The paradox of procedurally generated video games.

The Paradox of Using Algorithms to Create Infinite Video Game Worlds

The Paradox of Using Algorithms to Create Infinite Video Game Worlds

The citizen’s guide to the future.
Oct. 5 2016 7:45 AM
FROM SLATE, NEW AMERICA, AND ASU

The Daggerfall Paradox

Algorithmically designed video game worlds are more personal because they lack personality.

Daggerfall no man's sky.

Photo illustration by Slate. Images by Hello Games and Bethesda Softworks.

Few video games have generated more furor and frustration than No Man’s Sky. Featuring literally quintillions of explorable planets, all of them generated algorithmically, the title—from a tiny British developer called Hello Games—aimed to be as big as the cosmos itself, thanks to a technique known as procedural generation. In the wake of its August release, players who’d spent years growing evermore enthusiastic about the game were left disappointed by the dull sameness of its starscapes. Though reviewers found something sublime in its endlessly unfolding universe, angry complaints cluttered gaming message boards across the internet, and many players sought refunds, even after sinking hours into the title.

No Man’s Sky’s fall from the heavens remains 2016’s largest gaming story yet. Much as it’s a tale of our times, however, it’s also one that’s woven deeply into the history of gaming. Though Hello Games—which has largely gone silent online—is arguably a victim of the hype cycle it helped initiate, its fate was predictable for reasons that have far more to do with the tools it used to create its game. Indeed, looking back, one sees its destiny written in constellations assembled long before the company coded its own stars into being.

Advertisement

Understanding what went wrong—and what might emerge from the rubble—entails looking back 20 years to a largely forgotten PC game, The Elder Scrolls II: Daggerfall. When it arrived on the scene in 1996, Daggerfall—a first-person role-playing game that served as a sequel to the earlier Arena—stood out for the size of its playable world. Half as large as Great Britain, and peppered with tens of thousands of towns, dungeons, and inns, its virtual terrain remains one of the largest game worlds ever assembled. Though there was plenty to do and discover within its borders, and though your actions could change the destiny of the kingdom, that terrain didn’t feel as if it had been specifically built for you. It was simply there, and your adventures were almost incidental to everything else that was happening.

To achieve that sense of realistic scale, Daggerfall’s developers relied on a strategy conceptually similar—though more primitive in practice—to the procedural generation that assembled the enormous universe of No Man’s Sky. To say that a game is procedurally generated is effectively to say that its designers randomized its world instead of laying it out by hand. In Daggerfall, for example, the game would autonomously assemble dungeon maps from a wide array of smaller blocks, making each cavern or catacomb’s layout unique.

Theoretically, certain algorithmic rules guide the connections in procedural games—to ensure that, say, the topographic features of a planet in No Man’s Sky interlock properly. Such systems don’t always function well in practice: Some of those Daggerfall dungeons initially failed to include a door that would let players back out into the wider world. But implemented properly, procedural generation allows developers to include more playable environments than even the most massive team of developers could ever assemble.

Daggerfall was not the first game to rely on procedural generation. In a 2015 New Yorker article that helped bring mainstream attention to No Man’s Sky, Raffi Khatchadourian points to an especially prominent antecedent, the 1984 space adventure title Elite. Earlier games, including the influential 1980 dungeon crawler Rogue, had also employed randomized design mechanics to similar effect. And many, many more games have embraced such methods since then.

Advertisement

Despite that rich history, Daggerfall still makes for the most apt point of reference largely because, as with No Man’s Sky, the results of its procedural world-building are dull for the very reasons that they’re sometimes astonishing. Though a few gamers have noted the spiritual bond between the two games, discussion of those links has been largely absent from most prominent discussions of the newer title. There are, to be sure, differences aplenty between them. For one, Daggerfall is more random than its distant descendant. In Daggerfall, each player will see slightly different things at most locations; in No Man’s Sky, players can theoretically visit the same worlds and witness the same sights, but there are so many locations, it’s unlikely to happen often. Nevertheless, the sentiments they aspire to produce—and the ways they produce them—are strikingly similar.

To understand how Daggerfall’s world—one where I spent countless hours—came together, I called up Hal Bouma, a coder who helped create the game’s map. Fresh out of college in 1995, Bouma joined Bethesda Softworks and the game’s team when the project was already well underway. Twenty years later, the Elder Scrolls series is one of gaming’s most venerable franchises—thanks to the cult classic Morrowind and the massive successes of Oblivion and Skyrim.

