January 23, 2004

Reflections of a Larger Issue

by Andrew Stern · , 8:58 am

Our making, not telling terminology debate is a reflection of a larger issue, often framed as game vs. story or ludology vs. narratology, but I’d rather frame it as high-agency vs. low-or-no-agency. Let me quickly state, I’m not saying I don’t like stories or experiences without agency — I love them! I consume tons of books and movies and comics and music — but they’re not the new form I and many others are envisioning here… I think “gamers” or “ludologists” often have a distaste for interactive stories in their current forms not because interactive stories are not “games” per se, and not even because they’re often text-based instead of visual, but because the interactive stories built to date don’t have much agency. I feel the same way. (Go here for more on story vs. game.)

If you’re a developer or critic shying away from all-out-advocating of high-agency experiences, perhaps experiences leaning more towards the “telling” end of the spectrum are just what you prefer, or you don’t want to advocate / prefer one form over another, that’s fine. But another reason I could imagine is because high-agency experiences are really hard to design, and even harder to implement. But a lot of us believe they’re an extremely promising future for interactive entertainment, a holy grail that people want, evidenced by the most popular interactive experiences out there, e.g. the Sims, GTA3, which have high agency relative to others. Many of us have faith (e.g. informed by experiments over the last 15 years, often with mixed results) that high-agency interactive stories are theoretically buildable, with enough time and effort.

I’m not intending to attack anyone with this post, by the way, I’m merely stating my position on the directions I’d love to see developers and critics pushing towards.

19 Responses to “Reflections of a Larger Issue”


  1. nick Says:

    Perhaps they were peripheral to the earlier conversation, but I wouldn’t really frame my comments about “story” and “telling” as being part of the game vs. story, ludology vs. narratology, or high-agency vs. low-or-no-agency debate. They are just suggesting that if you do want to make stories interactive, or make the telling of stories interactive, it helps to distinguish what is being told (the content plane of the narratve, a.k.a. the story) from the telling of it (the particular narration). Then you can decide where you want the user to have agency (one or both of these levels) and to what extent.

  2. michael Says:

    One of the distinctions Andrew and I make is local agency vs. global agency. The player has local agency when the world responds to her moment-by-moment actions in a rich and satisfying way (e.g. characters change their behavior, speak to you, machines turn on and off etc. as you, the player, do things in the world). The player has global agency when the global shape of the experience is determined by player action, but not in a simple-minded (and easily unmasked by the player) branch pointy way. So, in the context of interactive story, global agency would mean that the final ending of the story, and the particulars of the arc that lead to this ending, are determined by what the player does, in a smooth continuous fashion (no branch points in some pre-authored story graph), and that, at the end of the story (one playthrough of the experience) the player can understand how her actions led to this storyline. Perhaps then she’ll be intrigued to replay the experience to see what other stories can happen. The ideal high agency interactive drama would have both local and global agency.

    In Facade we were attempting to build something with high total agency (both local and global). We ended up with more local than global agency, though we’ve done more with global agency than I’ve seen in other experiences.

    So Andrew’s probably framing the story vs. telling argument within the high-agency vs. low-agency debate because a high-agency telling experience (variation in the narration) would inevitably have less agency than an experience with high local and global (variation in the story) agency.

  3. andrew Says:

    Yes, thanks for that clarification of my point. I have to admit I find it a bit odd to think about player agency in the narration only, not of the plot itself. Perhaps someone could expand a bit on this idea for me. This would mean the player could interact with the narrator him/herself, but not having any effect on the plot of the story being narrated? What would be examples of affecting the narration but not the plot? When I think of the idea of a person telling me a story, and I interact with them, I imagine primarily wanting to affect the plot of the story being told to me (e.g., “wait, I want the princess to eat the frog! make her eat the frog!”). Otherwise, I suppose we are talking about affecting the pacing, style, etc. of the narration?

  4. michael Says:

    Actually, I think almost all (perhaps all) examples of story in contemporary games are examples of effecting narration without effecting plot. The plot is fixed, generally dispersed through a space, and, as the player navigates the game world, they experience the linear plot. But the exact details of narration, details of battles, exploration, confusions solving puzzles, etc., depend on player interaction. In the IF community as well, the story is generally fixed (I acknowledge that many IFs aren’t necessarily concerned with story per se – I’m talking about “story-based” vs. “puzzle-based” IF). I think we’re the odd ones for thinking about agency at the plot level – local agency without global agency is the norm.

