May 12, 2003

How to Destroy Possibilities

by Nick Montfort · , 5:56 pm

I’ve been corresponding with IF author and theorist Emily Short recently about an issue that relates to all sorts of digital practice and interactive design — perhaps to all sorts of art-making. One thing I’ve heard over and over in discussion of the design of virtual spaces, computer games, and other sorts of works is that creators have to constrain the interactor, limiting a world of possibilities to just a few so that the experience can be controlled and contained.

I don’t like this assumption. It harkens to the “Pirates of the Caribbean” approach: put the docile participant in a little car on a track and bring them the experience of the space in exactly the order you want. Of course, some might say that this destruction of possibility is exactly what Oulipian techniques, of which I am so fond, enforce. Italo Calvino talked very directly about this idea of eliminating possibilities, specifically in relation to the computer generation of literature, in his lecture “Cybernetics and Ghosts.”

I wonder if this is a problematic or at least unexamined assumption. Why don’t people ever seem to think about interactive design as building up possibilities from what are, initially, no possibilities at all? (I’ve tried to think in this way, at least at times, but I can’t say that a masterwork has resulted from this approach.) It seems to me like a world that is constructed perfectly, is so complete and coherent that it could not possibly contain anything less or anything more, would present exactly the right possibilities. Perhaps Will Wright has already accomplished this with Sim City and its progeny … perhaps Andrew and Michael have done something similar with Façade. It seems that those creating many different sorts of work could fruitfully take this perspective. Authors, artists, creators: Do you find yourself more often wondering “how can I force the interactor do this cool thing?” or more often thinking “what great possibilities can I create for the interactor?” And do you feel good about that? Has that perspective always been a fruitful one?

17 Responses to “How to Destroy Possibilities”

  1. Emily Short Says:

    In my experience of writing IF, actually, I wind up doing a little of each. The simulationist aspects of the game are all about creating possibilities: making the world model flexible and responsive, both as an entertaining toy and so that the player can find a variety of different solutions to the puzzles. The richer these combinations, the happier I am.

    The narrative level of a game, though, tends to require limiting the player again. Even where I’m writing a multilinear story — the player can choose to work with one of two opposing factions, for instance, and the choice affects the rest of the game — I find myself thinking a lot about how to make sure the player is reaching the pieces of the story in a comprehensible order. And this process often does involve trying to put one piece of information where it has to be seen before another, or making one experience a necessary precursor to the next.

    I hardly think of the player’s range of action as a Pirates of the Caribbean-like track. One metaphor I’ve sometimes used instead — which I have also heard from other IF authors on occasion — is that of education. The player is offered a certain range of interactive possibilities: talking to other characters, for instance, or using the magic system of the game, or doing research about the game setting. Once he has learned the elementary lessons, and demonstrated his understanding by a successful interaction, he’s ready to go on to experience the next portion of the game. But if he reaches that portion out of order, he won’t know enough yet to be able to interact with it meaningfully.

    The range of options in a work of IF isn’t just a function of what you’ve put into the code. It also depends on what the player perceives his options to be. By giving him some simple choices early on that demonstrate the interactive range of this particular game, you equip him to approach the later stages knowing, for instance, that this is the sort of game in which he *can* expect to make a choice about which faction to join, and moreover he knows how to communicate that choice in terms the game will understand.

  2. nick Says:

    I don’t want to hog the comment board, but your point about the interactor’s level of awareness, and how imporatnt the perception of different options can be, is a particularly good one. It’s somewhat easy to lead someone directly (via some sort of blatant in-game instruction or hint) to see what the possibilities of an interactive fiction work are. It’s more difficult for these possibilities to be made obvious as the interactor’s understanding of the world grows.

    The idea of “education” happening in-game makes me think about the way a good literary riddle can lead a listener along to see things differently, from the new system of the riddle’s world. The way the riddle is written can allow that understanding to come about in different ways, or not to come about, if it’s written in a way that is instrutable. I do think the workings of the riddle, and your idea of educating (rather than confining) the interactor, are important concepts for the design of IF and similar works.

  3. Sean B Says:

    Generally, I think modern IF has a tighter coupling of puzzles and story than we tend to see in commercial games, but I think in truth, it’s often a coupling of the *goal* of the puzzle; it makes narrative sense for the player character to be trying to achieve this goal at this point, but the manner of solving often isn’t that story-connected.

