Chris Crawford on Interactive Storytelling, second edition (Chris Crawford)

Chris Crawford on Interactive Storytelling (Chris Crawford). Long ago, I wrote about the first edition of this book. Since then, I got to know Crawford a bit, and attended one of his Phrontisterion sessions; he solicited IF community input on a second edition; he kindly sent me a review copy of the second edition, but I was too busy to get to it (this feels like my permanent condition of existence); I then failed to write anything about the second edition. I also wrote recently about Chris’ Encounter Editor project. But now, as I’m going back and looking at the variety of writing available on this topic, I’ve taken that review copy down so I can finally do what I should have done a couple years ago. Apologies, Chris.

In the interests of getting a fresh perspective on the book, I have read it without referring back to the first edition or even rereading my review of the first edition.

This lead to some thoughts like: was this in the first edition? I don’t remember it from the first edition. I feel like I would remember if I’d read this before:

“Your designs should aspire to the ideal of metaphorically having sex with your users.” (37)

especially as it is followed a few pages later by

“The overall quality of an interaction depends on its depth as well as its speed.” (40)

But these moments aside, the second edition is — as I remember from the first — opinionated, sometimes correct and insightful, but also at times reductive, patronizing, or willfully indifferent to work done outside his own sphere. Crawford is passionately committed to the procedural power of computers, and so pays even less attention than usual to related art forms like immersive theatre, tabletop RPGs, LARPs, or other rules-based interactive story structures. Indeed, he thinks that if you don’t share his emphasis on procedurality, you don’t belong doing computer-based interactive storytelling, or you are, in his words, “prostituting your Muse.”

He also doesn’t have much time for standard video games and largely ignores their contributions. Unlike many of the other contributors to the advice-for-interactive-writing genre, Crawford is no longer actively engaged with the game industry and has no interest in perpetuating industry norms, outfitting readers to become professional games writers, or staying on good terms with potential employers. He sees himself rather as a prophet unappreciated in his own era, a quixotic seeker who may not have time to find what he’s looking for before he dies, and his aim is rather to inspire and instruct the next generation on a quest for the holy grail of storytelling. He tends to assume that the reader of his book is technically very ignorant, and talks down on a regular basis, including little admonitions to get over one’s fear of elementary algebra, even if one does not in fact experience such a fear. And his view on women, expressed at several points in the book, is that they are socially gifted but nervous of math and possibly logic in general; and will need to be bucked up a bit in the numbers department in order to make the contribution to interactive storytelling for which they are otherwise destined.

At the same time, there is much that I recognize. Chris shares my own desire for a future of interactive storytelling full of dynamic, persuasive characters who react to the player in rich ways. In that pursuit, he’s brought passionate commitment and dedication over the course of decades, and made many attempts, both paid and unpaid, to build what he wants to build. We are very different in temperament, style, and our approach to projects. But the purpose that underlies this book speaks to me much more than the purpose underlying a lot of game writing manuals about how to keep your game cheap and appeal to the broadest audience. The latter considerations are valid, to be sure, but I wouldn’t be in this business in the first place if it weren’t for the thing I want to achieve.

So I read through a haze of mingled familiarity, sympathy, and exasperation. I don’t think any of that reaction will surprise Crawford himself: the review copy is inscribed to me with a wry note acknowledging both my help and the fact that the finished product isn’t likely to be universally pleasing to me. Here is what I found.

In the early parts of his book, Crawford strongly argues — and I would agree with him here — that you need to understand your verbs well to make a mechanically compelling interactive storytelling experience. He emphasizes that the first question in software design is “What does the player/user do?” And he also argues (and I would again agree) that the player’s actions at any given moment can or should be contributing to the longer-term outcome of the story. (This is an area where Choice of Games and Chris Crawford would rather unexpectedly align, though Chris would probably not be satisfied by ChoiceScript’s level of procedurality.)

Here’s a nub where Chris’ advice is especially far from my own experience:

A storyworld does not offer the player a few momentous decisions; instead, it offers many small decisions.

…

The fourth element of interactive storytelling is that the decisions must permit unsatisfying outcomes… If the player wants to deviate from the plot, let him deviate! The mistake lies in setting up a plot in the first place. (58)

On point one, I get the desire to use stats and build up a cumulative effect for every player action, building the sense of agency.

