Shadows of Worlds Unto Themselves

In the early 20th Century, the last resident wolf pack in Yellowstone National Park was killed. Following human expansion in the area through the 19th century, chunks of land were taken for agriculture, diminishing the wolves’ hunting grounds, and driving them to prey on livestock and come into increasing levels of contact with people. The wolves were seen as something of a menace, an undesirable threat to more serene species like the elk. Consequently, they were targeted for elimination, remaining a rare sight in the park for much of the 20th century.

That is, until 1995, when the inherent value of species diversity was better understood, and a reintroduction effort was made. The effects of this reintroduction were, however, more far-reaching and significant than anyone could predict. While elk weren’t hunted in huge numbers, their behavior changed significantly. Being wary of the wolves, they avoided areas they had been exploiting, and moved around far more. This prevented overgrazing, particularly of willows, which they had been consuming in great quantities. The now burgeoning willow populations provided valuable homes for a variety of songbird species, and were used by beavers to build dams. These dams reshaped the entire hydrology of the river systems, providing yet more habitat diversity and regulating water levels in the area. What had been conceived as a species introduction blossomed into a veritable habitat restoration. This is all to say that natural ecosystems are highly dynamic and unimaginably complex.

Now, imagine you’re hunting in Red Dead Redemption 2, Rockstar’s Old West epic, often praised for the complexity and interconnectedness of its simulated North American ecosystem. You’ll be taken aback watching different animals come into conflict through the game’s clever emergent creature AI. You’ll watch as Arthur methodically takes down his quarry, and skins its corpse. You’ll receive some meat and a pelt to use in crating – and that will more or less be the end of it. Even if you devoted hours of time to hunting down and killing every last wolf, the game’s ecosystem will eventually always rebalance itself to a tidy equilibrium, one that’s focused on providing a tailored gameplay experience. These beautiful habitats, and the animals within them, ultimately exist in stasis, and as such will always harbor an undercurrent of unreality and sterility. So how can we bridge the gap between the complexity of the real natural world, and the games which attempt to emulate it: beautiful, impressive, but ultimately shallow simulacra? And should we even bother trying?

Looking at these issues from a “bottom-up” perspective can help us to better understand the magnitude of this problem. In the abstract adventure game Genesis Noir, players will journey from the first instants of the Big Bang, through the origins of life on earth, to the rise of human civilization – all presented through the lens of a noir murder mystery (no mean feat!). In one sequence, you witness the formation of the first cells, forming in the primordial soup of a much younger Earth. Phospholipid bilayers chaotically twist and collide, eventually forming the first cell membranes. Cells begin to mutate, acquiring features which cause them to differentiate from one another. These features constantly recombine, as different combinations of cells compete and reproduce with one another, creating a diversity of organisms within minutes.

This is, of course, a necessary oversimplification of the processes of mutation, evolution and speciation, yet it’s the first time I’ve ever experienced a game which presents the complexity of life from this kind of “first principles” perspective. Understanding that the labyrinthine network of interactions that exist in modern ecosystems has arisen through hundreds of millions of years of these kinds of competitive and complementary interactions gives an idea of the impossibility of attempting to disentangle the natural world as it exists today.

One effective way of circumventing these issues is with a fantasy setting. The master of this approach is the Monster Hunter series which, with each new iteration, provides new and vibrant ecosystems, filled with fantastical beasts which coexist and come into conflict with one another. Monsters exist in a hierarchy of power, meaning these interactions resemble something like a  food chain. There are also instances of symbiosis, such as the benign thunderbugs which inhabit the horn of the zinogre, allowing it to use their natural electricity in combat. These interactions provide a living game world which is believable, without the suspension of disbelief required when trying to emulate real-world locations.

An equally successful approach can be found within the robotic ecosystems of Horizon Zero Dawn. While the machines in the game are functionally the animals or monsters of the game world, they are also completely synthetic and programmed, and so any artificiality in their nature is entirely consistent within the world of the game. Any rote AI behaviors and walking patterns are completely believable, and are actually highlighted by the game’s focus system, in a clever metatextual trick. This approach works because so much life and character is imbued into these machines with their animations and sound design. They interact, have defined habitats, and even scavenge and forage, but only to the extent that their programming requires. In the narrative of the game, their design was inspired by the animals of Earth – the perfect narrative justification for an ecosystem which feels at once alive yet manufactured. The one downside of this is that the real animals which exist alongside the machines suffer by comparison, somehow feeling all the more unreal by comparison – shunted into the uncanny valley by a grumpy robot crocodile.

A big limitation of videogame ecosystems is their strong focus on predation and conflict. The interactions seen in Red Dead 2 are a good example: An intricately animated bird of prey plucks a snake from the undergrowth, while a bear fights off a wolf from its kill, or two stags lock antlers in a rut for dominance. These moments are inarguably well-realized, but they capture only a small slice of the life history of these creatures. In Rockstar’s defense, these are some of the more immediately engaging and exciting behaviors you might hope to see (hence their ubiquity in nature documentaries), but I would argue that the quieter moments included in these emergent behaviors – such as a bear rubbing itself against a tree and marking its scent – can do just as good a job at placing you in the world. Animal behaviors, particularly those of large charismatic mammal species which are often favored for inclusion in open world games, are rich and complex and it would be fascinating to inhabit a world in which animals mourn their dead, playfight, shun a diseased member of the group, or pick out a mate from the crowd.

The obvious hurdle to many of these points is the computational effort required. In my day job as a biological researcher I know first hand the difficulties with attempting to measure and quantify the myriad elements of a natural population. An animal’s age, sex, infectious status, genetics, among countless other factors, will affect how it behaves, and how well it survives, given the environmental context, which itself is constantly shifting. Interactions occur between and within species, between populations within species, and even between the parasites and bacteria which exist within an animal (which, of course will be different across a population). Trying to pick out clear patterns in all this noise is a constant issue with ecological studies, and so trying to maintain dynamic simulations anywhere near this level of complexity in a game is on its face an unreachable goal.

But I would argue that it’s a goal worth keeping. By dedicating a provision of design and computational resources towards this end, developers can provide a world that really feels lived in, where the wonder of seeing a herd of horses cross an empty moonlit field in Breath of the Wild is something that could be repeatedly captured, but never exactly repeated. Where killing the last wolf in the map would have actual consequences. These choices would almost by necessity need to de-prioritise the agency of the player – but if used in the right way, they could be a benefit rather than a burden. Imagine discovering an animal population with a deadly new disease, and being torn about whether to destroy them, not because a local quest giver asked you to, but to protect the species at large, and ultimately preserve your own game world.

As with many tricky elements of game design, these ideas have been explored outside the AAA space, with games like Sapling, Eco and Creatura all attempting to present worlds which more closely follow actual laws of nature. I would, however, love to see this approach spearheaded by some larger, industry-leading studios (maybe the long-gestating Everwild will be something along these lines). The Assassin’s Creed games have in recent entries taken painstaking care to recreate accurate historical settings, to the point where they have been used in educational settings – yet they still succeed as hugely-successful mainstream action games. Hopefully studios will start to take similar approaches with their recreations of the natural world too. There is an ongoing question surrounding the diminishing returns from constant investment in high-end graphics, which invites creative approaches to how else increased computing power might be used to tackle other elements of game design. I would suggest that breathing life into a world where your players may spend hundreds of hours is an investment worth making.

———

Jonathan is a biological researcher by day, but spends much of the rest of his time obsessing over games, music and music in games. You can follow him on Bluesky and at Medium.