In interactive fiction, are all sentences created equal?
Aaron Reed might not think so. His latest work, Almost Goodbye, procedurally generates what he calls satellite sentences, bits of prose that anchor setting and color dialogue.
When I think of that buzz phrase, procedural content generation (PCG), I think of Dwarf Fortress, in which vast geographies and histories owe their existence to the ruminations of computer algorithms, shaped by human creativity and tempered by practical know-how. It’s easy to get lost in such a world. When you build a fortress that is so unmistakably yours within a setting that is so exquisitely randomized, you participate in a conversation of mutual respect with the game author. Your voice feels just as important as that of the developer, and the performance that emerges from this controlled chaos is unexpectedly personal and relatable.
But enough about Dwarf Fortress. I think it’s safe to say that the implied goal behind PCG is replay value. It caters to the school of thought that sees pure games not as contained stories, but as infinite possibility spaces. Movies, books, songs: these are things that we take in once, then discard or re-experience later. But let us talk of the game that has infinite replayability. This is what we’ll take with us to our desert island to experience over and over again, as the chains of time rust and finally fall. Such is the Mecca of procedural content generation.
Adventure games and interactive fiction arguably owe more to traditional narrative forms than they do their more ludic predecessors. Almost Goodbye is a brief sci-fi affair that attempts to bridge the gap between the two disciplines. Our protagonist, Dr. Muriel Ross, is preparing for a one-way trip to the other side of the galaxy, and she’s got a few hours left for farewells. That’s where the “interactive” part comes in. You get to decide the order and location of her visits, and you’re thrown a handful of dialogue choices. A display to the right lists the time of day, location and some descriptors reflecting Murial’s mood: afraid, driven, sure. Your choices affect what is displayed on this side of the screen, which in turn affects the procedural content that sprinkles the permanent text: “Time passes“…..”I remember to breathe“…..”The summer afternoon seems like it could last forever.” The game gives you the option to highlight the generated content, so you can more easily discern your effect on the prose.
Reed’s hope is to encourage player agency without surrendering authorial voice. From his academic paper on the project: “I consider the minimum amount of PCG that might make a human-authored story computationally interesting but still authorially sound.”
I can see what Reed’s going for. Building a computer brain capable of generating a work of fiction comparable to The Sound and the Fury, or hell even Twilight, is most likely an impossible task. So it makes sense to keep most of the story grounded and let the computer handle small portions of text, phrases that lie on the periphery.
But just as the phrase implies, Reed’s satellite sentences feel removed from the crux of the narrative. The phrase “A pulsing pain behind my eyes” is awkwardly sandwiched between clear descriptions of the character opposite the protagonist. You get the sense that these satellites are paying attention to what’s going on, but they stick out from the permanent body of text. The resulting sensation is not one of freedom, but of interruption, as a consistent narrative voice is sacrificed for the dream of agency.
I don’t mean to sound overly dismissive of Reed’s work. Almost Goodbye is a fascinating project. If the generated voice were distinctly separated from that of the static narrator (Maybe in the form of some robot character. Other, less bad ideas are welcome in the comments), I think this problem of consistency could be overcome. In addition, I understand that this work is an experimental blueprint of sorts, and I admire its ambitions. I also like that Reed seems to understand the narrative potential of PCG. I’m not really excited by the idea that a game will rearrange itself if I play it twice. For me, the appeal of generated content doesn’t stem from a desire to play through one game repeatedly. What excites me is the potential to affect a larger narrative thread; the opportunity to collaborate with a game’s author in telling a story. That’s what replay value means to me. Procedural content generation in interactive fiction could perhaps feed this desire for collaborative authorship, but it’s got a long way to go.
The satellite sentences feel so disquietingly out of context that the only way I think they could be justified is if the protagonist was insane. Voices, etc.
Haven’t played the game myself, but what you say clearly shows it’s an experiment. If Reed keeps pursuing his goal, maybe further iterations won’t feel so dislocated. So, something to hope for in the future, perhaps?
Thanks for chiming in, Ketchua. That’s an interesting take. How about an asylum setting where the protagonist’s thoughts are part static and part procedurally generated? I do think there’s promise in Reed’s research, but I’m not certain that trying to disguise the generated text is the best approach. Maybe it would be more useful if the randomization were embraced as it’s own distinct voice and incorporated more intentionally into the story. I know that suggestion sounds really muddled and vague, to the point where it’s not very helpful, but I need to think more on this topic. If you’re interested in the game at all, you should give it a play. It only takes about ten minutes.
Well, the way they are now, there’s no disguising them. They’re painfully apparent (again, I have no first-hand experience with them, which will change soon). But I do agree with you – instead of trying to smuggle them in as “man-placed”, the author(s) should celebrate them for what they are.
Aaron Reed’s interactive fiction was pretty interesting outside of the whole satellite sentence concept. The premise fit well with the strength’s of Int. Fic. and I liked how the choices of who the narrator spoke to first had actual consequences, albeit to a limited extent.
I do, however, disagree with one point you made. Personally (although perhaps I may be wrong on this), I feel that the importance of procedural generation is not so much in the replayability of an interactive medium or game. Instead, it is the unpredictable divergences from player expectations that make procedural generation so unique.
You mentioned Dwarf Fortress which is a perfect example. Although the brothers who developed the game know all the lines of code they inputted into the game, they often admit to being surprised at the resulting scenarios and patterns that emerge from the stochastic properties of their game. In the context of barebone rogue games like Net Hack, perhaps the main appeal of procedural generation is replayability, and that isn’t a bad thing at all. But what is most exciting about procedural generation is one step deeper than simply replayability which I think is too broad of a term. (At the same time I admit, I might just be describing one aspect of what makes proc. gen. replayable, thereby making my whole statement redundant and worthless.) Anyways, just my thought.
Sam, I agree with you. Sorry I wasn’t more clear in the post. What I was trying to say was that the aspect of procedural generation that interests me the most is the promise of alternative outcomes, not replayability, and that folks who use procedural generation to try and make a game “replayable” might be missing the point. (Though now that I retype that idea, I’m thinking I might be comparing two sides of the same coin). The idea that my play experience can be different from another person’s is enough for me. I like your point about games defying player expectations. But then I guess you don’t necessarily need procedural generation for that. You can surprise everyone with a purely authored experience if you do it well enough. I think, your pointing to the DF brothers as an example suggests that PCG is exciting because it can break off from a game author’s original intent and develop its own distinct voice, which is pretty wild when you think about it.
Oh ok my bad. But speaking of awesome procedurally generated games, I was reading an article on Spelunky on The Verge website(http://www.theverge.com/gaming/2012/6/22/3106547/spelunky-the-everlasting-platformer)
Seriously if Derek Yu ports it to Steam/PC I would throw so much money at him he has no idea.
That was a great article! They really asked all the right questions. Have you played much of the freeware version of Spelunky?
Yea I played the original spelunky for a good bit. But yea, I was surprised the verge had an article on spelunky, they usually just review tech stuff