Mark Bernstein used my series on frustration in interactive media as the occasion for his comment, so I’m trying to catch this wave of interest passing through Grand Text Auto.

Mark Bernstein was generous about my post, and I appreciate his attention, however I think he may have misunderstood me in the process of making his point that IF critics ignore his argument on the impossibility of interactive tragedy. I did not clearly label my post as the fourth in a series, so perhaps he missed the other three posts. Thus he paraphrases my argument as:

the underlying puzzle nature of IF is an intrinsic limitation that can only be partly assuaged by careful hinting and skillful writing

… when in fact, I’m more interested in amplifying than assuaging that puzzle nature.

Limits vs. Limits

When Bernstein talks about limitations as something to be “assuaged,” I assume he’s thinking of ‘limits’ as ‘inadequacies,’ whereas I’m thinking of ‘definitions.’ The formal limits of a medium are constitutive, like the limits of the canvas to painting, or the camera to film. As a sonnet structure is both the scourge and the support of a poet, the ‘limitations’ of a given medium are the beloved antagonists of its artists.

This was why I began my series trying to imagine an aesthetics of frustration:

I criticized the piece for being frustrating, then took a step back - what about interactive art whose point is to frustrate, or ‘disruptive’ art which tries to frustrate well?

What would it mean for a piece or a medium to be “better at frustration?”


While it is true that I didn’t cite Bernstein on Hamlet, I use tragedy to quite the opposite purpose. Dramatic irony is a perfect example of an artistic effect that depends in part on the incapacity of the audience to intervene - however much they might desire intervention. Why do we assume that a medium that includes interactivity makes frustration less artistically important or valid? It seems more likely that the opposite is true. Giving the reader agency (things they can do) inevitably highlights incapacity (things they can’t do) - and incapacity has a long history of aesthetic effect. IF is probably *better* at tragedy for its ever-present, frustrating limitation on free agency, and thus it doesn’t surprise me that recent IF is full of tragic devices that depend on irony. Just a few examples:

  • Babel hinges on what the interactor/character cannot know about himself
  • Rematch hinges on the fate which both interactor and character know, but can’t communicate to others
  • 9:05 rereads to a completely different conclusion once the interactor has learned the secret of the character

Some think interactivity and dramatic irony don’t mix because the audience understands what Hamlet cannot understand. The assumption is that the knowledgeable interactor will either actively destroy Hamlet’s story or else be unable to participate in it at all. The list of texts above are some examples of how this can be reconciled in Hamlet-like structures, but I also feel confident that Hamlet himself is not beyond IF’s reach - although it may be an experience beyond the reach of a particular kind of participant.

Sane and Reasonable

That participant would be Bernstein’s “sane and reasonable” interactor who, somewhat like the infinitely rational subject in economics, calculates the straight line between two points and then follows it. Peter Brooks says that, like an organism motivated by Freud’s death drive, a narrative is driven to end, and that the entire life of a narrative is a series of elaborations and deferrals of that end. Puzzles and games can also be understood as occurring only while such a process of deferral occurs. If we assume that the normative goal of all empowered interactors is to short circuit this process, then the “sane and reasonable” interactor is in itself a model for bad narrative or no narrative at all. Because all interactors are assumed to behave in this way, and such interactors will read tragedy badly, interactive readings of tragedy are bad. Quod erat demonstrandum. On the other hand, if we accommodate the rational goal, we get something like this:

You are Hamlet, Prince of Denmark. Tonight, on a rampart, the ghost of your father has told you that he was murdered by your uncle Claudius (now king). The ghost demands revenge.


You eventually do, but die in the process, as do many of your nearest and dearest.


… well, that was… quick. The parser didn’t give us any guff about not being able to see him, or “with what?” yadda yadda… but the experience still left something to be desired. And what if instead the ’sane’ response is a curveball, some kind of limits-testing action (see Mark Marino’s Perverting the Bot) which is out-of-paradigm for the piece like IMMOLATE HORATIO or INVENT GOTH POETRY?

You are Hamlet, Prince of Denmark. Tonight, on a rampart, the ghost of your father has told you that he was murdered by your uncle Claudius (now king). The ghost demands revenge.