  5. B. Rickman Says:

    I see “player agency in the narration” as the primary form of interaction in a simulation. When you’ve got an internally represented world filled with triggers for the user to find, the user experience is one of discovering or uncovering the story, but never really changing the story.

    But I also don’t think the user is much capable of distinguising between what is being told and how it is told, as Nick suggests. I know he’s talking about something the author should be aware of, but many times the author tends to confuse himself with the user and forgets to pay attention to “the how” and “the what”. The experience of a story is the experience of how the story is told. This, to me, is the reason why so many story generating projects have such a hard time creating “interesting” stories (e.g. Brutus), because they are oriented on “the what” of the story and not “the how”.

    Getting back to simulations, if the story depends on how it is told then why spend so much effort on the internal representations? If the user doesn’t witness some event, why does it need to even happen?

  6. ian Says:

    Michael — I very much like the way you describe local vs. global agency.

    As for changing the story, I believe (fundamentally in fact) that the user always changes the story, but the way that change manifests implicates not only the rules, but also the player’s mind. Non-interesting stories are inevitably ignoring the player.

    Michael will hear more about this from my very maw next week :)

  7. nick Says:

    The local/global distinction is a great one to make, but it isn’t the same distinction that I was making. I think distinguishing the levels of story and narration can also help to provide some insights here.

    Narratologically, the things that happen are part of the story (the content plane) and the way these events are told makes up the narration. In IF, I direct a character to do things within a simulated world. I can make the character jump, pick up objects, and talk to other charcters. These are events. They occur in the world. They may be meaningful or not, but they’re part of the story in this particular narratological sense. So the interactor in IF is controlling the story whenever the interactor issues commands. (The long version of this argument, along with some other applications of narratology to IF, is in my article “Toward a Theory of Interactive Fiction.”) So, these alterations in story may not have important causal effects on what happens later, but that’s a separate dimension. You might affect the story in trivial or profound ways, just as you might affect the narration in trivial or profound ways.

    The interactor in IF also controls the narration, usually in some less interesting ways, e.g., by typing “superbrief” so that no descriptions of areas are given or typing “verbose” so that descriptions of areas are always printed. This is not usually a very important aspect of IF; it is included for convenience, not narrative effect. There are other, perhaps more interesting examples of interactive works where the interactor can only control the narration. One is the Infocomics series done by Tom Snyder Productions and published by Infocom. In these, the story is fixed. You get to choose the character or group of character who will focalize the narration of the story, and you make several choices of this sort along the way. Perhaps a cleaner example is Composition No. 1 by Mark Saporta, which consists of 150 loose sheets that are to be shuffled into any order and read. The events that happen in the story of Composition No. 1 do not change for any total reading of the work. I believe that their temporal ordering and causal relationship does not change, either. What changes from one reading to another is the order in which the reader learns about the events.

    One issue here is that the casual use of the term “story” is not the same as the use of the term within narratology (when used alongside “narration.”) So, in the ordinary, casual use, if two people told us two versions of Little Red Riding Hood with slightly different events happening in each one, we’d say that of course they both told the same story. However, the content planes of these two versions of Little Red Riding Hood are not the same.

    “John died as a result of eating a ham sandwhich” has the same story (content plane) as “John ate a ham sandwhich, and, as a consequence, died.” But those are two different narrations in which the crucial information is given in a different order.

    “John died as a result of drinking milk” and “John died as a result of eating a ham sandwhich” are not the same story, on the other hand.

  8. andrew Says:

    Michael writes: But the exact details of narration, details of battles, exploration, confusions solving puzzles, etc., depend on player interaction.

    Brandon writes: I see “player agency in the narration” as the primary form of interaction in a simulation. When you’ve got an internally represented world filled with triggers for the user to find, the user experience is one of discovering or uncovering the story, but never really changing the story.

    I always viewed the details of battles, the solving puzzles, or the manipulation of objects in a simulated world not as variation in narration, but instead as not-so-profound or sometimes trivial variation in the plot itself (the content plane, as Nick put it). Actual events that occur, e.g. I uncover secret door B before I uncover hidden treasure Y, I move this lever, I attempt this key on this door, I make 10 points of damage with a sword, I construct an object in SimCity, are not narration, they’re small to minute details of plot. Yes, they’re being narrated to me by the fact that something has to tell me they’re happening, but the variation itself is in the content plane.

