"Massive Environmental Storytelling": An Interview with Mark Johnson

You mention Jorge Luis Borges as an influence, but also Umberto Eco and Neal Stephenson. What's their influence on URR?

Borges wrote a lot about a lot of themes which have so much potential in procedural generation: mazes, labyrinths, infinity, mirrors, life and death, the nature of truth and falsehood. Eco's best novels and work offers a kind of "cultural detective work" (or maybe semiotic detective work?) into understanding symbols and their meanings and how they circulate in society which seems a fascinating thing to base a game on.

Stephenson, meanwhile, has produced in his "Baroque Cycle" a very particular and idiosyncratic presentation of the Renaissance/the Scientific Revolution which really appeals to me, and there are some parts of that which I'm trying to emulate.

I’d wager that Eco’s Abulafia, the program that randomly ‘connects’ historical facts to suggest conspiracy theories in “Foucault’s Pendulum’, fit exactly in that mould … It’s a great way of looking at the relationship between procedural (in way: ‘blind’) generation and our brains’ need to connect facts and find meaningful connections, I think. However: Dark Souls’ understated and scattered storytelling does the same, but with carefully crafted text, not ‘blind’ - how will URR handle that? Will its narration, its plot rely on procedural generation as well, or is there a crafted ‘core’ to the game, to be replayed in a random world?

Abulafia is definitely a great example; although I read Foucault’s Pendulum long before I actually become interested in making my own procedural worlds, it definitely had a lasting effect. It’s very interesting to look at how the mind makes connections, particularly in looking at the attempts from players to “reverse-engineer” procedural games and try to figure out how they create their worlds. As for URR vs DS storytelling, the narration in URR will rely on procedural storytelling, though the basic core plot/quest/story is the same, but distinct each time: which is to say, the same basic story will be told across a different range of religions/cultulre/nations each time, so the story will emerge differently, even if the same basic themes - about historiography, “civilization”, metanarrative - are there each time. Then the information on that story should be distributed throughout the world in a Dark-Souls-esque manner, and leave the player to piece it together.

There’s also a question there though about how much variation should there be each time, and I intend to develop a system whereby the specificities of each generated world yield a very distinct version of the story each time, rather than different nations simply being different actors in the same play. Though I do realize that it’s a pretty nebulous distinction!

I like that URR is not a 'literary' in the way this is most commonly understood. Rather, the whole concept of procedurally generating a world is a very Borgesian concept. You name 'Tlön, Uqbar, Orbis Tertius' as an obvious inspiration (and in a way, you yourself, as creator of this project, could very well be imagined as a Borges character, I'd say.). Could you elaborate on that?

A few years ago, copyright expired on the works of James Joyce, and the very idea that somebody might produce a video game of his works was used as a kind of horrific worst-case scenario or even a threat . Do you think that we'll see more of a 'sellout' crossover between games and literature? Which projects would you like to see?

Ha, I didn't know that - I'm no Joyce fan, it must be said, and that does sound pretty horrifying. I don't think we'll see much crossover for a long while (books are not "events" like films are "events", so games based on films will surely continue to dominate?), but in terms of projects I'd like to see... I think there's a lot of potential in some very interesting SF works for games which are both mechanically and thematically interesting (which, for me, is the standard by which games should generally be judged).

Normally, if and when literature or even philosophy play a part in game design, they are only skin deep; Bad Mojo famously took Kafka's 'Metamorphosis' and made it into a game where you are a tiny cockroach. Do you see URR's as a radical new approach in that regard?

I'm trying very explicitly to make the themes and the gameplay the same thing. You cannot complete the game without understanding the themes; and you can't really get the themes without playing through them. The more the play understands the kinds of ideas about historiography and culture/semiotics I'm trying to convey, the "better" they will be at the game (in terms of skill).

It's tricky, and it's hard to pull one's brain outside the dominant discourse of "shoot a thousand dudes in the head" game design (as Jonathan Blow accurately puts it) and think "how can I make a game mechanic out of this interesting idea?".

With text-game tools like Twine easily approachable, we've seen a rise in 'experimental' narrative games, smaller, personal vignettes that might also be termed 'literary' in a way. URR is somehow the polar opposite of that - it's (don't take this the wrong way) 'impersonal' in that it relies on procedural generation, it's huge and also a massive piece of work. Do you see your game and this other approach as opposite sides of a coin?

Definitely. Twine stuff is cool, but I agree, I see it as in many ways the opposite of what I'm attempting. I'm reminded of that quote from Sergei Eisenstein along the lines of "protagonists should be eliminated": although you play as a single character in Ultima Ratio Regum, I'm trying to emphasize the lack of power of your single character (in some ways) and to make the game about far broader cultural/societal/religious concepts/questions, rather than the very individualistic history that games tend to write.

Although the player will determine the outcome of the world, there are a thousand other agendas at play, and how the player moves within and between those, whilst not necessarily having too much power to alter them and only having the power to use them or adapt to them, seems like an interesting thing to explore.