Scores of programmers, artists, and writers now labor over those games and their worlds. Daggerfall’s team was comparatively tiny, despite its massive ambitions, and procedural generation helped it overcome that limitation—as it would years later for the similarly small Hello Games. Bouma struggled to explain why Daggerfall’s massive scope felt so necessary at the time. “It was just to feel open-ended,” he told me, suggesting that they were trying to create the impression of really living in a massive fantasy setting without forcing players along a single path. The trouble with that approach was that everywhere you went ended up looking relatively similar: Towns had few characteristics to distinguish them, and dungeons were rarely memorable.

According to Bouma, this feeling of repetition didn’t trouble Julian Lefay, the project’s lead designer, since it actually contributed to the world’s realism. “He would talk about how if you’re driving through suburbia, all the towns are the same. He felt it was OK for the game to be that way,” Bouma told me. If anything, Bouma now wishes they could have implemented more coherent consistency to make the game’s thousands of destinations feel more like actual locales. “If we’d gone further down this road, I would have asked that we add more blocks specific to crypts so that if you’re in a crypt you feel like you’re in a crypt,” he said.

Advertisement

Even without such specifics, the Daggerfall team’s approach could yield compelling results. Playing the game shortly after its release, I took its open-ended mission to heart. I recall traveling to the bank of an outlying province, a region I didn’t expect to revisit, and taking out a loan so large my character could never repay it. Back in the comforts of a more central city, I invested in a sprawling house. Much later, when a mission called me back to that distant capital, its streets were crowded with angry bounty hunters, agents of the debt collectors thirsty for my blood.

My outlaw adventures get at what makes procedural game design—especially really bold stabs at it like those attempted by the Daggerfall and No Man’s Sky coders—special. As Bouma put it, each player’s experience in the world of Daggerfall was genuinely idiosyncratic, shaped by individual choices and random design details. “What came out of this—for No Man’s Sky, too—is that your game, the game you play, is unique to you,” he said.

While procedural generation eases some elements of the design process, that resulting specificity imposes difficulties of its own. When something went wrong for a player—and in the bug-infested Daggerfall, it too often did—it was often difficult for Bouma to determine who or what was at fault. Because he couldn’t re-create their experiences on his own, he would have players literally send him their save game files, and he would try play through the problematic section in an attempt to discern what had gone wrong. (“I wasn’t a proficient player. I would literally have to activate invulnerability mode in seconds—otherwise I would be killed,” he told me.) He and his collaborators had, in other words, created a game too complex to be managed on its own terms, a digital land that they could only explore by peering through the eyes of those who were traversing it.

While that’s a remarkable approach, it leaves game designers in the position of absentee gods, powers who refuse to meddle in the universes they’ve made. Rereading the New Yorker’s article on No Man Sky now, few beats feel more ironic than Khatchadourian’s observation that Sean Murray, the game’s lead developer, “left a successful career with Electronic Arts” because he was “frustrated by the impersonal quality of corporate game development.” Whatever its successes, the game he produced after making that shift is almost painfully impersonal. Without a meaningful hand to guide you, it leaves you only with the pleasures of getting lost.

Significantly, many who’ve been left cold by No Man’s Sky complain that it’s a fundamentally lonely game, especially since you can’t explore its planets and moons with others. Even if you could travel with a friend, however, it would still feel desolate, if only because the world around you seems untouched by human hands. As in Daggerfall, that very absence—that feeling of radical isolation—allows us to extract experiences from the game that are wholly our own, whether or not we take any pleasure in doing so.

Call it the paradox of procedural gaming: The less personality a game has, the more personal our own time with it can become. This isn’t, of course, true of all procedural games, but it may be an inevitable consequence of Lefay and Murray’s admirable, maddening desires to render whole worlds. However much computing has changed in the intervening decades, one truth remains constant: Then as now, it’s precisely when designers go big that they leave us to be ourselves.

This article is part of Future Tense, a collaboration among Arizona State University, New America, and Slate. Future Tense explores the ways emerging technologies affect society, policy, and culture. To read more, follow us on Twitter and sign up for our weekly newsletter.