At the same time, what we found in Versu was that you could have a system where the player was able to make a character .25 more annoyed every turn, and it wasn’t actually all that satisfying to prod someone four times and have their rage go off suddenly on the fourth.

No, what you wanted was a sense of rising stakes, an awareness that as you kept prodding, you were coming closer to the border of something. Feedback from the fiction that the mechanical boundary line was close at hand. Maybe the momentous interaction is only possible because the player has built towards it through many prior interactions, but it’s still important to have one there, something memorable and rewarding rather than a damp repetition of many equally low-significance moves. After all, doing the same action over and over again in order to make a small incremental progress towards a threshold is pretty much the definition of grinding.

Most tabletop role playing games I’ve tried are in some sense aware of this too — hence concepts like drama dice or the “but only if” mechanic in Polaris — which enable some moments to be more colorful than the norm, or let players raise the ante on a particular interaction.

As for the wholesale denial of plot: no no no.  I share Chris’ view that the player need not be guaranteed success, but that’s not the same thing as giving an unsatisfying result for failure. A consequence of a major action in a story should always be satisfying, even if it’s the perverse dramatic satisfaction that comes of seeing your kingdom dissolve in fire and war. I’ll come back to this one later as well.

What else is in this book: quite a lot of swift, shallow summary of fields as diverse as astronomy, evolution, theology, law, and finance, generally to draw some kind of parable about human agency and understanding. There’s a sermon — familiar to those who know Crawford’s work — about how one needs Math to do interactive storytelling effectively, but alas, most humanities people fear math, and being female likely makes it worse. There’s some advice for both technical people and artists which anyone on either side is likely to find irritating. (E.g., to artists, “You want to avoid C++ and the professional languages, but there are plenty of cute little programming languages that are easy to learn and fun to play with.”)

Amazingly, this chapter ends with a further sermon on why Europe is better than the US due to its elevation of Culture. Now, I like Europe; I live here. But it definitely contains outlet malls and service stations as well as galleries, cathedrals, and museums. Now perhaps there is slightly more of a tradition of really art-game projects coming from Europe than from the US (and I’d want to run some numbers on that). But if there is an effect in need of explanation here, I’d guess it’s because government grants and socialized medicine make such pursuits economically (sort of) viable, and not because of a lack of arty people in the US.

Anyway. I’m not quite sure what purpose is served by Crawford’s fantastic collection of generalizations and stereotypes. And some of them actively reinforce existing gatekeeping in the industry.

*

In the Part III, Crawford takes on several forms he claims do not succeed at interactive storytelling at all: branching narrative (meaning CYOA, hypertext, gamebooks et al); interactive fiction (by which he means the parser-based sort, with a focus on Inform 7); CRPGs; and video games.

One, branching narrative, he writes off on the grounds of combinatorial explosion, before eventually admitting that state variables help with the problem. It’s a very shallow coverage even of the kinds of structures you find in CYOA and gamebooks.

Second, interactive fiction. This chapter I had seen before: Chris came by the IF forums while he was writing it asking for advice, and I offered some direct input as well. He gives an overview of Inform, as well as some thoughts on Make It Good and Blue Lacuna. In his view, Inform still falls short because it is not focused on mathematical formulas (though he admits TADS is more programmer-ish and mathematically oriented). Blue Lacuna he refers to as “probably the best IF has to offer” and admits the complexity and variety of Progue’s emotional effects. Then he concludes:

IF has become a serious form of literature. It does not, however, constitute interactive storytelling in my opinion. The answer to the fundamental question, “What does the player do?” always boils down to “Solve puzzles.” The puzzles these days are more literate and more meaningful; sometimes they advance the narrative in a useful manner. But the player doesn’t interact primarily with characters; the player interacts with puzzles. (139)

I wouldn’t say this was true of every piece of parser IF (and Crawford is focused on parser IF here; Twine would fall in the branching narrative category). He also doesn’t give any attention to visual novels or dating sims at all, despite that being a very relationship-centric genre. (What exactly he would make of Black Closet, I’m not sure.)

