Self as Fractal

In defense of the Actor Stance

Recently, Victor Gijsbers (The Baron, etc.) posted "Actor stance, make way!", an essay criticizing his perceived ubiquity of the "actor stance" in interactive fiction. Some context: Figaro, as the title implies, is a demo work based on The Marriage of Figaro. The main feature of Figaro is its approach to decision-making. You play Figaro, hiding in a trunk and spying on Susanna and a mysterious man. Who is this man? That's for the player to decide - literally; the game outright asks the player to choose one of three options. Once he (or she!) has been identified, Figaro can - well, he can do one of three things, depending on whether he's got an axe, a camera, or a cream pie. Again, the player chooses.

This setup is an example of the "author stance." The terms "Actor stance" and "Author stance" (and "Director stance," which isn't really discussed much in Gijsbers' post) are defined in detail at this link:

In Actor stance, a person determines a character's decisions and actions using only knowledge and perceptions that the character would have.

In Author stance, a person determines a character's decisions and actions based on the real person's priorities, then retroactively "motivates" the character to perform them. (Without that second, retroactive step, this is fairly called Pawn stance.)

Gijsbers argues, in essence, that part of progress in interactive fiction is moving away from the Actor stance, where he claims it is currently, to the Author stance. I think progress can be made equally well moving toward it.

I admit that I do not have experience with RPGs, so I will not discuss that analogy in much depth. Instead, I see IF as more akin to static fiction. If we're invoking the old story/game dichotomy, I lean heavily toward story and believe it better suits the Actor stance. There's plenty of support for this in the actor/author/director model. Also from the link:

Consider the previous example of a group who has arrived at the agreement to role-play a vampire-character game...what sort of Stances might be most common during play, from each of them?

...

Another is interested in experiencing and Exploring [sic] the nuances of the story as it is presented from an external source (perhaps a sourcebook and/or a GM), and spends a lot of time in Actor Stance.

There are no sourcebooks or GMs in interactive fiction, of course, but this approach still works. The player, in his or her interactions with a work of IF, draws out various nuances of the story by making the PC do things. In the process, he or she becomes immersed in the work. Immersion, too, is discussed:

Immersion is another difficult issue that often arises in Stance discussions. Like "realism" and "completeness" and several other terms, it has many different definitions in role-playing culture. The most substantive definition that I have seen is that immersion is the sense of being "possessed" by the character. This phenomenon is not a stance, but a feeling. What kind of role-playing goes with that feeling? The feeling is associated with decision-making that is incompatible with Director or Author stance. Therefore, I suggest that immersion (an internal sensation) is at least highly associated with Actor Stance.

So this is the great advantage of Actor stance: facilitating immersion. Before I go any further, I'd like to clarify a few related terms. In his essay "A Triangle of Identities," part of the DM4, Graham Nelson identifies three main identities at work in interactive fiction: the player, the protagonist (or PC), and the narrator. The player is the actual person reading the work of IF, the protagonist is the character depicted in the story, and the narrator is the voice relaying that protagonist's thoughts. The protagonist and the narrator are functionally identical to their counterparts in static fiction; the player, however, has a greatly expanded role. I'd like to add a fourth identity to this: the parser. Not the parser as in the actual source code that processes commands, but the parser as in the voice that tells you when you've entered something invalid.

What I propose is a way of writing and reading IF that's rooted in the Actor model, where the player temporarily inhabits the mind of the PC. One common representation of the Actor model is the player asking him/herself "What would (PC) do?" but this isn't quite accurate. It still interposes a barrier between the actor and the PC. People don't go around, after all, asking themselves what they'd do. They come up with ideas, and then do things. I'm sitting down right now, but I could at any moment get the urge to jump, or go into a jig, or break a window or run out of the building screaming. In the Actor model, the player provides these urges. The protagonist either does them and reports back, or rejects them, providing his or her reasoning.

To return to Graham Nelson's roles, this model conflates the protagonist and the narrator. In static fiction, this is roughly analogous to the first person point of view. The narrator *is* the PC - maybe not the "main character," but the one telling it. There's one big difference here, of course: the text (usually) isn't first person, but second. This facilitates immersion - the player, being addressed, is brought into the work. Any printed text, then, is communication from the PC to the player.

Some existing works of IF are particularly conducive to this model. Stephen Bond's Rameses, for one, is particularly illustrative. (Spoilers ahead, for those unfamiliar.)

