Interview With Tarn Adams



February 28, 2019

From seeing the city as an organism to laying it out according to Taylorist principles of efficiency, our urban imaginaries are always colored by the dominant knowledge-models of their time. As any glance at a smart city white paper will tell you, our current view of the metropolis is dictated by our obsession with bits and bytes—cities today are envisioned as information. Data is the twenty-first century governor: If only we accrue enough of it, and ensure that it travels freely, all our urban problems will be solved. Traffic will flow unimpeded, smart sensors will stop crimes before they happen, and pollution will recede in the rearview. In the smart city, populations and their environment combine into one perfect, cohesive, information-rich system. Of course, the reality is somewhat different: GIS datasets crash our computers, political slime gums up the works, and Russian trolls steal our secrets. In short, information-fueled omniscience is forestalled by the equally powerful force of entropy. We try to combat chaos with an internet of things, but our perpetual battle for order is in vain.
None of this would be news to the video game designer Tarn Adams. He recast the world as a battle between information and chaos way back in 2002 with the creation of his masterpiece Dwarf Fortress, a fantasy world-building game as rich in complexity as it is poor in graphic sophistication. Foregoing three-dimensional graphics, Dwarf Fortress’s fantasy landscapes are rendered entirely using CP437 text symbols (the character set of the original IBM PC): trees sprout up as upside down triangle symbols, water flows as a texture of blue ~s, dwarves scurry about as little orange smiley face symbols. But don’t let its digital austerity fool you—a whole lot of algorithmic brain power lies behind these simple symbols. Contrary to typical video games, Dwarf Fortress does not consist of a limited set of pre-designed worlds; instead, from an increasingly complicated set of initial inputs (like Sarah Winchester’s perpetual home building, Tarn continuously expands his game’s catalogue of parameters), each Dwarf Fortress game begins by automatically generating a unique and highly detailed landscape—its programming language specifies everything from geologically accurate ore deposits to the erosion capacities of its streams—and subjecting it to years of historical growth.
Emergent world at hand, the gameplay begins: with only a small expedition team at her disposal, the player goes about the incredibly difficult task of mining her digital landscape for building materials, shepherding pack animals, establishing rudimentary farms, harnessing the ideological powers of religion, and establishing a colony. Gradually, through this precise data manipulation, architectural form is imposed on the digital ether: rectilinear homes, castles, and processing centers turn a field of green Us into a minor city. This urban patterning lasts only so long, however, for Tarn, ever the masochist, has programmed into Dwarf Fortress a formidable set of enemies. Soon, ravenous goblins and carnivorous carp set upon the player’s fortress, war descends, entropy reigns, the screen becomes a Pollock rendered in CP437. Game Over.
To its hardcore fans, this difficulty-via-complexity is the appeal of Dwarf Fortress. The game’s unofficial motto is “Losing is Fun!” and it certainly seems true that watching the rise and inevitable fall of a civilization seems to constitute the game’s core pleasure. In fact, most players are so enthralled by the gradual entropic destruction of their digital societies that they produce, via screengrabs, fan fiction, webcomics, and other means, elaborate narratives of their campaigns, which they publish online for the reading pleasure of Dwarf Fortress everywhere. And so, like a Smart City sensor producing dataset upon dataset that no one will ever have time to read, Dwarf Fortress’s intensive data-driven worlds only produce more data, information-rich histories to populate message boards the world over.
We spoke to Tarn Adams—the genius behind the game, the man who any city’s data analytics department would kill to add to their staff—via email earlier this month. His lengthy responses came back within mere hours, as though procedurally generated. In the end, I suppose, all true great works of art are but imitations of their makers.

Paprika!: Lots of construction and management simulations model real life scenarios (city-building, theme park design, etc.). Why choose fantasy for this game? What do you gain from using a fantasy lexicon?

Tarn Adams: We’ve done games in other settings, mostly unreleased, and one thing you gain is not having to worry about accuracy in specific details quite so much (whether that’s technology or history or culture, etc.). Also, range (of all kinds: travel, weapons, sight, communication) becomes a problem the further in the “future” you go, and that’s hard to do well on a local grid. Including magical effects and miracles etc., that all comes down to taste, more or less. The game has done pretty well without having much of it to this point, but the planned push toward creation myths and magic allows us to play around with world-building ideas we’re interested in, and allows us to have more diverse systems acting together. That’s just part of the fascination that’s driven the projects from the beginning. For instance, we have a few immortal creatures (elves, etc.) in Dwarf Fortress right now, and that’s already introducing not entirely unexpected problems as historical populations are tracked over hundreds of years; the sort of thing that’s often hand-waved in a novel, but we have to deal with it head-on in a simulation on the computer, or everything falls apart or becomes at least unsatisfying. Not to say we’ve fully addressed these issues; I’m thinking about them again due to recent complaints, he he he.

P!: What are the fantasy worlds that have influenced the design of Dwarf Fortress? Have any real-world spaces also influenced its design?

TA: I think some of the influences are pretty clear—it’s a Tolkienesque critter. My brother (who works on the game with me) studied ancient history, so we see various Assyrian and Hittite matter there, but I’m not the one to ask about that. We also grew up with Dungeons & Dragons and Conan and all that, and I’m sure it rubs off in places. Romance of the Three Kingdoms was important, just seeing tons of characters moving around early in life impacted what we thought of as possible; back to Tolkien, The Silmarillion is similar. At some point, we just started hoovering in the whole of Wikipedia though.
In terms of real spaces and buildings, geez, I’d hate to drag anybody into the crap we currently have. I looked over various village plans and so forth when doing the town generator, but all of our city and building maps are so basic I’d hesitate to credit anything yet.