    Michael, Brandon, if I’m interpreting you correctly, you’re saying that when the story is fixed on a high level (linear), then trivial variation of details of the unfolding of the story (e.g., the details of puzzle solving) aren’t considered plot at all, and get lumped in as narration, because they’re not affecting the plot on a high level? I don’t know, that seems misplaced to me; they are plot, just not-all-that-meaningful variations in plot. (One might ask, why are they there then? The act of solving puzzles can be fun in and of itself, especially when performed in the context of / immersed in a story; and causing even superficial variation in plot is pleasurable to some; sometimes it offers the illusion of more meaningful global agency, too.)

    Brandon, in the case of simulation, since often the sum total of the experience doesn’t much hold together as a story proper, actually I wouldn’t say you’re “never really changing the story”, since there is no story there in the first place. (Unless we’re calling a non-well-formed sequence of events a story, which could only be called a very bad story.) Or, if it turns out the events in a simulation do hold together as a story (a subjective measurement, of course), then in fact you are creating the story in real-time. The space of all possible stories was bounded by the potential of the simulation, but within that space, you’re creating a story.

    Nick’s examples (“superbrief”, the two ways of describing death-by-ham-sandwich, Infocomics) seem to me to be the only examples here of variation in narration. Hence my feeling that pursuing player agency in narration seems like a less critical thing to focus on.

    The Composition No. 1 example is getting pretty close to what hypertext fiction is, in that you’re reading fixed pages in varying orders; but couldn’t that be interpreted by the player / reader as variation in the plot, not variation in the narration? I’d argue the former is possible. That is, even though the content of the pages are fixed, by viewing subsets of them or viewing in them in certain orders, aren’t we effectively creating plot variations? Because the reader understands the plot in real-time, and the order you receive the information can change your understanding. Perhaps not terribly meaningful variations in understanding, but again, that doesn’t mean we should not call it plot and instead call it narration… (Okay, I suppose if the content of each page was very explicit in describing its intended temporal and causal ordering, allowing the reader / player to understand a single plot no matter what order the pages are read in, then this cannot be interpreted as plot variation, and is merely narration variation. But often an author wants to allow varied interpretations, temporal and/or causal, and therefore we could call that plot variation, yes? No?)

    Brandon writes: Getting back to simulations, if the story depends on how it is told then why spend so much effort on the internal representations? If the user doesn’t witness some event, why does it need to even happen?

    I think a satisfying experience depends both on what is told and how it is told. (What-is-told always goes through how-it-is-told, but that doesn’t mean all we need to do is focus on how-it-is-told.) I agree that story generation research often ignores how-it-is-told, but to date these systems also tend to produce not interesting enough what-is-told. To get interesting variation in what-is-told (the plot), especially ones meaningfully affected by player interaction, we need substantial internal representations. That said, I like to believe there’s tons great stuff one could do without a lot of internal representation, and your The Dr. K Project suggests that is possible. However, as we’ve mentioned in the past, with Facade we (er, especially I) hoped that good design would get us a lot further than it actually did; but after years of trying to squeeze water from a handful of moist pebbles, I’m much more convinced that this is going to need more sophisticated internal representations that afford more generativity.

    (To extend the rocks on the beach metaphor, and quote our GDC paper, systems need to “offer open-ended sandbox-style play in which the system knows how to make sandcastles, and tries to collaborate with the player to do so.”)

    Now, would you call The Dr. K Project (which presumably most of readers haven’t seen, but can be read about it here) purely variation in narration, and not story? If you did, I’d imagine you could because the experience is actually about interacting with a narrator? But even then, I would argue that a story (the content plane) is being created in real-time, where the narrator is both the narrator and character in the story, so in fact you are varying what-is-told.

  9. nick Says:

    Nick’s examples … seem to me to be the only examples here of variation in narration. Hence my feeling that pursuing player agency in narration seems like a less critical thing to focus on.

    It’s easy to advance a contrary argument. Since allowing agency in telling/narration hasn’t been widely explored, maybe there are a lot of interesting things that can be done there. Early work in plot variation (e.g., TaleSpin) didn’t have interesting results either, so applying the same perspective there would have led us to think that we shouldn’t focus on plot variation as a critical thing.

    The Composition No. 1 example is getting pretty close to what hypertext fiction is, in that you’re reading fixed pages in varying orders; but couldn’t that be interpreted by the player / reader as variation in the plot, not variation in the narration? I’d argue the former is possible.

    Is your argument about Composition No. 1 specifically, Andrew? I disagree that varying the order of pages varies the plot, and I’d suggest that it’s probably more useful to discuss this point after reading Composition No. 1. I think the variation in narration that can occur between readings of that work is interesting. Even if plot variation does occur (I could possibly be persuaded that it does, but my sense is that it does not), the changes in the content plane are not where everything interesting happens.