The player, in Rameses, is a voice in Alex's head, who suggests various actions he could take or words he could say. Alex, however, no matter how much he might want to comply, usually doesn't. He's too timid, or cowardly, or self-absorbed. The story unfolds unnervingly, as the player desperately tries to do something - anything! - to break Alex out of his shell. Some people have criticized Rameses as being uninteractive, or too much like static fiction. These people, in my opinion, are completely missing the point. This could be done in static fiction, technically, but it wouldn't be nearly as effective. It's one thing to write about how Alex is trapped inside his head, and quite another to have the player re-enact it.

Gijsbers seems to argue that it's dangerous to conflate the player and a fictional character in this way, bringing up Adam Cadre's Varicella. I disagree with this. There are plenty of completely depraved protagonists out there in film, fiction, theatre, and other media, and players identify with them all the time without becoming depraved themselves. If I read Hamlet and decide to identify with Hamlet, to closely follow what he does, that doesn't mean I'm going to put down the book and go murder the whole town. Ditto Macbeth (Shakespeare does this a lot). If I watch Sweeney Todd and start to root for Sweeney, I'm not going to get up after curtain and start my own barber shop. I'm not alone. As Gijsbers writes, most people can dissociate themselves from a fictional character. Immersion is what happens when they decide not to for a while. It may be disturbing to identify with an unsavory protagonist, and good works will embrace that.

To actually achieve immersion requires a good deal of implementation. Underimplementation kills, as reviewers point out all the time. Not only does the work have to be well-implemented, but the PC's voice must be strong. Anything that the work prints out has to be a plausible thought of the PC. If an action is trapped, there has to be believable reasoning behind it. (On a side note, this is one of the things I really like about Inform 7. Compared to I6, text substitutions are a whole lot less tedious, which makes it a lot easier to vary the text. After all, people don't repeat the same thoughts.)

It becomes even more important to make puzzles believable. It's bad enough to have soup cans without having to come up with a reason why the PC should care about them. Knowledge puzzles are the real threat. The player, after all, has thoughts of his or her own, and these thoughts might just be channeled into figuring something out before the PC does. This is why knowledge puzzles are so difficult.

A good deal of development, in the actor model, is devoted to getting rid of the parser voice. In an ideal playthrough, this voice would never appear; however, people make typos and use unsupported vocabulary and do all kinds of things the author didn't think of. So the parser intervenes with predefined messages. Most of them are fine responses - Graham Nelson and Mike Roberts are no slouches - but the PC most likely isn't either of them. The voice intervenes, then, and it's either someone completely unrelated, or a conduit for error messages. This is not good for immersion, and is a factor in many common IF problems. Guess-the-verb, for example, is frustrating in part because the parser has totally taken over. The player wants to get back to the PC, but can't. And again, underimplementation kills, because there are all these default parser messages where there should be PC communication. Fortunately, both Inform and TADS make it easy to change them.

A concern that can arise out of the actor model: If the work of IF represents a player's thoughts, why does the PC's mind work like a command prompt? There are two ways to approach this. The first is more of a joke that can only be done once or twice - maybe the PC literally DOES think like a parser! For most works, though, it's something the player has to accept, like the fact that the PC moves in compass directions. This requires suspension of disbelief, yes, but it really isn't all that much worse, on paper, than wondering why a novel's protagonist never mentions all the times she uses the restroom or has a snack, or why she inserts chapter breaks after suspenseful things happen, or why everybody she knows is portrayed by a celebrity, or why she breaks into song (if she's in a musical). It's less familiar, is all.

Another issue, as Gijsbers correctly points out, is out-of-character actions. As examples of these, he mentions saving, checking the score, and undoing. I'd extend that to meta-commands in general - and there are more of them than you might think. "Look," for example, might as well be a meta-command for "print a room description," since that's almost always how it's used. These too can jar the player out of immersion, but messing with them is a bit more problematic. You're not likely to make many friends, for instance, by disabling UNDO. There are a number of ways to get around this: disable them, make them actual commands somehow, preserve them as meta-commands but have the PC report them, or ignore them altogether, among others.

---

The above is the standard actor model, but it can be twisted a bit. An example can be seen as recently as Violet by Jeremy Freese, the winner of 2008's IFcomp. Violet generally follows this model, with one major change: the narrative voice isn't the PC's own voice. It's the PC imagining what his girlfriend would say. Everything is written in her voice, down to the meta-commands, and it works wonderfully. Also from 2008, the third place winner, Jim Munroe's Everybody Dies, uses several distinct personas in one head.