You try to contemplate this course of action, but like iron to a loadstone your mind returns to your damned father and your thrice-damned uncle.


Your thoughts jumble, and you feel a wave of fatigue. What if you are confused, or your vision was in error? Perhaps you should seek the council of friends.


… well, that didn’t work either.

If we assume that interactors can never be expected to act writerly, then yes, they will seldom succeed in tragedy. If in place of character-based or situation-based performances they choose sanity or libertarian free agency, then yes: the sane interactor will never play Hamlet. Hamlet just isn’t an infinitely rational kind of guy, and his most likely actions are moping, scheming, and monologing. These can’t even be cued to a reader who is violently opposed to being directed, and desires in place of negotiated interactivity a kind of pure activity.

Such unlimited activity can be had more readily from a blank MSWord document.

Role Playing

However, these problems for the sane interactor don’t trouble another kind of participant, the role-playing interactor - a reader who wants to improvise within the scope of acting out a role, who receives artistic direction from the text itself on how to proceed, and adopts an attitude that Jane McGonigal refers to as “suspension of belief.” For this kind of reader, Bernstein’s quote would probably ring false:

The imaginative reader is bound to think of things the creator never envisioned, and the reader’s best thinking inevitably generates the dullest response: “I don’t understand.”

This is like saying that the very “best” answers to crossword puzzles are inevitably the ones that don’t fit in the grid. A different perspective is that the “best” thinking instead involves the achievement of understanding (Nick Montfort’s Twisty Little Passages), not some untrammeled creativity, and the pain of frustration is not simply that one is not understood by the parser, but that one must also struggle to understand the situation in order to proceed - making moments of failure at least as important as moments of success. Rather than trying to impose agency, one attempts to perform a shared understanding. Not action, interaction.

I don’t mean to airbrush the deeply serious challenges that IF faces in making advances into leveraging natural language processing, weak AI, and so forth. However trying to design interactive storytelling for such sane readers doesn’t seem to be a productive way to expend energy - not when we could be writing for people who want to participate in the slow unraveling of Hamlet’s family, fortunes, and mind… and who are willing to get a bit frustrated along the way, just as we are when we watch the play.

We often begin conversations with the shared assumption that frustrated agency is something which must be minimized in order to create good interactive art / fiction. Fewer stymied moments, and more “it shall be done!” Yet this assumption seems unwarranted in IF, as it is in horror films and tragic drama, in crossword puzzles and in MMOGs. “Frustration is inherent in this medium” is NOT equivalent to saying “this medium is flawed.”

If frustration is the flaw of IF, it is truly tragic, for it is a flaw which simultaneously brings out greatness. The art is the error message, and the error message is the art.

17 Responses to “Frustration, Irony, and Sanity”

  1. 1 Dirk Scheuring

    Good post, Jeremy. I think the advocates of the “sane and rational” idea overlook a fundamental fact regarding the nature of tragic stories: classic tragedy draws its power of impact on the audience from the very fact that there is no “sane and rational” solution to the story’s problem. Tragedy works by putting the spectators, via their identification with the characters, into what I think psychologists call “double-bind situations”.

    Look at Hamlet: the old king was murdered, and the new king has taken over his reign, riches, and even his wife. The old kings’s son knows that the new king killed his father and now fucks his mother, and he’s pissed, plus his father’s ghost eggs him on: “Go kill the mofu!” Everybody else, even his mom, wants to bury the memories, but the active spectator - in the sense that she identifies with the Hamlet character - understands that Hamlet can’t be like everybody else: it was his dad, after all! And that bastard is fucking his mom! Shakespeare’s target audience - a very motley crowd - certainly did see that, from his point of view, Hamlet’s thinking is “sane and rational”, and everybody else appears insane.

    On the other hand, with not everybody being a naturally born tragic hero, a good part of the spectators that are putting themselves in Hamlet’s shoes would probably opt for the “fuck and forget” strategy themselves if they were to chose, just like Bernstein does. So Shakespeare shows that Hamlet isn’t a naturally born tragic hero either, and agonizes over the decision of whether to be or not to be. He trys, in all kinds of persuasive ways, to expose Claudius as the killer, so that the other characters take the weight off his shoulders by establishing communal judgement, but lack of luck makes him fail. Even worse, his efforts end up increasing his problem, because they make him look ever more insane to the people he tries to persuade. We see what motivates the character to change while he’s changing, and we see how him reaching the point where he decides that not to be is the answer becomes more and more inevitable. He’s right! But he’s wrong! He’s right in being wrong, and he’s wrong in being right. And that’s called tragedy.