The game - and its myriad possible procedural incarnations - or rather, the algorithm creating content is very 'artistic' in that each unique generation is in a way a 'readymade', or even an objet trouvé, which fits nicely with the exploration gameplay. Did you have that interpretation as a piece of art in mind? (After all, DF is in the MoMA these days.)

Very interesting question. Hmm. I have definitely written the algorithms in the game in such a way that they should emphasize difference. That's the most important thing. In the real world, how much difference is there really between, say, a castle and a fortress? That seems like mostly semantics to me. On the other hand, in URR the differences are massive - so I've identified a few very minor differences and "scaled them up" so that there is a clear gameplay distinction between the two terms; a technique of that sort has been used throughout the game in order to give the kind of impression you describe, that every nation/religion/etc should be somewhat "artistic" in its own right, should stand on its own merits as something noteworthy.

There are a lot of special algorithms in the game which also delete content which is too similar to other content: so if a given world generates a religion with a certain symbol, for instance, the game can tell if another religion symbol is too aesthetically similar, and if it is, it'll make a new symbol, and keep doing that until it is different and interesting.

Eron Rauch wrote a bit about the problems presenting games like DF in a museum context here . Do you think that this 'appropriation' of games into an high-art context is problematic? Is it even possible?

Not at all, I think it is very desirable, and we (meaning intelligent and thoughtful game players/designers/academics/etc who want the best for our medium) should be pushing for that as strongly as we can. However, I see the points made in the article linked: for me, the biggest problem (especially with a game like Ultima Ratio Regum, or actually other things like Dark Souls) is that so much of the gameplay happens in the player's head. So much of "the game" takes place in what the player is thinking, imagining, figuring out, and that is damned hard to display on a static display (and, in many cases, hard to display on a gif or a looping video or whatever too). I'm not really sure what the solution here is.

The best case solution seems the same as you get in "exhibitions" of films: i.e. short clips and stills. It's not perfect, but I don't see a better solution; you can't get a random who wanders into a game museum to play Dwarf Fortress for 20 hours to figure out why it is so impressive, and maybe it's impossible to make a game like DF "meaningful" to someone who doesn't have the kind of required implied game literacy to really get anything from it? I'm not sure: this seems like an open question to me, honestly. I try to "sell" URR by displaying still imagines of interesting procedural generation, and to use the aesthetics of the generated graphics to lure people in and get them to experience the gameplay I'm working on. It's not perfect, but I can't really say "well, you'll be playing this game for X hours, and then you'll shortly come to realize that the game's AI has written basically thousands upon thousands of procedurally-generated aesthetic and textual puzzles hidden throughout the world..."

Using the rogue-like as your genre might seem quite surprising at first, as this ancient niche seems quite technical and mechanical to the uninitiated. Why did you go for that?

A short answer this time! I chose a roguelike at first simply because I found a tutorial for making a very simple one, and I'd never made a game before or even written a line of code so wanted something simple and easy, and I think it's a fascinating genre with so much gameplay potential; the thematic/literary ideas came much later, but seemed to mesh strikingly well.

In our series on roguelikes 'Mille cavernes', our very own Joe Köller argued that procedural generation is capable of producing the sensation of experiencing the 'sublime' in the philosophical sense, an overwhelming sensation of grandeur and our insignificance next to some rogue-like's sheer size, their difficulty and also their indifference toward their players. Would you agree?

I would fully agree. I think roguelikes emulate, in many ways, the depth, complexity and experience of "being in" (in the Sartre sense, maybe?) the real world; that the real world has some skill, some luck; some people are fortunate, some are not, some are "skilled", some are not, and the world is a hugely complex "thing" we don't fully understand.

Again, a Borges quote: "It is clear that there is no classification of the Universe that is not arbitrary and full of conjectures. The reason for this is very simple: we do not know what kind of thing the universe is". We don't really know what a roguelike is until we master it, and this generally takes a lot of time, and even then we have to use that same mastery to negotiate these challenging procedural worlds.

I've seen URR described as an 'ASCII walking simulator' (although that may be because of its unfinished state), and this was actually high praise in my eyes - an ASCII game capable of generating its fascination without relying purely on gameplay mechanics to keep players engaged is a great thing! Will the final game try to keep this focus on mere exploration?

Ha, yes - it is basically an absurdly detailed simulation at this point. That's because the world must be there first: you can't spawn an NPC until you know what religion they worship, what nation they came from, what clothes they wear, etc etc. I think it's cool that people do enjoy working on it, but yes, the final game is very exploration focused: you'll be trying to explore as much of the world as possible (only your home nation will start "explored") to pick up the clues and hints and information you need.

That doesn't mean it's going to be a "walking simulator" in the sense of games with no challenge/objectives/etc - it is naturally permadeath, and I want it to be extremely challenging - but that's a fair definition right now. But for not long! 2015 is the year of starting on gameplay.

I love the simple, yet impressive art the game generates; some of the larger images you shared look almost like ornamental mandalas. The game's aesthetic, its careful use of ANSI and color, seems very important to you. Can you tell me a bit about that aspect of the game?

It is very important in URR! It wasn't at first: it just came about by chance. I tried generating a few ASCII/ANSI images, and people just loved them, and I realized this was a way I could really make the game stand out and be distinct from any other game out there, really. Then a few months later I realized that not just could this make URR unique and distinct, but I could use this for gameplay.

