryways into the narrative aims to bridge both approaches– the existence of both the haiku table of contents pages organizing parts 1 and 2 and the mid-text links allows the reader to determine what degree of linearity they want to experience the story.

There is so much more potential for analysis beyond these discussions of linearity– the content of the haiku and choice of linked words, both in the representation of their pages, the two parts, and of the characters, the magical properties of water and rivers, and more story secrets, such as (SPOILER WARNING) the identity of the person writing Anna love letters, (SPOILER END) to name a few. With all these tools and mysteries at its disposal, Disappearing Rain justifies its existence as hypertext fiction and establishes itself as a multimedia art form, setting a satisfyingly-high bar for future works of hypertext fiction to reach.

Tomaszek, Patricia, and Mat Anderson. “Electronic Literature Directory.” Disappearing Rain | Electronic Literature Directory, 19 May 2010, https://directory.eliterature.org/node/516.