  2. 2 Dirk Scheuring

    Addendum: Since I come from a country and descend from a population that has allowed and supported killers to govern it and build up their power structures over years, I find Bernstein’s suggestion that enjoying oneself and waiting for an “opportunity” particularly flawed. In such a case, I believe that with each day passing, such an “opportunity” becomes less likely to appear. Claudiusses make inhumanity profitable for large-scale populations, and therefore, they must be stopped. We don’t need less Hamlets, we need them to be more successful.

    For somebody who’d like to try and make an interactive version of the play that is better than the one that exists, it might be worth thinking about whether the player’s range of possible interactions with the play could be designed so that the player could make Hamlet succeed in making the other characters see Claudius with his eyes without killing anybody.

  3. 3 Mark Bernstein

    A very intriguing and thoughtful discussion. Thanks.

    My point is not that a hyper-rational reader could make the drama collapse, but that dram depends on a specific confluence of character and circumstance. It’s not that a clever player could avoid Hamlet’s fate: almost *ANYONE* in that situation could and would avoid it. Hamlet cannot. Understanding why this is true is the subject of the play.

    Same with Oedipos. A lesser man would let sleeping dogs lie. A less responsible ruler would say, “no man can control Gods or diseases.” A less mystical fellow would say, “people die every day; I’m sorry about the plague but this is just the way things are.” A more callous man would say, “Damn! I really messed up. Sorry. Have the senate pass a resolution. I’ll put up a statue or something. How about a nice Burgundy with dinner?”

    When I wrote of a “sane and sensible reader”, I didn’t mean rational — merely a protagonist who doesn’t possess the specific combination of traits that propel this particular drama (whatever it is) to its illuminating but necessary resolution.

  4. 4 Jeremy Douglass

    Thanks, and interesting observation, Dirk - especially on Hamlet’s borderline personality. If choosing what kind of man he is going to be is part of the first-person drama of Hamlet, then we need to provide real choices (direct vs. indirect, personal vs. communal) that will all still culminate in tragedy - which may mean that, as Mark Bernstein said, we need to rig the game.

    I actually don’t have any problem with rigging the game, any more than I have a problem with rigging a puzzle - to the extent that it is rigged, it has a solution or solutions. I think that many of us disagree less on implementation than we do on what constitutes a problem, and this sometimes makes us talk at cross-purposes.

    However, I can see why making series choices as Hamlet and always getting the same results would be frustrating. How could we create a drama of participation and choice that holds his bloody fate invariant?

    Supposing we let Hamlet ACT CRAZY, and this enabled him to advance the plot - if he doesn’t at key moments when he is being observed, then bad things happen because other characters have their guard up - for example, Rosencrantz & Guildenstern never allow him to steal the letter.

    Our imagined version of Hamlet, Varicella-like, involves intrigue, however it is of the learn-through-dying kind - we emphasize that Hamlet is in real danger of being killed by his uncle, and even hint that a return to school would be walking into a trap. Hamlet will arrive at the sword-fight eventually, with small differences, but to the extent that he chooses to ACT CRAZY at key moments, the ellipses in the story get opened up, allowing more exploration and discovery. In the end, how Hamlet feels about his fate is what matters, and that will depend on what he has learned.

    The complicating factor in this design as that the more you seize opportunities to ACT CRAZY, the more observation becomes available to you, but the less possible it is to learn things through conversation (people won’t talk to you or humor you without engaging). Furthermore, the more Hamlet chooses to ACT CRAZY, the more the parser begins to implant artifacts of madness and paranoia into descriptions of his direct experience, showing things that the player know may not be literally happening.

    In the end, the game is “rigged” in that it always culminates with much the same scene - a bloody pile by the foot of the throne - however a certain reader might play again and again in order to experience the central argument of the design - that Hamlet struggled in his choice between direct and indirect, sanity and madness, knowledge and ignorance - and that this determined not whether he would die, but who he was when he died.

