Types of Action and Types of Agency

I’ve been thinking again about actions and how they’re expressed, and how the communication of an action relates to player agency.

Let’s say, for now, that agency is the player’s ability to affect the world and story, and it depends, in turn, on whether the player can form a reasonable guess about the results of an action before taking that action. If the player cannot guess or does not care where the action will lead, there is no agency; the player is providing the energy for forward motion but is not meaningfully steering the work.

The thing is, different kinds of actions are themselves susceptible to different degrees of agency. We’ve been moving slowly (and with varying degrees of success) away from having all IF commands be of the sort appropriate to physical action.

Hypothesis

If we consider the types of agency involved in different kinds of action in the real world, we will be able to come up with better ways for the player to command non-physical actions.

Physical Actions (Default)

Physical actions such as TAKE BOX, DROP MATCH, OPEN DOOR. The player has a clear, well-delimited idea of what should happen when he makes the action; there is a strong possibility of agency. This translates easily into other game media as well: shooters and many puzzle games allow the player a limited set of physical actions that can be performed within a consistent system.

If one wants to hang a narrative on this, the challenge is to find a way to make a story where physical acts are freighted with emotional meaning. Usually the way to handle this is to create set scenes of crisis (see: the end-game of Portal).

Since it is easier to act physically on bodies than on minds or emotions, these conclusions are often (though of course not universally) violent.

Undirected Exploration

Undirected exploration such as EXAMINE CUP, THINK ABOUT SAMANTHA, and so on. The player usually(*) does not intend any particular specific result, but if the responses are implemented at all the exploration is likely to be successful, in the sense of revealing new information to the player. This is even more true of THINK than of EXAMINE, though THINK is relatively uncommon in IF. (On the other hand, sometimes EXAMINE produces the player character’s insights and commentary about an objects.)

One can safely “drive” this kind of interaction via keyword. Blue Lacuna relies a lot on this kind of interaction, which is one reason that its highlighted-keyword interface is effective in conveying so much of the possible action to the player.

Hypertext can do the same, which is perhaps why so much hypertext fiction is about mental experiences — memories, thoughts, juxtaposed visions — rather than physical ones.

When I think about a topic, I have no intention in advance about what I am going to conclude as a result of that thought. I may have some distant notion that thinking more about topic X may reveal a similarity between X and Y; or (on a more emotional plane) that thinking more about X will irritate me. But this information about the probable result of the thinking is often vague and intuitive rather than specific and verbal.

Hypertext about intellectual journeys sometimes, but not always, provides enough clues to replicate this aspect of thinking. One of my favorite examples is Le Reprobateur, about which I have written some agency-related comments before; it also has the advantage that there is an implicit goal, to find out as much as possible about the story, pursuing its themes in whichever order is most satisfactory. The Space Under the Window also gives some (limited) contextual cues about where it might lead to follow up on particular words, but its success here is variable.

As we dig deeper into the problem of characterizing the player character, I wonder whether we might not want to look again and more closely at this sort of interaction. There’s a lot of IF that has explored transferring mental and emotional states into a physical realm via metaphor. Some of it has been fairly effective and some rather overdone, but in all cases the metaphorical handling an intellectual maneuver that tends to distance the player from the very emotional states that the work is intended to investigate. On the other hand, many players have reported a direct visceral reaction to art games such as Jason Rohrer’s Passage or Gregory Weir’s (I Fell in Love With) The Majesty of Colors. These use extremely simple gameplay and narrative content but offers the player very intuitive controls and feedback.

The goal of exploring more in this direction would not be to convert IF into hypertext or to do away with the actions from a story, but instead to present interludes in the storytelling that would allow for internal monologue; and to allow the player not only to explore but perhaps even to establish the PC’s motivations without, on the other hand, being reduced to answering yes/no questions about them. In a comment on a previous post, Conrad wrote

Similarly, choice-points ought not be points. They ought to be networks of moves, each move clearly working toward the chosen effect… Plus, being able to move around inside the exploded view of a decision node is just more interesting than push-button contigencies.

At the time we were talking about a conversation scene that moved gradually towards one outcome or another, but the same could also happen in interior monologue, as the player character moves gradually towards one emotional state or degree of resolve rather than another.

I have some implementation ideas about how to hint to the player about the outcome of thinking/emotional choices, enough to endow these scenes with agency and give a more intuitive feel to traversing them. But I need to try a few out and see how they go.