Maybe you're told there's a clue in "the chamber of the four fragile dragons"? Well, go and find a room with four vases in it, and all four of those vases have generated with dragons on them, and you know you're in the right place. And then extend that to every single image and concept in the entire game. Now I'm trying to ensure that every image or shape or colour can be "transformed" into a clue or a piece of data, basically, and getting the player to examine these images and aesthetics and use them to understand the world.

I think it's something really unique and distinct - I mean, yes, of course we have those old "hunt the pixel" adventure games where you just have to somehow guess what the hell the game designer was thinking (looking at you, Toonstruck) and where they wanted you to click, or that they wanted you to combine the fish with the tractor with the plunger and then click on the box hidden in the top-left of the screen, and whilst these did use "visual puzzles", they were fixed: once you know the solution, you can instantly solve it every game. But getting the game's AI to procedurally generate visual puzzles?! That seems amazing to me, and it's a big part of URR.

URR and Dwarf Fortress are both hugely ambitious in their approach towards procedurally generating massive worlds, down to an almost microscopic level. I use to call DF 'the Alpha version of The Matrix'. What's the most interesting part of generating an environment at this crazy level of detail? Does the code keep surprising you as well?

The most interesting thing for me is integrating the macro - nations, religions, cultures, history, etc - into the micro of individual people, houses, what people have in their homes, clothing styles, etc. It's an awesome way to make this type of stuff not just lore or background like in so many games (I have a long-standing hatred of games that "tell their story" by just having you pick up little fragments of text or recordings or whatever - it seems so lazy) but rather something which is integral to the player's success. I gave a talk on that last year!

Do you think that procedural generation will migrate more into the broader games industry? 'No Man's Sky' comes to mind.

Yes - the phrase "with roguelike elements" seems to have become the new "with RPG elements", though most people don't really understand how to usefully integrate procedural elements into otherwise handmade games (though I'll be speaking about exactly that topic at the Independent Games Track at GDC Europe in August).

No Man's Sky is a great example and seems extremely promising: in other cases, though, I think AAA companies will have to really re-learn (or maybe de-learn older models first?) how to make games to make interesting and original procedural games. Then again, AAA games don't exactly pride themselves on originality these days, so maybe they'll be content to make endless clones of the same procedurally-generated FPS? Urgh.

DF demonstrates that a (relatively) simple set of AI structures makes for interesting (and sometimes scarily realistic!) behaviour. DF generates anecdotes, stories, even epics, and all from code, the players' minds filling in the gaps. Do you think that this kind of procedural storytelling will become more and more refined? How will URR address that?

Procedural storytelling is an interesting thing. At the moment we're clearly only just really scratching the surface - but the question is, are we only scratching the surface because generating stories is inherently harder than generating levels, or is that just because we've become so used to generating levels that very little time/research/money/effort/energy goes into generating narrative? It's hard to say, and it's like asking whether Go remains so tricky to write good AI for whilst Chess is now so trivial because Go is "actually" harder, or simply because 1000x more money is put into Chess AI than Go (or, at least, that was the case in the past).

I'm not really sure. I think it will definitely improve, and URR's type of storytelling (through symbols, signs, books, poetry, paintings, what people say, etc) is more of a kind of... massive environmental storytelling? I'm not really sure how I'll design it. I have an intriguing masterplan with taking the "stories" of what a given player character does in the game and then doing something incredibly cool with that story, but that won't be happening for at least a year, and probably more…

In my academic work, I used to work with Stith Thompson's 'Motif-Index' , a very thorough, systematic approach to narration in folk literature. Thompson basically analysed literature for what he called 'motifs' and categorized them, coincidentally in a pioneering IT project, using punch cards. Do you think that what we might call 'good narratives', eg those 'handcrafted' by humans, will ever be rivalled by procedurally generated ones?

I think that's a tricky question. From someone with my academic background - sociology, history, humanities - I want to say the answer is no, and there is something special about the human creative process which cannot be emulated. And I do think the answer is no for some time, decades, maybe even centuries. But then the fundamental question is: is there anything about the human brain which cannot be replicated, i.e. anything which is not physical and materialist? And I have to say the answer is no, and therefore I am inclined to think that some day the "work of art" (of whatever form) created by a computer will become indistinguishable from that made by a human.

For example, right now, Tommy Thompson (CS Lecturer at Derby) is doing some really interesting work into the human ability to distinguish level design from humans and level design from AI, and people seem to generally struggle. That's placing platforms in an infinite runner rather than writing novels, but I think that indicates where things are going, even if it's still very early days at the moment (compared to writing compelling/convincing fiction).

Wouldn’t the kind of ‘massive environmental storytelling’ you mentioned earlier be, at a grand enough scale, actually impossible to design by hand? I also mentioned Thompson because I’m kind of hoping that someday soon, there’ll be credible ‘narration engines’ that basically construct narratives meaningful enough for our brains to fill the holes with fleshed-out stories - kind of like Eco’s Abulafia was used as a ‘cue-giver’. Might URR be that game?