  5. 5 Jeremy Douglass

    Mark - thanks so much for clarifying. My substituting the word “rational” for “sensible” after making that economics comparison was actually my editing error rather than an attempt to overstate your argument or build a straw man. I’ll fix it the misquotation and the header in the post above.

    I suspect we are in agreement about many things, and talking about ’sanity’ or ’sensibility’ distracts us from the point that the perspective of an interactor isn’t better or worse than the tragic protagonist - just different.

    We agree (I think) that there are going to be interactors who are acting more or less in sympathy with the protagonist - not necessarily a literal sympathy with those goals, but at least an understanding of that perspective and a willingness to perform or role-play towards those goals.

    We agree (I think) that interactors who are totally out of sympathy with the tragic perspective and who had free agency - instead of sane or infinitely rational, we could just call them “unsympathetic” - these interactors would simply avoid the tragedy.

    We agree (I think), that it would be nice if this gap between the perspective of the reader and the character we want them to experience could be narrowed. It might be narrowed through discipline-and-punish style feedback that teaches the interactor to be Hamlet, or else. It might be narrowed through incentives, plot points and experiences that are only unlocked through good role-playing. It might be narrowed in a way completely external to design, through a reading community of interactors who desire to understand Hamlet rather than defeat his problems.

    Probably it will be narrowed through all of these, and more! However what seems less likely is that attempts to directly accommodate interaction completely out of sympathy with the design (”LEAVE, GO TO SCHOOL AND GET LAID”) will ever bear much artistic fruit. My mental test is to imagine the most talented improv storytellers I know standing in for the machine, and imagine them dealing with this kind of behavior while still trying to arrive at a proper tragedy. Weeks of preparation, and now you have to age all your characters 5 years, re-imagine the political landscape after five years of rule under Claudius, and explain how the supernatural structure (ghost, purgatory, etc.) deals with Hamlet’s sojourn. Even in the most skillful table-top role-playing circle (which relies not on ‘AI’ but on ‘I’), the results of dumping 95% of your structure 5 minutes after you start are seldom pretty. For computers, “rigging” seems to be the only solution - oh, but what a rigging it will be!

    At this point, if you’ve lost interest in the remaining possibilities of IF, I think we’ve parted ways - because this is my starting point. W e need a collaboration between interactor and structure, rather than an adversarial model. When the interactor willingly conspires to lead Hamlet to his doom, and the design traps errors by generating suggestions and providing options for how to doom him (after his own fashion), then we have interactive tragedy and an aesthetics of frustration and error messages. My impression is that is what is already happening in IF now.

  6. 6 Jeremy Douglass

    For any visitors unfamiliar with The Most Lamentable and Excellent Text Adventure of Hamlet, Prince of Denmark, which is among other things a demonstration of Robin Johnson’s Nondescript javascript IF engine, I’ll mention here that it is definitely not what Mark Bernstein et. al. are talking about - not because it is good or bad art, but because it is screwball black comedy rather than tragedy.

    Who really killed Hamlet’s dad? What does King Richard III want with a horse anyway? And where did the gravedigger get that gorgeous pink dress? Avenge your father, defeat your evil uncle and ascend the throne of Denmark in William Shakespeare’s long undiscovered text adventure.

  7. 7 Christy Dena
  8. 8 Mark Bernstein

    Again: I’m not necessarily talking about an interactor who is trying to ignore or defeat the spirit of the work. I’m talking about an interactor who is doing exactly what we ask them to do: be themselves, or someone better than themselves, when thrust upon the stage.

    The character notes that make this tragedy Hamlet (and not just a puzzle) are disclosed through the working of the drama. How do we know how to (inter)act? And, if we have no real scope of action, if we can’t choose who we are and what we’re to do, what sort of interaction is this?

    It’s not just tragedy. Consider the thriller: we are minding our business one day when, suddenly, we see the wrong thing and we are unexpectedly in grave danger. They (the government, the Mob, the aliens) are after us. We’ve go to do something!