Directed exploration

Directed exploration. I said above that the player usually does not intend any particular result of exploration, but sometimes there is in fact very specific information that the player wants to extract. Proponents of heavily simulated worlds have touted as a benefit of such a system the fact that they would then be able to type questions like HOW BIG IS THE BOX? or WHAT COLOR IS THE KERCHIEF? and receive meaningful answers that, in turn, might improve their understanding of the pictured world and perhaps help them to solve puzzles. (See also: David Welbourn’s brand-new 69,105 Keys, where arguably the entire point is to get the player to ask more and more refined questions of the model world.)

One concern about this, in turn, is the fact that specifying all this data, even where it isn’t relevant, may place an undesirable burden on the author. A couple of further options would be to implement such commands with a “That doesn’t matter” response to most “how big/what color” sorts of questions, directing the player only to consider the qualities of objects that have puzzle value; or to rely upon invented specificity, which is probably what a human game-master would do.

Player: “What color is this kerchief?”
GM: (never having previously considered the question) “…Pink.” (It is now.)

In such a scenario, the game would default to most objects having indeterminate color, material, etc., but would select a plausible one (and some work would have to go into crafting the definition of plausibility here) in response to player inquiry; after which, the object would be permanently defined as a pink kerchief and retain that characteristic for the remainder of the game.

I think directed exploration may have some interesting potential, but I am not quite sure how it might improve narrative interactivity as we usually see it in IF. I can fantasize up some directed questions the player might ask about story, like WHY IS SAMANTHA ANGRY? or WHO KILLED STEPHEN? — and it might be really interesting to play a game that was able to introduce new scenes explaining the answers to these questions, though it would be a lot of work to program and would probably require conversation-like prompts in order to make the possibility space clear enough to the player.

Besides, the result is the player acting like a reader, not the player acting like a character — not necessarily a bad thing in interactive narrative, but a different thing than we usually do in IF.

In practice, I think something like this would come out a bit like a conversation game, only the interlocutor would be the parser/narrator rather than a character within the story.

Expressiveness

Expressive actions such as SMILE, FROWN, and CRY. The player wishes to convey or connote something without necessarily having any effect on the game world at all. These are equivalent to the natural facial expressions, gestures, and expletives that we produce thoughtlessly during the course of daily life; they’re naturally low-agency actions, in that we tend not to think about or intend any consequence at all.

(I’m not counting, here, calculating gestures that people make with the intention of manipulating one another; those might go under the heading of Conversation instead.)

Along somewhat related lines, part of Chapter One of Blue Lacuna invites the player to type free-form responses to various events. If it recognizes specific keywords in these responses, it tweaks the replies accordingly; even if it does not, however, the fact that the player has been challenged to consider her reactions and feelings adds impact to the scene.

The questions to the player in The Baron are similar in some respects, in that they invite the player to consider moral questions and suggest answers; I found that a little more cerebral and distancing, though that may not be a bad thing. (Would I have wanted a deeper visceral connection to the protagonist of The Baron? I don’t think so.)

Expressive actions are the kind that can most benefit from completely freeform input, where the parser acknowledges whatever it can but never actually rejects anything (since no expressive reaction is ever “wrong”). Within the context of a mostly parsed game, of course, that’s a bit tricky, since the parser can and should reject physical action commands it doesn’t understand, and it will only recognize expressive commands if it has been explicitly programmed for all of them.

Blue Lacuna gets around this by dropping into a visibly different parsing mode. That works for its purposes, but part of the fun of expressive actions lies in being able to incorporate them freely into the rest of the gameplay (see: The Tale of the Kissing Bandit, Lost Pig).

I have one crazy idea that is probably lame: let the player type an escape character to shift the parsing into “expressive” mode, at which point it would respond to keywords if it could, but otherwise indicate a generic acceptance. Downsides: this feels a little unnatural to do; it is an additional convention that would have to be taught to the player; it requires the author to spend some time thinking about emotive responses the player might want to communicate at each stage of the game and build in keyword responses, which is an extra chunk of work that might or might not enhance the experience.

Possible upside: one of the things I enjoy about beta-testing is making little escape-charactered comments back to the game. I wonder what it would be like if the game were primed to respond to those but in a different register from the regular interaction. On the other hand, I may be totally alone in enjoying this.

Like I said, probably lame.

A handful of IF games even recognize what we might call “direct feedback”. For instance, to quote from some JayIsGames feedback to Suveh Nux:

Haha, when I used my first spell I typed in “awesome” as a command and the game replied “Thanks.” I love that level of responsiveness.

This is obviously fun, but probably belongs in a different category, since it is communication with the game from the persona of the player rather than communication in character.

Conversation

Conversation is itself a salad of all these things. It can express intended results in the emotional or physical model (INSULT QUEEN), (PERSUADE QUEEN TO DONATE MONEY FOR OCEAN VOYAGE); it can be undirected exploration (ASK QUEEN ABOUT THE KING) or directed (ASK QUEEN WHERE THE KING IS).