  10. andrew Says:

    Nick, yes, I certainly wouldn’t want to say it’s not interesting to explore variation in telling/narration, in fact ideally an interactive story system would offer high degrees of both variation in narration and content. Your point about Talespin is well taken.

    My whole goal here with all this argumentation is to try to understand (and advocate to others) which research directions are urgent to pursue for offering players rewarding experiences. I.e., what do people want a lot (not “the most”, not the only thing, just a lot). This is informed by anecdotal evidence, observing trends what kinds of experiences tend to be more popular and rewarding than others, and my own personal feelings and reactions. And by the results of interactive / generative story research and theory.

    (An aside — I find it’s often tricky to take a stand on an issue without occasionally coming across as a touch arrogant or dismissive, which is perhaps occurring here. It’s a fine line, something I’m finding one gets better at with practice. :-) The potential for misunderstanding is especially high when we’re debating the definition of terms, e.g. what is narration / telling, what is not. However I’m one to take the risk and err on the side of pushing buttons, and enjoy and appreciate it when others do the same.)

    Is your argument about Composition No. 1 specifically

    No, the more general form of reading fixed lexia in varying orders, e.g. what some hypertext fiction is. That it’s possible for such an experience to feel like plot variation to a reader, not narration variation, I’m guessing depending on the flexibility or ambiguity of the content of the lexia themselves.

  11. B. Rickman Says:

    I suddenly feel a bit fenced in by the terminology being used here, not so much by Nick’s narratological concerns but by the terms of high/low agency. I don’t want to outright disagree with the concept of “agency”, but I want to poke at it a bit to see if it moves.

    To paraphrase Michael, the player has agency when the world responds to their actions in a satisfying way. Agency isn’t big on meaningfulness; a player can have agency when playing with a collection of wooden blocks, in that the player can be satisfied by the experience without any real meaning taking place. (Blocks that have been glued to a table do not offer agency.)

    But does playing with blocks actually satisfy the definition? Do the blocks respond in a satifying way, or are they simply obeying the laws of physics? I think perhaps since the blocks lack any real mechanism for response, they don’t really allow for any functional agency.

    When a player interacts with a narrative generating machine, he has agency when the machine reponds to his actions in a satisfying way. But what if the machine responses are simply the result of a set of fixed rules, such as when the narrative elements are constructed from internal representations. Is this functional agency? Perhaps.

    If the machine operates with fixed rules and internal representations, then there is a point at which the user’s interactions may continue to be satisfying, but cease to be meaningful, once the machine has presented all of the interesting semantic relationships at its disposal. But would you place value on the agency offered by such a machine?

    Hopefully it is clear that I’m trying to split meaningfulness from agency. It isn’t enough to say that agency does or does not depend on meaningfulness.

    Now, suppose you built a narrative generating machine that offered open-ended agency to the player. The player may now engage in satisfying actions indefinitely. But what about the meaningfulness of the activity? It would seem that with truly open-ended agency, there are no semantic relationships to be revealed to the player, because everything is up for grabs. Agency without meaningfulness.

    I’ll stop there for now.

  12. nick Says:

    Brandon, your point is a good one. Being able to make substantial (noticable, large-scale, etc.) changes to the simulated world does not equate to being able to mainipulate the world in a way that is personally meaninful to the interactor. What happens at the formal level can’t tell us everything about what happens at the levels of reception (how does the interactor understand?) and operation (why does the interactor take particular actions?), although it probably relates to that in important ways.

  13. andrew Says:

    In an earlier discussion about what is and isn’t agency, I suggested that any effect the user has on the world, however trivial, transitory or ultimately meaningless — such as pressing a button that actually doesn’t do anything — could technically be thought of as a teensy bit of agency, since you are having an actual effect on the world for at least a brief moment; I also brought up the distinction between local and global agency. In such a definition of agency, you are concerned about how much agency you have, it just or local or is global, etc.

    But Michael suggested it’s really only useful to think of agency in terms of effects that have some importance to the player, such as the phenomenon of the player successfully expressing a substantial intention through action. I’d agreed that is a more useful definition than my initial suggestion. I understand “importance” to be something that matters, that has at least a bit of “meaning”, to the player. That is, action has to reach a threshold of meaningfulness before we bother calling it agency; otherwise it’s just something that happened.