    As a result, IF authors do end up implementing alternate solutions to puzzles, even during beta. They primarily have to simply balance the alternate solution’s difficulty so it isn’t too easy, although they are also compelled to make sure the puzzle can’t be solved too early in the narrative.

    It seems to me that this is a very small example of authors starting with constrained possibilities and building up more. Unfortunately, just on a very small scale.

    Certainly some of the public comments that have been made about Deus Ex 2 have been along the lines of ‘in Deus Ex, we hardcoded 3 solutions to everything; we’re going to try to make it more open this time’. Sadly, I don’t know what sort of magic electrons they’re putting in their computers to achieve this. But at least some people are thinking about it.

    SimCity is a great design of a rich possibility space for the player, and it may have even evolved as an accretion of abilities for the player, but since building in a rich dramatic experience to SimCity is as unsolved as the general interactive story, I’m not sure what to conclude about it.

  4. andrew Says:

    I wonder if the “Pirates of the Caribbean” approach to designing interactive experiences has been made all the more obsolete by the recent rise of rich virtual worlds as gaming environments, e.g., Grand Theft Auto 3 and its ilk. I think the bar has really been raised in the last couple of years. (Not that I think that today’s virtual worlds have figured it all out; see the discussion in the “introductory post” thread on the problems with virtual worlds.)

    In the best work out there that I’ve come across, typically games but also in some net art, the “how can I force the interactor to do this cool thing” approach has been succeeded by “wow, look at the things I can do, and they’re all pretty cool”. (Or at least actions that are cool to gamers.)

    Generally speaking, it’s about expression – giving the player the means to express themselves in non-rigid ways. For example, if you give the player a very broad set of powerful actions – say, at least 30 – all of which the player can do at any time, and all are effective most / all of the time (ie, you don’t arbitrarily disable player actions just to make the implementation easier), then I think the player will feel reasonably unconstrained.

    Frankly I still haven’t found an interactive experience that gives me a set of expressions that I want – but I’m sure it’s possible. And I don’t think you have to be a huge team of people with millions of dollars to pull this kind of thing off – I think it’s possible for small groups, even individual authors / artists, to achieve this level of richness in their work. Essentially I’m suggesting that artists need to make their work deeper. When deciding how to invest your time and energy into a piece, perhaps tradeoff breadth for depth.

    At the keynote of the TIDSE03 conference last March, Chris Crawford urged designers of interactive experiences to start by thinking about the verbs, not the objects, that you will give the player. Think about what the player will actually be able to do, and then design the objects, environment, story, etc. around those verbs. (He also suggested that, for a decent interactive story, players will need a minimum of 1000 verbs, ideally 2000-5000. Yikes.)

    Nick, you mentioned Façade in your original post; in that project our goal is to give the player a non-trivial new set of things to do, things you don’t usually get in a game, such as socializing, taking sides in arguments, flirting, saying provocative things, all performable through a natural language & gesture interface. (Internally, everything you say or do gets turned into one or more of ~40 parameterized “discourse acts”.) Of course, once the project is done and released, players will have to be the judge of whether this set of actions is satisfying or not, or more specifically, if we did a good enough of a job implementing these actions and the system’s reactions.

  5. Nicolas Szilas Says:

    It seems that the idea of story is inherently linked to the idea of contraints: most of situations that can be generated by a simulated world are not narrative at all, so we have to apply constraints to obtain a satisfying narrative result. But if the constraint is applied to the interactor, then we loose the very reason why we wanted to do interactive stories: eventually we tell one given prewritten story!

    It is not the interactor that we should constraint, but the computer! Forcing it to produce events that makes a good story, according to what the interactor chooses to do. The difficulty is that the constraint is generic: it is not anymore about constraining to a specific path, but constraining to a set of satisfying paths. Most research on Interactive Drama is based on this idea of defining a kind of generic criterion for a good story.

    This way, we might manage to give the interactor some choice (let’s say 30 at each turn…), yet letting him experience a story. I am not sure we need for this 5000 verbs!! Most of stories are a combination of a limited number of basic elements… but this is another issue!!