IF has developed a range of input conventions to allow the player to do all these kinds of things, depending on circumstances; and one of the things I like about hinted conversation (as seen in TADS 3 and in Eric Eve’s I7 conversation extensions) is that the same game can safely incorporate all of these different varieties of conversational agency in the same place.

If the prompts are written sensitively, they allow some characterization via the gap between intention and execution. For instance, imagine a scene in which the prompt is

You could flirt with the mariner or try to make him laugh. (Men like a lady with a sense of humor.)

The MAKE HIM LAUGH option could lead to the PC telling an awkward joke, making an offensive remark, or throwing a pie at his face. Which one happens will give us a lot of information about the player character, even if her action is not what the player intended. Rameses has a few bits where the player’s command and the player character’s execution have surprisingly little to do with one another, but in ways that illuminate the character.

Some problems.

Agency gap. This last point opens up a sticky issue. If the player’s actions always go exactly the way he intends, he has a lot of agency, but he’s probably bored by the narrative direction. If they never go the way he intends, he has little or no agency, and may feel that he is being toyed with by the author.

Input uniformity. Another problem is that mixing all different forms of command structure — free parser-based input, hinted conversational input, moments of multiple-choice decision, and hyperlinked exploration — may lead to an off-puttingly complex experience for the player.

On the other hand, some amount of this may be necessary if we are going to offer the player the different sorts of agency that pertain to non-physical acts.

14 thoughts on “Types of Action and Types of Agency

  1. There are a bunch of games that have THINK ABOUT FOO commands to provide some kind of background about a topic; there are also some that use THINK as a hint dispenser, where the fiction is basically that the PC is solving the puzzle rather than the player doing so.

    I don’t know that there are any where the game actually shifts into a different “thinking” game mode to allow the player to freely explore an idea space, though.

  2. Interesting article.

    Broadly, I agree with what you’ve laid out, although I’d rather use “Expressive” commands label to box up all those ways the player can express his current feelings to the game, rather than simply those commands that allow the PC to express feelings to an NPC (such as Smile, Cry, etc). Having the player express himself, and be understood, is perhaps the bedrock of a satisfying interactive experience. The IF may be a little black-and-white for this sort of thing; compare “WALK CAREFULLY” with moving in gentle steps through the darkness in Thief.

    I’d also comment that the “Agency Gap” is the nub of the matter when it comes to good storytelling in pretty much any medium: protagonists in static fiction have some agency, enough for us to identify with them and think of them as protagonists, but sufficiently little that they can’t just “WIN GAME” and nullify the story before it starts. I’m not convinced that the Agency problem in interactive fiction is much different: it’s just a matter of balancing the progression of events against the exposition and plot-drivers.

    • I’d rather use “Expressive” commands label to box up all those ways the player can express his current feelings to the game, rather than simply those commands that allow the PC to express feelings to an NPC.

      Yeah, possibly my examples were badly-chosen, but I would certainly include other ways for the PC to express feelings, even if not directed at an NPC. (Expressing the feelings of the player out of character is a tricky other issue, and I’m not sure I put that in the same box.)

  3. Interesting as always, Emily.

    As you undoubtedly know, the classical view of causation considers four types of cause: the matter of the phenomenon investigated; its form, which is to say its nature and the rules governing it; the conditions and forces which have effected this particular instant of the phenomenon; and its purpose. Our modern scientific-type idea of cause mostly means the efficient cause.

    In a well-written IF work, player agency can address the efficient cause: that happens when the plot responds well to the simulator. The final cause is not usually considered, although it does come into play in the experience of the game: years ago when I was DMing a lot, a useful question was: “Now when you do that, what are you aiming to accomplish?”

    I can’t imagine a way of handling that question programmatically, but I imagine game designers keep in the back of their minds the purpose of shaping player intention, and making certain of the options more obviously compatible with certain intentions.

    The formal cause — the phenomenon’s nature and the rules that govern it — is defined by the game; defining the formal cause of the different situations in the game is designing and programming the work.

    The material cause — what the phenomenon is made of — has to be considered in disparate ways. Insofar as we are interested in the simulator, we must consider the simulated events to be caused by appropriate simulation physics. If we are interested in the emotion of the characters, we must consider the stuff of emotion, how text and text games can be shaped to look like emotion, how it can interact with the player’s emotionality, and so on.

    And that’s more difficult, because the rules of physics are better understood.

