How Evolution Can Make Video Game Minions Into Big Bads

Biologist Barrie Robison and computer scientist Terence Soule are on a mission to prove a radical, unique hypothesis: If virtual bad guys change and adapt according to principles of evolution, games will be more fun. Together, they are the faculty leads for Polymorphic Games, a design studio at the University of Idaho that draws on the expertise of students from a variety of disciplines to build awesome, biologically inspired video games. They aren’t exactly the Darwin and Malthus of gaming, but they’re as close as the industry is likely to get.

The program’s most recent success is a arcade-style game called Darwin’s Demons; it’s similar to Space Invaders, except the enemies evolve as you move level to level. Each creature has a genome that controls what it looks like, how it moves, and how its weapons function. At the end of a generation, the “fittest” enemies — the ones that player struggled to murder — produce the offspring that attack in the next round. This means players’ weakness are isolated and exploited. Player failure and decision making drive selection.

Like Space Invaders, the game gets progressively harder as players level-up. Unlike Space Invaders, every level after the first looks different each time the game is played. That changes the nature of in-game decision making and a more dynamic product.

Inverse talked to Robison and Soule about how they’ll prove evolution makes for better games, and what happens when clever gamers outsmart evolution.

Why did you want to put evolution in video games?

Barrie: I think the answer to that question is so obvious that I just — I don’t even know how to state it. I feel like putting evolution into video games is similar to like back, say, 20 years ago back when people started putting physics in video games. It’s kind of a logical step, a logical system that can be put into games that have the potential to have them be really — a lot cooler, a lot better, explore different spaces and things like that.

Terry: It will make the games better.

But how do you know it will make them better?

Barrie: [Laughs] All right, Terry, you got us into this mess…

Terry: Okay. So, it will make the games better in the sense that the opponents will respond in more intelligent ways to the player’s decisions. So the opponent will literally adapt to whatever the player decides.

Barrie: I liken it to sort of procedural generation of enemies rather than environments. So you know how Minecraft pioneered the procedural generation of worlds? The player knows there’s a system and these biomes are created by some system that’s coded into the game engine. And so there’s predictability, and yet there’s new discovery every time.

A Darwinian Demon is a concept that describes a hypothetical organism that evolves without limits. How does that idea relate to your game?

Barrie: Left without trade-offs or limitations, evolution produces something terrifying, if your enemy is evolving in that way. And we thought, “Well, that’s an interesting kind of metaphor for an enemy in a video game.”

And, eventually, the enemy will always evolve to the point where it’s unbeatable, right?

Barrie: Yes with an asterisk: We discovered something recently that we’re now trying to fix.

Some very clever players — I mean these are like super, ultra gamers, and they knew it was an evolutionary game ——— set out to domesticate the aliens. They shot all of the nasty ones first and then they let the dumb ones that didn’t fire very much and just stayed up at the top and didn’t do anything live for a really long time. And,, after several generations,, they had space cows.

We were like, ‘As scientists, that’s fantastic; as game designers, that’s less than fantastic.’ So we’re now running some experiments to try to collect some data on, “Okay, do we want to nerf this strategy? How do we do that while staying true to our evolutionary model?”

In theory if the player plays the game like most players do, evolution will win. They will lose. It’ll be like Space Invaders and eventually you will lose. But if you’re clever about it, you can farm these things.

Terry: I think we would like to keep it where that’s a strategy that will allow you to survive longer, but not indefinitely. That was the real problem — it’s like, ‘Okay, now the game is boring.’

But this definitely represents a teachable moment right? The smart gamers learned how to steer evolution.

What we’d like is in the first stage people to say, ‘Wow, this is a lot of fun. I’m enjoying it, but I’m not getting past Generation 12.’ And then for people to say, ‘Hmm, let’s think about it as an evolutionary game. Now I can adapt the strategies like trying to domesticate them that allow you to get to Generation 20.’ And so in the process of learning to play better they’re also learning about evolution. They’re going to have to learn about the mechanics of the game, which means learning about evolution.

In nature, there tends to be an evolutionary arms race between predator and prey, but that doesn’t seem to be the approach here. Why not?

Barrie: We do kind of have an arms race in the game because the players can upgrade. We’re very careful in our game to specify the fact that the player does not evolve. Evolution is a population phenomenon not an individual one. The player upgrades. And so we don’t have any co-evolutionary dynamics in there — which would be a totally separate and awesome and totally fun game, I think.

The best biological approximation of what happens in our game is parasites and pathogens that attack long-lived species. And that results in the Red Queen hypothesis in evolutionary biology where the parasite and pathogens — their generation time is so fast — they’re evolving to adapt to a host that stays the same. And that’s really what we’re mimicking here.

Your game includes a function where gamers can gather the data generated by their play. Do you expect some them to start conducting their own little experiments in evolution from this?

Barrie: I think that would be a great result. I would be super pleased with that if our website had a wiki of some sort where there was a community of people who were like “Hey, I ran it under these conditions; I’m adding this to the data set — this is what happened.” If they’re doing those kinds of things, as educators that’s a huge win. In fact that’s wonderful.

Do you think the wider gaming industry will be putting evolution into video games any time soon?

Terry: The short answer is: Absolutely. I would love to see games like World of Warcraft where things are actually evolving as people are playing, and there’s no reason why that couldn’t happen.

Barrie: I think there’s tons of scope for these kinds of things to be implemented in the industry as a whole. I mean if you look at No Man’s Sky — they have the procedurally generated creatures. There’s lots of starting points there where you could take that but turn it into an evolutionary concept where the creatures are adapting to their particular planet or — I mean how cool would it be if something like Skyrim had a functioning ecology and an evolutionary component? The developers don’t even know what’s going to happen to you in that instance. I think that’s pretty exciting.

At some point in the future, do you imagine there will be simulated worlds that people can enter and spend time in, that will behave functionally much like a real, evolving ecosystem?

Barrie: I think unquestionably yes. I think that’s going to happen in some capacity. But I don’t know how fast it will happen. But yes, that’s my prediction.

Terry: Yep.

This is a bit tongue-in-cheek, but do you think there’s any chance we already live in a simulation — in some intelligent being’s evolving video game?

Barrie: No. [laughs] I’m a very practical person. I cannot prove that my assertion is true. But I assert it nonetheless.

Terry: My short answer is no, but from a more philosophical question you have to ask, “when does a simulation stop being a simulation?” In a sense, we’re not simulating evolution. What’s taking place in the game is real evolution.

Barrie: It’s just in silico evolution. So maybe we’re just big, biological computers. Yep. There you go. I don’t know. This got weird.

This interview has been edited for brevity and clarity.