  6. Athomas Goldberg Says:

    At the risk of sounding reductio, I believe it would be difficult to conceive of, let alone implement an “uncontrained” interactive experience (or any other form of artistic expression) While constructive games like SimCity are extremely open-ended and generative in nature, there is a very specific and limited set of tools at the interactor’s disposal and it is the nature of these tools and the interface in which these tools are used that defines the interactor’s experience. The same holds true for GTA3, Facade or any other video game, IF, board game, etc. What makes these games seem “limitless” is that they are “complete” – in other words, the typical interactor’s expectations of agency are satisfied by the set of choices available to them *within the context of the experience*

    I think quantitative measurements of “agency” are meaningless. Whether 5000 choices or 5 (or none) is adequate at any given moment is wholly dependent on the nature of the choices and the interactive experience in which they are made relevent. I think that creating compelling “structured simulations” in which the rules by which a player interacts with a system are generative of a specific form or arc of experience, is possible provided that the experience is complete. Determine what is necessary to achieve agency within a given experience *is the art*

    No matter how open-ended and unconstrained we’d like to make a particular interactive experience we can not avoid imposing constraints on the interactor. Even the decision to make the experience computer-mitigated, is itself a constraint that is not devoid of meaning. What made the Oulipo brilliant was that they understood that “established” conventions imposed constraints whether the author was aware of them or not, and what they set out to do was explore the meaning of those constraints and there alternatives through deliberate rather than arbitrary construction.

    I would like to write more, but I have to catch a plane for E3. I’ll check back during the week to see how he discussion progresses.

  7. Andrew Plotkin Says:

    Hrm. Okay, here’s my take. (Warning: a lot of this has already been

    said in comments, but I feel like putting explicit examples to it.

    Makes me feel like less of a theorist and more of a game designer.)

    NM’s direct question: “Why don’t people ever seem to think about

    interactive design as building up possibilities from what are,

    initially, no possibilities at all?”

    I don’t think this captures the distinction between SimCity/GTA3 and

    “Pirates of the Caribbean”. (And that latter label itself doesn’t

    usefully capture the range that we see between “good interactive”

    adventure games and “boringly constrained linear” adventure games. But

    that’s a different topic. (I see Emily has asserted this also.))

    A simulation approach gives you a well-defined, relatively small range

    of action, but it’s tightly packed: anything you do has a meaningful

    consequence. (Of course most of those consequences are bad, in a game

    like SimCity or GTA. It’s much easier to crash your economy or drive

    off the road into a tree than it is to make your city or your bankroll


    An adventure approach has a fuzzily-defined, huge range of action —

    apparently as large as real life. But it’s sparsely filled out. On the

    up side, every twig is hand-crafted, so when you *do* make progress

    it’s very interesting.

    “I wonder if the Pirates of the Caribbean approach […] has been made

    all the more obsolete by the recent rise of rich virtual worlds as

    gaming environments, e.g., Grand Theft Auto 3 and its ilk.”

    No, and here’s the rub: the simulation approach is *terrible* for

    generating narrative. Describing the rise of your SimCity, or the

    flourishing of your civilization in Civ3, is always a boring story.

    “Well, the area down near the waterfront got polluted for a while, but

    I pulled it out… then there was this volcano downtown, heh heh…”

    It doesn’t have a thousandth the narrative pull of a disaster movie —

    even a badly-written badly-acted volcano-erupting-downtown hack

    disaster movie.

    GTA3, and nearly every other action-adventure and CRPG-adventure game

    in existence, uses a simple trick: missions. You alternate simulative

    mission challenges with non-interactive cut-scenes. All the game

    happens in the missions; all the narrative happens in the cut scenes.

    I pulled a lovely trick in GTA:VC. In one mission, I had to shoot up a

    shopping mall and then escape through a cordon of irate police cars.

    My idea: steal a helicopter first, park it right next to the mall, and

    escape via the air! The simulation engine of GTA is powerful enough to

    allow this. It was very dramatic.

    In a text-adventure style game, the author could — would — have made

    good use of this outcome. A strongly-written paragraph describing the

    cops waving their fists furiously on the ground. A thematic tie to a

    dramatic escape earlier or later in the story. Maybe a character

    flashback — that awful moment in ‘Nam, when the choppers arrived too

    late to save Fred!