    I think where this shows up is in the influence the efficient cause has on the interface: the command-line prompt, which works so well for the physical simulator, no longer works well for conversation, because the stuff of conversation is not like the stuff of physical objects; nor is our ability to simulate conversation like our ability to simulate physical reality; and as a consequence the form of input must change.

    So, with regard to your intriguing hypothesis:

    If we consider the types of agency involved in different kinds of action in the real world, we will be able to come up with better ways for the player to command non-physical actions.

    I’d counter that, when we truly understand the non-physical material we want the players to be able to control, which means that we understand the rules which govern it, then we will be able to design, first, a minimally-intrusive interface which enables the player to grapple with, second, a set of contigencies which reflect and appeal to our intuitions regarding these phenomena — be they conversation, emotion, or whatever.

    Conrad.

  4. I think that expressive commands are too often ignored. One of the most memorable things in Eric Eve’s “Nightfall,” for me, was the depth of the scripted responses to emotive commands in the final confrontation. If you knew that you held all the cards, and decided to smugly “> SMILE,” you’d get an appropriately worried response from the NPC.

  5. Thanks for writing this. I’ve been grumbling about agency in NPC interactions recently, and not come to any so-well-formed conclusions. In playing through The Witcher, there are lots of places in high-stakes conversations where you don’t know what effect your speech is going to have; maybe it’s an artifact of bad translation, but it’s easy to think you’re steering the conversation one way and have it go the opposite way. There are lots of characters who want specific gifts, and give you no clue about what kind of bribe they’re looking for, or a clue so vague you have to give them a dozen things to hit on what they were after. There are characters who have certain conversation options open only during very brief windows of time; the action that closes them down seems mostly unrelated. There are characters who want one particular obscure thing from you right now, and if you don’t have it in your inventory at that time (or think you do, but were mistaken), that conversation branch is lost forever. The result, for me, is a huge impetus towards what roguelike games call “save-scumming”: if I have a conversation where I have a desired outcome, and fail at it because what I did seemed insufficiently clued as wrong, just reload and try again.

    This is complicated by the fact that some interaction mechanisms seem to be undocumented (drinking contents?). It also seems to have lead to more rampant spoiling on the web than in most other recent mass-market RPGs – you can find “shopping lists” by chapter of all the items you should gather, who you should give them to, or what future chapter you should be saving them for.

  6. Agency gap. This last point opens up a sticky issue. If the player’s actions always go exactly the way he intends, he has a lot of agency, but he’s probably bored by the narrative direction. If they never go the way he intends, he has little or no agency, and may feel that he is being toyed with by the author.

    Yeah, this is true of a lot of games and skills-y play. It seems we enjoy most those reponses from that narrow in-between zone of “I never thought of that!” It mediates between surprise and predictibility: this is where we have things that are foreseeable, and yet unforeseen.

    I think we respond this way when our actions have all the intended effects, but they also have further effects that we didn’t consider; or when they have effects very like our intentions, but the cumulative result is not what we were aiming for; or when the fulfillment of our intention itself defeats our goal — a Phyrric victory, in other words.

    This is probably the trickiest part of straight storytelling, but also the most rewarding. Stories that do this kind of thing really well always become my favorites. (_Run Lola Run_, for example, although that does a number of other things right too.)

    Mediating between frustration and satisfaction is general to a lot of art. It seems to me Umberto Eco talks about this in _the Art of the Striptease_, but it’s been a while since I read it, and I only vaguely recall the argument. I believe it amounts to Eco calling certain categories of art “pornographic,” based on where he puts them on the arousing-interest-with-no-possibility-of-followthrough scale (or only with self-serving followthrough?). Basically it seemed that Eco got too involved in a runaway metaphor, like a film noir hero with a treacherous blonde.

    But the basic idea I think applies: in straight fiction, we need to want something for the characters which is not provided — and yet it must be somewhere within sight. We tolerate neither stories that relentlessly torment sympathetic characters nor stories that endlessly shows nice things happening to good people. It’s a tease.

    Most games and logic puzzles, while they hold our interest, occupy this space where we have a goal which is being frustrated by a series of shifting obstacles. Solving a Rubik’s Cube relies on understanding the kind of “agency” you bring to the puzzle — as it becomes more complete, you need to understand better the side-effects of your moves, and take moves to counter-act them.

    Attainment of the goal is a tasty treat being held forth and yet held back; it gets closer as we do better; it is a tease, and yet if we have agency, it is a fair tease. (I’m sure there is an Umberto Eco essay in there somewhere.)

    Input uniformity. Another problem is that mixing all different forms of command structure — free parser-based input, hinted conversational input, moments of multiple-choice decision, and hyperlinked exploration — may lead to an off-puttingly complex experience for the player.