But he’s still rather kinder to IF than to RPGs or to video games. The latter category he criticizes by proxy, going through Interactive Storytelling for Video Games and explaining what he dislikes about each of their methodologies; this takes the place of Crawford doing his own research into the state of the art in video game storytelling, or even addressing any of the many other books about this topic. The video games chapter winds up being a gratuitous hit piece on the book (“If you want a job making the same old canned stories with clichéd heroes fighting clichéd battles as part of clichéd quests, Interactive Storytelling for Video Games is the book for you”) without amounting to a very interesting or serious critique of where video games are now.

This chapter caused me to immediately order Interactive Storytelling for Video Games from Amazon, because I am a stubborn individual.

*

In Part IV, Crawford starts talking about his own recommended methodologies: modeling emotional states and reactions numerically, for instance. The section on modeling is largely about how to write formulas to express ideas that are otherwise usually handled with a heap of if/then statements: for instance, for

IF ((Mom is strong enough) AND (Mom is desperate enough)) THEN she lifts the car

he would have us substitute

IF ((mom_strength*mom_desperation) > threshold) THEN she lifts the car

While this approach does tidy things up in complex code, it’s harder to read in the simpler cases. This is really only worth bothering with when you’ve had lots of prior in-game opportunities to alter Mom’s strength and desperation, and need to allow for the system to include incremental variations in Mom’s strength, for instance, in the same way that ChoiceScript’s FairMath helps an author produce cumulative agency over the course of a long interactive novel (though that requires some additional art to plan and tune). In the next chapter, Crawford talks about bounded numbers, which also gets us into FairMath territory; then, under personality traits, about building a concise and orthogonal set of traits for a character description.

The next few chapters are, to my mind, the meatiest in the book and the most likely to introduce new ideas to the reader. Chapter 15 covers the concept of drama management and environmental manipulation of an ongoing story, as well as the idea of scoring the player of a storyworld on how dramatically she has played, rather than on “winning” the story. (“In storyworlds, the reward is applause, not victory,” as Crawford puts it (225).)

Chapter 16 gets into Crawford’s ideas about verbs: how many you need for a good storyworld (an absolute minimum of a hundred, he says); types of information that you need to know about a given verb (does it take direct and/or indirect objects and if so of what kind; does it take time to execute; can other people see it in progress and if so how). In fact quite a number of these considerations line up pretty well (if not always in quite the same terminology) with what goes into defining the parsing, affected objects, and scope of reporting for a new Inform verb.

He also gets into recording the action history of the world (what’s happened so far?) and the problem of subsequent reportage (how does one character later tell another about what has happened and the flow of causality within a story)? — as well as mechanisms of gossip, secrets, and lies. This is an intriguing and fraught territory (see also some of James Ryan‘s work on spreading information through networks of NPCs); Chris mostly concludes that it’s difficult, which I can’t argue with, but also didn’t find very helpful.

In the subsequent chapters in Part IV, Crawford gives an overview of Façade; introduces “tinkertoy text,” essentially madlibs at the shallow end of generative grammars (see also Expressionist and Tracery); talks about desirable features for any kind of storyworld scripting language (an idea he lays out in some detail); and tells some entertaining war stories from his own attempts to build an Arthurian storyworld.

*

Part V, “Wrapping Up,” gives brief coverage to a few academic projects; Prom Week is mentioned here, for instance, and Crawford admits that it has a number of promising features, though it doesn’t go far enough for his tastes, in particular because, as he says, “these are not dramatically interesting characters.” Having got to the point where the characters do all interact with a wide variety of defined verbs and respond dynamically, he finds that he isn’t that excited by the result. He himself would say that this is because Prom Week is doing it wrong, but when he tries to explain how it’s wrong, he falls back on rhetorical flourish rather than analytical critique:

You don’t have to pull everything out of the research literature… If you don’t grasp the big picture of storytelling, whole and complete, in your mind, what in the purgatory are you doing working on interactive storytelling? …as long as it is fundamentally a work of computer science, as opposed to a work of art, it will never bridge the uncanny valley. (318)

And this to me feels like a key moment of the book, whether or not Crawford recognizes it as such. Prom Week does so many of the things Crawford advises: it provides many verbs, all of which affect the social model and knowledge state of NPCs; it propagates characters’ reaction to each other through the social network, allowing for a concept of reputation; it sets goals for the player; it remembers and calls back to interpersonal history; it robustly handles many types of ways the player might manipulate characters in the environment. It’s highly complex and ambitious. And I agree with Crawford that it is not particularly emotionally compelling or dramatic.