    In the GTA game engine, there’s literally no way to implement these

    narrative features. There are too many ways to escape from the

    shopping mall. Even if the authors added code to check “Does the

    player enter a helicopter during mission 37?”, it couldn’t take the

    context sufficiently into account. (Maybe I escape from the cops on

    foot, and then, when the coast is clear, mosey across the island to an

    unguarded helicopter. The “dramatic escape” scene would look pretty

    corny then! Anyhow, what if I dive into the ocean and escape by

    speedboat? Seaplane? Pizza-delivery scooter? Do we add code to check

    for all of these? Do we rewrite the character flashback four ways?)

    Anyhow. This is a long way of saying what people have already said,

    that “the idea of story is inherently linked to the idea of


    The interesting other thought is that this is *not* a deficiency of

    simulationist game design. It’s a mismatch between the techniques of

    game design and the techniques of static fiction (novel, movie,

    theater). The latter are all *designed* for the case where a single

    sequence of events is portrayed. When you write a novel, you create a

    sequence of events, a set of characters, a scenic background, a mass

    of prose. Each of these is carefully honed to complement and tie into

    all the others. The more powerfully these elements bind together, the

    stronger we say the novel is.

    Trying to cut the events loose — make them unconstrained — while

    leaving the rest of the narrative structure intact is a fool’s errand.

    Computer adventure games don’t even try to do this. They take the

    “mission” approach. Text adventure games (and, to some extent,

    graphical adventure games) do a little better: they have a fractal

    tree of “missions”, where little actions get a little narrative

    response, bigger actions have more, and big scene-ending actions

    provoke lots of narrative. But it’s still the same old

    everything-bound-together narrative form inherited from static


    But what other structure do we *have* for narrative?

    (The only other one I know is the paper-and-pencil RPG, where

    *everything* is fluid, coinvented by the players and game-master. But

    that’s just as far from the single-player computer game as it is from

    the novel.)

  8. Sean B Says:

    But that’s just as far from the single-player computer game as it is from the novel.

    I don’t think that’s particularly true; modern IF derives rather directly from a very early attempt to implement the experience of paper-and-pencil RPG gaming inside a computer–Crowther said so himself. (CRPGs are a separate species which share that non-computer ancestor.)

    And certainly pen-and-paper RPGing is frequently held up as an existence proof that it is possible to tell a dramatic, evolving, interactive story; the debate AFAIK is whether it’s AI-hard or not. The mapping of DM & players -> DM & one player -> computer program & one player isn’t such a stretch except for that little detail.

  9. andrew Says:

    Andrew P. wrote: “the simulation approach is *terrible* for generating narrative”

    True, as I bemoan in the “introductory post” thread. I suppose what I mean by the “pirates of the caribbean approach becoming obsolete” is, that, as authors and players, until we can find a new approach (if one can be found) that gives us the best of both – the satisfying structure of narrative and the freedom/agency of simulated worlds – we’ll be forced to choose between the two. My guess is that players are going to gravitate (or have been already) towards experiences that give them more freedom/agency, at the expense of narrative. Hence the “pirates of caribbean” approach will continue to fall out of favor.

    Could it be that people want to have real, varied effects on their virtual worlds, more than they want to experience a well-plotted story? especially when there’s an abundance of well-plotted stories available in theaters and bookstores? (Again, I’m optimistic we can figure out a way to offer both…)

    Athomas wrote: “No matter how open-ended and unconstrained we’d like to make a particular interactive experience we can not avoid imposing constraints on the interactor. ”

    That’s true too. I’m not suggesting total lack of constraint; I’m just suggesting practitioners push to build systems that offer players a far richer set of expression than we currently see. If we did this, it would open up new and interesting “contexts for experience”.

    I like Nicolas’ suggestion – here’s my interpretation of that. Theoretically, let’s say we have a system where you can say anything you want at any time, and the system can actually understand everything you’re saying. This is crazy unconstrained input, right? No. The system (eg, an NPC) can decide what it will and will not respond to, or even better, how it will map these unconstrained millions of inputs into a finite set of inputs it has responses for. The constraints lie in how the system will respond to this huge range of possible inputs; the constraints do not lie in some sort of arbitrarily limited interface (eg, you’ve only got these 20 commands, or these limited set of tools/objects…)

    Maybe this matters most when we’re talking about language input; when we’re talking about physical action, it doesn’t feel quite as constraining to only have a finite set of options or objects.