    That’s just a practical problem to be addressed with the game design. Complex interfaces are built up of simple elements, and the simple elements can be taught early on in mini-game, tutorial-type situations. Expository programming.

    On the other hand, some amount of this may be necessary if we are going to offer the player the different sorts of agency that pertain to non-physical acts.

    As I said before, I think the form of the interface must follow the logic of the simulator; and the logic of the simulator reflects your understanding of the non-physical action, contestation, and agency you’re writing about. So start with an interesting logic for the non-physical domain of inquiry, with a couple variables the user can control, a couple they can only monitor, and a clear cause-and-effect chain, and then work out a minimally-intrusive interface for your logic.

    I don’t think you can do it starting with the interface.

    Conrad.

    • The interface and the underlying simulation are (or ought to be) closely related in any given design, though; and it does no good to come up with a simulation concept if there is no reasonable way to drive it.

      Which is basically to say that I think there are multiple ways to brainstorm your way into a good system, even though sooner or later you are going to have to think about all of these aspects (how does the player exert control, and over what?) and about the whole they create.

      • On the one hand, I favor any brainstorming techniques that are effective. I even quite like techniques that are ineffective. So, sure: we can any of us tackle a problem any which way.

        On the other, form follows function. A tool must fit the hand that uses it, but how it fits that hand is determined by the purpose of the tool: a screwdriver, an axe, a keyboard, a shoelace, chopsticks — the gripping end of these things all conform to the human hand, but how they conform is dictated by the purpose and by the business end.

        The hand is maybe not infinitely adaptable, but it is highly adaptable; and the human mind, language, and computer logic are probably more adaptable than the hand. So I figure it has to start with the simulator. You don’t invent an axe by studying the hand.

        (Now watch Emily prove me wrong by inventing some freakishly wonderful general-purpose interface.)

        (I’d be fine with that.)

        Conrad.

  7. I winced just a little when I saw conversation was a group tacked onto the end. I like the idea of this breakdown, but I think you’ve muddled the means (saying, doing) with the ends (respectively: effecting, absorbing, seeking, expressing). I think this happened because you wrote down physical actions first because it springs to mind first, then finally added conversation later to contrast with it.

    Physical actions have a high degree of agency because they have an immediate and clearly defined effect on the world. The player definitely has control. But I believe there are a class of conversational utterances which also have this immediate & definite effects property: “I now pronounce you man & wife”, “I find you guilty as charged”, and any Big Reveal like “*I* was the one who killed the bastard!” The effects from these kinds of utterances (typically) have an immediate and definite effect on those around the speaker than most conversation. That’s why such utterances are so often used at a story’s more climactic points. I believe these utterances and the physical actions should be grouped together because they have about the same potential for agency, and the conversation group should be struck from your breakdown because of ends vs means.

    (As to what to call that category… well, I don’t have a pat answer. That speech act guy Searle had about five broad categories of speech acts, and that one was called the declarative: “speech acts that change the reality in accord with the proposition of the declaration”. That sounds like physical I-F actions to me. (Another category was expressives, BTW.))

    Let’s say, for now, that agency is the player’s ability to affect the world and story, and it depends, in turn, on whether the player can form a reasonable guess about the results of an action before taking that action. If the player cannot guess [...] where the action will lead, there is no agency; the player [...] is not meaningfully steering the work.

    I have a brain twister for you. Given an oppressive, clockwork, micromanaging government (think _1984_), a PC citizen like V from _V for Vendetta_ would have a lot of agency regarding his plans because he’d know the government’s response plus the effects of his rabble-rousing acts. But since his goal is the defeat of that clockwork government, the more chaos he sows the less sure he can be of the government’s ability to respond in the precise and harsh methods it typifies. (Citizens become increasingly less helpful to the government, lower-level government officials increasingly let things slip, etc. ) Near the endgame, as anarchy draws close, the PC is losing his agency because he can’t count on any of the clockwork still functioning correctly if at all. But, the very fact that he can’t reliably choose anything means he is succeeding, so he experiences agency rather powerfully.

    So does the endgame have agency, or not?

    • I think this happened because you wrote down physical actions first because it springs to mind first, then finally added conversation later to contrast with it.

      I think we disagree less than you think: as I said, conversation tends to include all those kinds of agency.

      So does the endgame have agency, or not?

      I guess I would say that the endgame allows the player to experience the results of earlier agency, but itself doesn’t offer much. Which seems entirely consistent to me.

  8. Pingback: Weekly Reader | William Patrick Wend

  9. Pingback: This Week On Twitter | William Patrick Wend

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s