But to my mind the “problem” here — if it is a problem — is not to do with the authors of Prom Week being soulless technologists with no interest in art. The ad hominem bit doesn’t stick: a major contributor of Prom Week was Aaron Reed, the same person who produced Blue Lacuna and the Ice-Bound Concordance among others; elsewhere in the book Crawford praises his other work. Meanwhile, I’d be inclined to direct a lot of the same criticisms towards the (mostly unfinished) storyworld work of Crawford’s own that I’ve seen so far. I think, rather, that Prom Week is doing quite a lot of what Chris Crawford thinks should work, and it doesn’t suffice to produce the experience he thinks he’s moving toward.

I have some theories about why this is.

One has to do with the concept of stakes and thresholds I mentioned above, talking about Versu. In order to navigate the storyworld confidently and with a sense of drama, the player not only needs to be able to guess the likely consequences of an action from their internal social model. They also need feedback from the game that the path they’re on is an interesting one with likely results of a particular kind; they need a sense of rising stakes, and a foreboding of probable consequence. This is something that IF does a lot, especially in the branching narrative/quality-based-narrative space Crawford has so little time for. Jon Ingold has talked a number of times about related mechanics, and this old post of mine also talks about some structural methods of approaching this within a choice-based system. While that particular implementation is focused on things one could rig up in Twine, it would be possible also to use Crawford’s concept of a “Fate” drama-manager (chapter 15) to handle informing the player about major plot points or outcomes that are within a step or two of being triggered.

That way, the player can decide whether to move towards or away from those outcomes: decide to say that one more insulting remark that will make the duelist call you out, or to hurriedly backtrack; to try to find exactly the right combination of kindness and discretion that will make the diplomat like you but not give away a state secret. Prom World doesn’t do a lot with this sense of escalating consequence.

My other doubt about Crawford’s method has to do, ironically, with the design of the verbs themselves.

While I agree with Crawford that one needs to plan around what the player can do, and that you get a lot of leverage from a consistently-behaving procedural system, I feel there’s a missing element of systemic design in his description. We have a bunch of verbs, the verbs move emotional and other stats, those stats are chosen on the basis of the themes and genre of the story we’re telling. So far so good!

But backing that, if we want to have interesting choices fall out of the system naturally, we need that system to be a good piece of game design still: one in which different types of strategic and tactical advantage are possible, in which tradeoffs become meaningful, and through which risks, rewards, resources and probabilities can be expressed. There are moments when Crawford’s discussion seems to touch on this idea: in the first part of the book, he acknowledges the importance of mostly balanced choices and constraining the player’s options to things that are currently interesting. But for the most part he is trying to separate storyworlds and interactive storytelling from the concept of games, and he eschews all of parser IF’s developments in this space as “puzzles.” It’s not clear how the simulations he describes in the second half can be guaranteed to be throwing up interesting choices. When he talks about designing the numerical consequences that should be associated with any given verb, he talks about how to express concepts like “people enjoy being complimented” or “people’s likelihood of believing a lie depends on their gullibility and their opinion of the speaker,” but not so much about how those might be balanced with respect to the storyworld’s array of available plot points and consequences.

In part, his interest in player expressiveness (an interest I share) has led him to conclude that a very large number of individual verbs are required (here I disagree). And I wonder if that’s created a stumbling block simply because it is hard to reason systemically about those numerous verbs and how they fit into a coherent system. There are other ways to get an expressive experience out of a smaller number of overall verbs (e.g., by allowing the player to double up inputs in a single move, or of course complex verb-noun combinations).

In any case, it may be that my own biases are at work here again. But in my view, Crawford has put aside traditional game mechanics because he thinks they’re irrelevant to the more artistic and elevated domain that interests him. However, (if well designed and well expressed to the player) such mechanics are precisely what would provide the sense of escalation, foreshadowing, and drama that would bring his storyworlds to life.

*

Side note: if you’re interested in the types of experience discussed in this book, you may also be interested in the work of Nicolas Szilas.