These are just two examples. Authors could come up with plenty of other variations on the model. You don't have to change it, though. Immersion is good on its own, and I'm excited to think about all the things that can be done by following this model.

---

Returning to Figaro for a second, I see no reason why this could not be done equally effectively with the actor stance. (Not to say that Figaro is bad. It isn't, at all. I quite liked it. This is just an exercise.) This requires a few assumptions:

1. People jump to conclusions based on faulty data. A lot. All the time. (Especially in theatre.)
2. People cannot always gather sufficient data.
3. People especially can't gather sufficient data when they're hiding in a chest.

So the intro would get into Figaro's voice, somewhat like what it does now. After a while, a mysterious man shows up. Assuming we're in the actor stance, it makes total sense for a person to examine the man. It'd take unconventional rationalization for someone not to. Even if that wasn't the case, the author designating the man as "mysterious" is pretty much commanding the player to examine him.

Let's be clear about what that means, though. In the author stance, "examine man" might as well be a meta-game command, roughly equivalent to "Give me some information about this man, parser." In the actor stance, "examine" means "examine." Here, it means that the PC is literally taking a close look at the man:

"Through the keyhole, you can't see much - just Susanna and some mysterious figure. A man?

>examine man

You strain through the keyhole, but it's so hard to focus on everything at once."

This is where the break happens. In Gijsbers' example, the player decides outright who this man is going to be, catalog-like. Keeping Figaro in the Actor stance believably would require the author to provide some way for the character to figure out who the man is. Let's say that the Count has a big ol' scar on his right cheek, and that Cherubino's got a paunchy belly, like a cute little adulterous Cupid. It sounds contrived, but people rely on signifiers like this all the time. How many times, for example, have you totally failed to recognize somebody because they got a haircut?

This information, of course, would go into the intro. In character, of course - you wouldn't just say "Oh, and by the way, the Count has a scar on his cheek and Cherubino has a paunchy belly." For the record. You'd instead have Figaro fuming about the miserably Count with his ugly scarred face - and that potbellied Cherubino is no better, the bastard! The key thing here is that this is a magician's choice. In other words, if you examine said body part, it's going to be scarred or bellied. (The game could prompt this by providing increasingly lurid but misinterpretable details of the two carrying on.) This version of Figaro, whether he likes it or not, has a vested interest in identifying an adulterer. Everything he sees is evidence to him. A man with a paunch? It's got to be Cherubino! The ONLY possible explanation. Example, again:

">x torso

You peer through the keyhole again, squatting a bit until the man's torso comes into view. It's paunchy. Too paunchy. Cherubino. The wretch!"

This, of course, would change the man's name from "mysterious man" to "Cherubino," and alter all the associated descriptions. Now for that inventory issue. This is a bit more difficult. You'd have to slip the three options - kill, photograph, or be-pie - into the intro, come up for plausible reasons why the PC can't do other things, and account for as many synonyms as possible. Then, whatever the player chooses determines what the character's carrying. If Figaro's going to kill them, he finds he's got a knife hidden in his clothes! (It'd be hard to hide an axe.) You'd have to deal with >inventory by turning it into a non-meta command. That is, changing it from "print my inventory" to "fumble around in my pockets to see what I've got" - only it fails. So, an example:

">i

You fumble around in your pockets, but can't find anything on a quick once-over. Maybe if you tried to find something in particular...

>kill cherubino

"YOU'RE GONNA PAY FOR THIS, WRETCH!" you roar, instinctively grabbing for that one fold of your robe, pulling out your knife, and doing the deed."

There's one big difference between this example and Gijsbers' game, of course. In Gijsbers' Figaro, your decision is pretty much the word of God. If you say the mysterious man is the Count, he is the Count. In here, your decision simply represents who the character thinks the man is. You can look at his torso, see his belly, and come to the conclusion that it's Cherubino - but maybe it isn't. It's just a paunch! Maybe the Duke from Rigoletto stopped by or something. Or Santa Claus. The possibilities are endless, but they don't really matter - for the story's purposes, what matters is the impression the character gets. Is it prone to lawnmowering? Yes, but no more than the original.

I'm not trying to say, either, that this version is any more valid. That'd be incredibly rude. It's just an exercise, again.

---

Is the Actor model the only one through which to approach interactive fiction? Of course not. This does not mean, however, that it should be discarded, or that it's somehow not conducive to innovation. Many, many fine games have incorporated aspects of this model, and I think there's plenty of room for progress in going further. There's room for more than one kind of IF, after all. If the actor model needs to move aside, it's only to let others stand by it.