    So, going with that, I’m not sure you can separate agency from meaningfulness. Maybe “satisfying” (the term Michael used this time around) is an imperfect requirement for deciding if something is agency. Having a substantial, even satisfying, effect on the world that somehow actually turns out to be “meaningless” — e.g., no important semantic relationships get revealed — is that just an extreme example of pressing a button that stays permanently depressed, but has no meaningful effect on things?

    If so, what we’re saying is that it is possible to have a “substantial” effect on things, but not necessarily a meaningful one. Who decides what is meaningful, the author of the system? — who presumably established the semantic relationships present in the first place?

    But if something is satisfying to the player, could it also at least a bit meaningful to her, simply because it’s satisfying? How do we decide when something is “meaningful”?

  14. B. Rickman Says:

    I like the description of agency as satisfying (and not meaningful) because satisfaction is something you can definitely ascribe to a person, whereas meaning isn’t so easy. Certainly an observer watching someone manipulate a machine can find meaning in the player’s actions, but that is because it is in some ways easier to find meaning in the actions of other people. But the meaning identified by the observer is a meaning between the observer and the player, not between the player and machine. The observer can ask the player, “Did you mean to do that?” or, “What door does that key open?” and the player can answer.

    In contrast to this, if you watch one machine manipulate another machine, it is harder to find meaning in either relationship (machine-machine and observer-observed). For example, a bot programmed to play Asheron’s Call. But at the same time, the observer is likely to find some meaning in how the bot was programmed; the programming reveals the ways in which the programmer understood the bot machine and the game machine. (e.g. “He doesn’t know about the teleport on the third level.”)

    Anyway that gets a bit confusing. What I’m really trying to get to is this: if the player has agency with a machine, then can the machine have agency with the player? Can the responses of the player be “satisfactory” to the machine?

    I think this is hinted at in IF. Take a simple puzzle where the player needs a key to open a door. Where the player wants to open the door (action), the game wants the door to be opened (state). When the player finally responds with the correct action, the game can be satisfied with the outcome. That seems to match the definition of agency (feel free to disagree).

    Meaningfulness comes into the situation through the single or multiple ways by which the door can be opened. If the door can only be opened one way, with one specific key, then the only semantic relationship that can be shared by the machine and the player is the one in which that key opens that door. And that’s the trick with meaning: it has to be shared, it has to sit inside of a relationship. If no other action will open the door, then there are no other possible meanings (within the player-game relationship) for that puzzle.

  15. Walter W. Kim Says:

    Agency is ultimately only a capacity to cause. What matters to us is our subjective experience of causing or bringing into effect. We require a meaningfulness in order to be brought into that self-causing or self-bringing into effect. Satisfaction from action is a prospect of engaging one’s agency that can ultimately be defeated, but the meaningfulness is always what must provide the grounds of being satisfied or disappointed.

  16. B. Rickman Says:

    Walter, what are the boundaries of your agency-as-capacity-to-cause? It seems to encompass both the player and the machine, to such a degree that we can no longer say, by your terms, that the player has agency. And yet, this capacity to cause remains subjective to the user’s experience, so it is only important that the user’s actions appear to be the cause of some effect, even when they aren’t. I don’t see how that interpretation of agency gets you anywhere.

  17. Walter W. Kim Says:

    Agency and causality are thought to be related somehow in terms of the development of their concepts in humans. Did our experience and subsequent concept of agency give rise to the concept of the cause? Some philosophers think so.

    In the abstract sense I don’t see agency and causality as any different. What we want to use ‘agency’ for, though, is to imply that there is some ‘internal’ goings-on underlying the action, something less direct than you hitting my knee with a hammer and me kicking. What we have is the sensation of being the cause, which we are (or can be tricked into thinking so), but in philosophical terms we certainly aren’t ‘originaters’ of cause, as in, “we defy the causality of the universe every time we decide and do something.”

  18. B. Rickman Says:

    Walter: “Did our experience and subsequent concept of agency give rise to the concept of the cause? Some philosophers think so.”

    It may be laziness on my part, but I fail to see how agency’s dubious heredity has much bearing here. Nor do I see a need to seek out internal causes for action, let alone universal causality. If the only purpose of agency is to delude someone into believing they are a cause, there’s no real need for a concept as specialized as agency. And yet…

  19. Grand Text Auto » GTxA Symposium: Future Directions Says:

    [...] hat, and drama before that — in addition to that potential, digital fiction can be a very hard, very challenging, and therefore very frustrating and rewarding prob [...]

Leave a Reply

Powered by WordPress