    What shall we do? There are lots of things we can try. We could slip away from our routine and check into a hotel under an assumed name, and we’re in Charade. We could slip into a dress, and we’re in Some Like It Hot. But it’s got to be our choice, doesn’t it: if someone *makes* you put on a dress, or if putting on that dress and join a female orchestra, if putting on that dress was the ONLY thing you could possibly try, then it’s not the same as choosing it yourself.

  9. 9 Jeremy Douglass

    But Mark, I’ve been saying repeatedly that I’m *NOT* asking the interactor to “be themself” - I’m asking them to role-play Hamlet, which is quite different, and my design will negotiate with them what that role-playing does and does not entail.

    If he wants to spy on Polonius from behind the curtain, or he want to listen at the door, fine - but if he wants to hide mother in the closet and meet Polonius wearing her dress, this falls outside the box of Hamlet, precisely because it is something the tragic Hamlet would not do in this particular rendition of him, according to my authorial concept of Hamlet as tragedy rather than screwball comic. How was the interactor to know this was my conception?


    You cannot bring yourself to touch your mother’s tainted goods.

    Hamlet might confront Polonius, or spy on him, or attack him, but the generic error message traps some other potentially valid “be yourself” creative impulses behind a general description of Hamlet’s character in this work - Hamlet will not use objects in the environment of his mother’s room because he considers them the spoils of prostitution, and using them would stain his honor. Note that the parser didn’t care which object was referred to or understand what wearing the dress would entail socially, however the message still coincides generally with Hamlet’ s odd sense of self-importance and ponderous dignity.

    I don’t think our discussion translates very well to a thriller, whose protagonists are quite varied in nature and which might be comic or tragic or otherwise - I’m responding specifically to My Friend Hamlet, and saying simply that I agree you can’t have it both ways. You can’t ask a volunteer Hamlet up on the stage, instruct her to “be herself” AND expect to arrive at a tragedy. There are more open (although still restrictive) architectures that allow the interactor to choose if the story will be Charade or Some Like It Hot - but these don’t answer the point of your original question - can we let someone onstage and have tragedy?

    Yes, you can - if many things are allowed, but putting on the dress is not one of them.

    You say it’s got to be our choice, and that being constrained is not the same as freedom - and it is true, they are not the same - but with constraint comes much of the power to craft genre. I’m wishing you luck in discovering a more “be yourself” way of creating tragedy, but this may not only be difficult but also unproductive.

  10. 10 Mark Marino

    I love that this conversation revolves around Hamlet, perhaps the least active protagonist this side of Beckett. I’m interested to in the value we place on tragedy, since it’s one of the few forms clearly classified by ancients.

    Jeremy, your idea of the craziness pushing the plot further seems to fit well.

    Let’s pursue the notion of frustration being constitutive element of IF, and in effect, that unrecognized phrases are going to be a part of IF.

    What does it mean to be unrecognized? To be unparsed?

    Hamlet cannot parse his environment. Perhaps to create the Hamlet experience, we need to be Horatio, Rosencrantz, Guildenstern, or even Polonius, trying to give advice to Hamlet. Put Hamlet on the other end of the engine, at some distance from us.

    >Tell Hamlet it is a dagger he sees before him.
    Hamlet turns, “But why would you say it is a dagger I see before me. There is no me before me.”
    >That’s not even blank verse.
    Hamlet whirls and marches off.
    It is a lost cause. In the corner, two itinerant actors flip coins.

    Sigh, comedy again.

    What’s interesting to me, in light of our other discussions of benchmarking, how many different ways one could render Hamlet in IF.

    You may have already discussed this, but Michael Mateas and Andrew Stern offer input to this quandary, by having a plot that rolls on inevitably with or without input. See their “Architecture, Authorial Idioms of the Interactive Drama.”

  11. 11 Mark Marino

    Jeremy, I can see that you are also distinguishing between the frustration at the interface (not being recognized, receiving repeated error messages), and the frustration within the narrative of not being able to find a resolution despite impending doom. I’m not sure how much pleasure we derive from the former (to each their own), but the latter seems part and parser of the genre. This line seems to mark part of the divide between you and Mark Bernstein.