  10. Dan Shiovitz Says:

    RPGs and IF do also share the characteristic that although there’s no single path to take forward, if you look back you do see a single path of action (ignoring save/restore and multiple playthroughs and so on). This initially doesn’t seem useful because the path hasn’t been particularly directed, but I think it’s not unreasonable to use information from the past to guide the plot in the future, and thereby make past events seem more significant and lend some shape to the overall narrative.

    The trick, I think, is giving up on trying to create drama for an individual event — I don’t think in a simulationist world you can say “ok, the player will have a cool escape from the mall in a helicopter” since they might instead sneak out the back and take a jet-ski. But you can say “ok, the player *did* have a cool escape from the mall in a helicopter — so now I’ll send in some more bad guys in fighter planes to give chase” or “so now at the end of the game I’ll put in another helicopter on the landing pad to set up a parallel”. It’s the same situation historians have to deal with: you’ve got the whole civil war to look at, so which parts do you

    pull out to make a narrative?

    Similarly, the player in SimCity deals with a variety of problems, and some they might handle trivially (add another airport to handle traffic) and some might be more complicated (a fire gets out of control and burns half the city). You can look after the fact and pick out which ones were most interesting, and make those happen again. Or be connected. Then the story is no longer “I fought a fire, and then the sewer line broke, and then there was an earthquake, and then there was a volcano in the middle of downtown” but “I fought a fire, and then a while later there was an earthquake, and eventually I realized it was all because of the lava forcing its way up in the middle of downtown, creating a volcano”. Or “I fought a fire, and then another fire, and another fire, and eventually there was a car chase and the police busted the serial arsonists.”

    This might not be a helpful reformulation of the problem; I suppose “interesting” may end up being just as AI-complete as anything else.

  11. Emily Short Says:

    …Chris Crawford urged designers of interactive experiences to start by thinking about the verbs, not the objects, that you will give the player. Think about what the player will actually be able to do, and then design the objects, environment, story, etc. around those verbs. (He also suggested that, for a decent interactive story, players will need a minimum of 1000 verbs, ideally 2000-5000. Yikes.)

    I think that’s an overestimate; in fact, I think having that many verbs might actually reduce the player’s sense of control. Yes, one should think about what the player will be able to do — in fact, this is the starting point from which I almost always begin a game design. “What aspect of the world does the player control? How will the interface work? How do I make it consistent, and how do I make it produce interesting results?” The first part of that is the simulation part; the second part is where the narrative begins to creep in.

    However. One of the things I’ve discovered in trying to write conversation-based IF is that playtesters become uncomfortable if given too many verbs, because they don’t know which they’re supposed to be using at a given time. They apparently feel a bit as though they’ve been put in the cockpit of a jumbo jet and don’t know what to do with all the buttons. The more sophisticated the parsing and the more subtle the world model, the more likely it is that the effects of the player’s actions will be opaque to him — that even after he’s pressed the mystery button, he’ll have no idea what it did. The player is robbed of intentionality just as much as if I only had one button that said PUSH ME. Hence the need for some balance.

    Naturally I realize that not everyone is interested in writing games per se, but, for what I want to do, the loss of player intentionality is a serious problem. What I am interested in doing is making a system that does recognize some event of major narrative importance (whether this is the solution to a puzzle in traditional IF, or a critical decision in something more story-driven) and respond accordingly, no matter by what simulationist means the player arrived there. In the GTA example, for instance, it wouldn’t have the ‘player gets in helicopter’ criterion, but it *might* have a ‘player gets away from waiting police in an exotic vehicle’ option.

    The critical difference between this and the ‘missions’ design — at least as I’ve seen that design implemented thus far — is that the player would (ideally) be led to perceive what the narrative options are, rather than being told them at the outset; and that there would be more than one possible outcome.

    Of course, this still requires that I have, if not a specific linear story, then at least a set of narrative possibilities in mind. But if these possibilities are moderately diverse, and if the means to reach those possibilities are *also* diverse, then the player does have a non-trivial degree of continuous control. And he’s able to build his particular story out of a range of, say, forty or fifty possible stories. It’s not infinite, but I think it is potentially satisfying.