P!: Architecture plays an interesting role in Dwarf Fortress. It only minimally affects the world—designed “high quality” architecture can trigger a happy thought in dwarves and very well-designed architecture can “boost your fort’s overall architectural wealth,” but it doesn’t do much else. Have there been any occasions in Dwarf Fortress where the architecture has mattered in a bigger, unexpected way?

TA: There used to be the 7×7 cave-in rule, back before we had a third coordinate. Subterranean open areas 7×7 tiles in size would collapse, so it was important to leave natural pillars or build supports. That led to some surprises for people! There’s also all the route planning—it’s important to design for traffic, and you can get all sorts of problems that come from having work areas that are far away, or hallways that are too narrow. Architecture also matters with respect to invasions and wild animals. A lot of creatures can jump over a one tile space, or climb up a short wall, including regular human invaders. If you don’t think about this when designing your defenses, you can find yourself amused by your lack of foresight when the time comes.
I’d like to do a better job here, but architectural features have always been a weak point of the game. I have poetic forms and musical instruments generated in some detail now, but architectural features haven’t come up. Part of that is how the tiles work one at a time in the 2-D space; anything beyond a single tile is hard to express/organize/design/order/maintain/etc. This is why we do have quite detailed engravings, where players carve into a single wall tile, sometimes recurring images or designs. But no real concept like “arch” yet.

P!: Periodically, architects get obsessed with the idea of self-programming buildings, where a set of parameters determines the design in lieu of the architect’s hand. What do you think about the designers who take the approach you’ve taken with Dwarf Fortress and attempt to impose it on the physical environment?

TA: It sounds like I’m being pulled into some beef in the architecture community, ha ha ha. I’m not educated enough on the subject to really comment. I’ve read Phantom Architecture, a book collecting several unrealized projects, but that’s about it. In theory, on paper, or in a computer, I think it’s cool to mess around with whatever, but it strikes me that space and material are much more expensive and exclusive in the physical environment.

P!: The world of Dwarf Fortress, actually, seems highly relevant to environmental concerns, with resource extraction being a dominant factor in the game. Does Dwarf Fortress contain any lessons for our world?

TA: Although resources are mostly finite in DF, the scale of the few dwarves working on a given site is small enough that you don’t really feel that in a way that’s modern, unless you want to draw the more fanciful and traditional connection between digging deep and being attacked by monsters. There’s also no air pollution outside of localized miasmas and smoke, and the smoke doesn’t come from industry, which is all wrong. More general industrial air and water pollution is rough on the processor, especially with the 3-D grid; the way tech constraints bump up against teaching lessons should be kept in mind, though we do have to try harder. I say that, but there are lots of different axes along which we need to try harder, and no time. So, best to look to the real world, not our poor troubled game.

P!: That being said, you have achieved an enormous level of complexity by relegating the visuals to the very simple language of ASCII. Was this all just about being economical, or did there ever come a moment when the aesthetics of this form of representation became a serious factor?

TA: There are proponents for text-based graphics, and I happen to like that style as much as I like others (since I grew up on it, I have no difficulty parsing it, at least until the symbols become overloaded and ambiguous, which is a problem). However, DF began coming off the slow collapse of our similarly-ambitious 3-D game Slaves to Armok: God of Blood. We’d been using text graphics in many of our other side projects, and it felt like adding graphics would ruin our ability to work on the game, given our failure in the earlier 3-D attempt. The DF modders have done a good job adding graphics though, and it hasn’t impacted development negatively, but that’s a bit different from doing it in-house.

P!: On the subject of complexity—what do you see as the end game of Dwarf Fortress? How do you decide which variables to put into the game? Is complexity about elaborating as many variables as possible or about establishing realistic relationships between them? How do you decide when the game is finished?

TA: That’s not easy to answer, since it depends on everything that’s going on at the time, and also what we plan to do in the future, filtered through decades of practice. I try to keep the new features pointing toward the broader strokes of what I’m going to add later on, and I think about whether or not a story thread can pass through a given feature and its implementation. Relationships are more important than individual elaborate features; however, once a feature is sufficiently elaborate, you should probably already be thinking of it as several simple features related to each other, so that you can more easily expand it, so there isn’t really a difference, unless you are talking about over-emphasizing a single area, which is certainly something to think about. I don’t think we’re realistically planning on finishing the game anymore, in the sense of being done with it and moving on before we’re dead.

P!: Does a forever-evolving project like this have a core essence or basic artistic identity? When Dwarf Fortress was displayed at MoMA, for example, a video of gameplay was shown on a loop. Should the code have been on display instead? Or is it the act of playing the game that is Dwarf Fortress? What about the many worlds generated, all the stories that Dwarf Fortress produces?

TA: Part of the reason that MoMA showed it just as a video on loop, and didn’t make it playable, was that the game was deemed too complicated for a museum setting. But as to its artistic “essence,” I’m not sure I can narrow it down. I’m sure there’s some kind of artistic merit to the trundling mass of code (although in theory it would need to be released and discussed to achieve that status). At the same time, there is an “outsider art” argument to be made, especially thinking of the game early on. But now I think you have to consider how DF interacts with people at large. In addition to the story playthroughs and graphical output, there’s the people who use the products of DF to produce their own artwork—pen-and-paper campaigns, cartography competitions, etc. I sometimes joke that the people that read DF stories are the “players” while the people that actually use the game are more like playtester-writers, since it takes some time and isn’t immediately rewarding, while the stories that people read have a human-interpretation element to them.