  12. 12 Jeremy Douglass

    Thanks Mark M (now I see the danger of being on a first-name basis in comment threads) -

    While I like the idea that Hamlet’s legendary passivity and aggravating refusal to take action makes him a problematic role for the interactor, I’m not sure it is entirely fair to describe him as inactive. He has a busy time of it - the plot to hire the players, the substitution scheme against R & G, killing Polonius and driving Ophelia mad… and that all leads up to a sword fight!

    On the other hand, a plot that rolls on inevitably without him has a certain appeal. If the consequence of doing nothing is eventually being poisoned or being shipped off to a foreign land (and killed), that certainly helps clarify motive during subsequent traversals.

  13. 13 Mark Marino

    That’s a good point about Hamlet’s activity.

    I’ve been looking at Chris Crawford’s Chris Crawford on interactive storytelling. He’s got a few ideas to throw into the mix.

    One he thinks the user “should be able to make lots of dramatically interesting decisions” (53). This goes against my inevitable plot idea.

    Second, he offers a bad example of a story, or an example of a bad linear story about a knight who rides away from the castle and dies after falling off a horse. The story, he argues is unsuccessful. As a game, however, we would find this satisfying.

    Now Crawford is making a point about the difference between stories and games, but there are a few aspects of this that play into your post.

    When we play games our identification with the avatar raises the stakes of the action. The story is not: There once was a knight who rode on his horse, fell off and died, but instead: I am a knight. I get on my horse and make it ride. We pick up speed. I can barely control it. I fight with all my might to stay on. Ah, the agony as I fall to my death only to try again tomorrow.

    I’m stretching a bit, but you get my point.

    The second idea is that games give us the promise of something more beyond starts and stops, whereas the printed story (aside from serials) is the whole enchilada.

    In the first case, identification through play and guiding the character seems to add an intensity of plot that stories must work much harder to build.

    Being Hamlet, then, might be an intense experience even while walking down a hallway towards the King, never knowing what mouse might be crawling behind the tapestries.


  14. 14 Jeremy Douglass

    I haven’t read Chris Crawford’s latest, but here are some thoughts:

    First, your knight example implies that first person is more emotionally engaging than third - which it might be, if it is immersive. It might also be less engaging if the agency of the interactor is alienated (”Ah, the agony as I thought I was hitting the X button but was actually hitting the Y button”). I’m not sure identification is the key difference. Third person seems to create intense identification just fine, as witness how common it is to cry over novel and film outcomes vs how common it is to cry upon failing a game level.

    Agency, on the other hand, raises the intensity of immersion / alienation - although now that I think about it I’m not entirely clear the extent to which “identification” and “immersion” are continuous or distinct. I suppose I define identification as emotional metaphor - “we lost” meaning you identify with the sports team - whereas immersion is emotional simile - Scott McCloud’s great example from Understanding Comics being “He hit me!” being indicative of how the participatory immersion of driving changes the thought “His car hit my car!”

    As to your second idea, limits are tricky - sometimes the most limited experiences imply the richest offstage or off-page worlds.

    The final point, on the intensity of playing Hamlet, seems to bring us full circle. I think it is an open question - which is more emotionally intense, the experience of walking into a dangerous situation (first person) or watching someone else do it (third person, as in horror films, above)?

    The obvious answer might be first, but then, I think that depends on one’s stake. Which is more intense - walking into a dark alley at night, or watching a closed-circut video of your children walking into a dark alley at night? And can one fear for the video-game knight (who, after all, may die one thousand times) as much as we fear for a character in a script, who presumably has but one life’s course to run?

  15. 15 Mark Marino

    The point about first and third person makes a lot of sense to me, though I’m starting to think that moving an avatar around a stage is distinct but not separate from the effects of point of view.

    I want to emphasize the role of this character movement in one’s own mental construction of the “story” that occurs in the game. I’m sure someone has written about this more eloquently than I’m phrasing it. My question is what is the difference between the told and played version of that knight episode that make the first so “unsatisfying,” assuming it is.

  1. 1 WRT: Writer Response Theory » Blog Archive » Of Games and Hamlet
  2. 2 WRT: Writer Response Theory » Blog Archive » Computers in the Compostion Classroom (Great Debate)

Leave a Reply

thesis writing service