    I think (assuming I haven’t misconstrued) that this is something like what Athomas is suggesting:

    I think that creating compelling “structured simulations” in which the rules by which a player interacts with a system are generative of a specific form or arc of experience, is possible provided that the experience is complete.

    Building the world to be self-consistent, letting the player discover the rules, and then teaching the player to use those rules to control the narrative nodes I’ve set up: this is how I envision such completeness.

  12. nick Says:

    I realized I was quite confusing in my original entry, not that this prevented an interesting conversation from taking place. I was actually thinking of a perspective on design/authorship/creation, not particular works, despite my mentioning SimCity and otherwise being specific.

    As Andrew Plotkin’s comments describe, the pieces of a more “simulationist” world can fit together in unexpected but pleasing ways, while this isn’t the case with all sorts of works. Some works let the player or interactor construct unusual, possibly unanticipated outcomes.

    But there’s also the issue of how you approach the design of any sort of interactive digital work. Imagine you’re creating Adventure, or Afternoon, or Superbad. You can take the perspective of adding possibilities where there are none, or you can take the perspective of removing possibilities where all exist. This relates to Mark Bernstein’s distinction between calligraphic and sculptural hypertexts – do you start with no links and add them, or start with all links and remove them – but I don’t think the question of the designer’s perspective is completely answered by a particular authoring system and what functions it enables, nor do I mean for the question to be answered by pointing to particular works that are in one or the other category. You could have designed exactly the same Adventure, Afternoon, or Superbad whether you imagined yourself removing possibilities or adding them. It might predispose you to certain sorts of works, but that’s at the next remove. My initial musing was about whether it’s interesting to think about a poetics or a design perspective as being possibility-creating or possibility-destroying.

    That seems to be an oversimplifying question, though. Certainly designers of any complex work have to think in both ways, and perhaps other ways, at different times. Emily’s concern for “balance” is an example of a more sophisticated perspective, I think.

  13. magy Says:

    I think the measure of engagement or immersion within an experience depends highly on the situation setup. Any game has a certain set of rules and no matter how big or complicated they are, users will eventually learn to work with them. I don’t mean to take an old example, but take Chess for example, there is a set of rules that constrains the players’ actions, yet many find the game enjoyable.

    Frustration occurs when the situation or interface creates an expectation for a hugh range of agency that is simply not allowed. For example, when you allow the user to type anything that he wants or dreams of, you are then telling him/her that he/she can essentially say anything and the game will respond. If that illusion is broken once, then the player will be frustrated and the sense of engagment breaks. Of course, there are ways around that, Andrew and Michael address that brilliantly, I think, in facade.

    I agree with Emily, my experience with giving too many actions/options to players has been negative. But I think that also depends on the game, the setup, and the player’s experience.

    RE: Writing

    During my acting classes, we learned a few techniques for narrative construction that I think may be very useful for interactive narrative. Instead of focusing on sequences of events that lead to a fable. We, as actors, broke down our characters’ goals and motivations into character beats and tactics. Instead of a scene being a sequence of beats or events, characters will choose a goal and tactics that achieve their goal, which is part of the scene goal. Characters will also change tactics depending on how well a tactic is working. This is very similar to the character-centric approach of screenwriting, I don’t remember if Mckee actually discusses that type of writing, but there are several books that describe that, especially playwriting books.

  14. jill/txt Says:
    An email yesterday announced an interesting new blog, run by Michael Mateas, Nick Montfort, Stuart Moulthrop, Andrew Stern and Noah Wardrip-Fruin: is about computer mediated and computer generated works of many forms, including intera…

  15. GamesFilter Says:
    How to Destroy Possibilities
    – grandtextauto talks about freedom in ‘interactive fiction’. Should developers see themselves as adding more freedom in where there is none, or removing freedom so that they can push players down a proscribed path? Some good discussion in the article …

  16. This is Not a Blog Says:
    Blogroll Addition
    Check out my latest addition to the blogroll: Grand Text Auto, a group blog that discusses digital narrative, computer aesthetics, and – of course – video games, and all with a nice academic background. Yummy. It’s very new, but the…

  17. Grand Text Auto » Credibility Currency Says:

    […] talk than deep one, touching on the idea of limits in interactive stories (reminding me of Nick’s first GTxA post), as well pointing towards procedurality. […]

Powered by WordPress