"The Process"-Footnotes: An interview with Jake Solomon (Firaxis)

Recently Rainer tried to introduce the readers of Der Standard to the wonders of procedural generation with this article, but I had already plunged the further depths of this topic in last year's article for games publication GameStar. During my extensive research, I conducted a series of in-depth interviews with a few notable game designers on procedural generation in games today. In this interview, the third of a series of diminuitively nicknamed "footnotes", the AAA sector speaks up: I talked to Jake Solomon, lead designer on XCOM: Enemy Unknown.

Let’s start with something simple: Are there any games you particularly like that make prominent use of procedural generation?

[Laughs] Pretty much every game that I like makes use of procedural generation. I do think that procedural generation is the real power of games. There are some really great scripted, linear experiences in games and I certainly love those as well. But for me, the real sense of wonder that games can create – that’s what to me makes games really special –  does comes from procedural generation.

That’s probably not a surprise with me working at Firaxis. The Civ-series uses that extensively in turns of creating terrain – the random creation of the terrain creates completely different games every time you play. But there’s also something like Minecraft, which shows how incredibely powerful that can be. In Minecraft, there are two kind of things. There’s the creation side, but there’s also the procedural generation of the world and the environment, which makes it special every time you play. It’s actually hard for me to think about games without thinking about procedural generation on some level.

With XCOM, our gut reaction was  to do procedural levels first. It didn’t work out like that, but it’s still something we saw a lot of value in. What it comes down to is this: You want the experience to be unique to the player. As a player, let's say when yu are playing Survival in Minecraft... you just know that you are experiencing something that is unique and different, and you do get a sense of ownership over it. So it is really important to make that experience authentic. On the other hand, if you’re playing through something that is a little more linear, you can have an incredible, crafted experience – but you do know that it’s the exact same experience that lots of other people are having as well.

That aspect of ownership is interesting in light of something you said in another interview about your reasons for not using procedurally generated maps in XCOM. You said that there could be a serious problem with players losing soldiers they felt attached to for reasons that felt too random, and therefore unfair to them. Do you think it’s difficult to find a balance between providing that feeling of it being an unique experience and this unfortunate aspects of procedural generation?

Balance is an issue. I mean, you can make a sort of linear game that feels like an interactive movie, and that’s incredible.  Whereas with procedural generation, you can make something far beyond that. You can make a world. That said, the problem is the following: When you give up the control as a developer, you really run into problems. You certainly cannot balance the experience nearly as well. It’s a very big challenge whenever you use procedural generation in any kind of system, not just in terrain or environments. You have to be careful.

Let’s take Civilization as an example: There is procedurally generated terrain, but terrain factors heavily in how your civilization is going to grow, how an enemy is going to get resources that you don’t, and so on. That could feel incredible unfair if you get the impression that you have got a bad starting spot, without it being your fault. At the same time, it forces the player to have new experiences that they would normally not go through. Those are the sort of things people love telling stories about.

Or let’s take XCOM as an example: The classes of soldiers you get is procedural. You cannot choose what class a soldier will be. To some extent, that can create frustration, because you really want a Sniper, and you do not get one. On the flipside, somebody is going to be like: “Ok, now I have got four assault soldiers, what do I do?” And they find themselves using tactics that they would never have experienced if they had had control over the choice of class. You know, a lot of times, the player finds a successful strategy and uses it over and over again. So they quickly grow bored of the experience, precisely because they have got too much control.

It’s a sort of tug of war: You want the players to feel like they have enough agency not to get screwed by the game. But you still have to make it an organic experience where the player is facing new challenges, so he has to respond with new answers. That’s really what procedural generation gets you: An unending experience.

You know that story about that guy who has been playing Civilization for ten years? That’s the  sort of thing that procedural generation can get you. It’s always fresh, it’s never going to be the same experience. That is something that is incredibly powerful, and it’s something that is unique to games. Still, you have a responsibility as a developer to balance, as best as you can, the game on a systemic level to make sure that it doesn’t go completely too far one way or the other.

The indie sector uses procedural generation often in a way that makes a significant difference in the gameplay. Some of the developers I interviewed wondered why the AAA-sector is hesitating so much in using procedural generation in more meaningful, visible ways. What are the reasons for that, in your opinion? You already said that giving up control is an important aspect, but are there other reasons as well?

Handcrafted content until recently was and is always going to look better than procedurally generated content. Making realistic procedurally generated content is incredibly difficult. It’s not only about making something that looks realistic, it’s about making something that looks realistic and also magnifies the game experience. It’s hard enough to make procedurally generated terrain, but to throw on top of that art that also can support all those game rules – that’s an incredibly complicated challenge that involves huge technology, huge design investment, and huge art investment.

That being said, certain games like Minecraft have a unique look to them. That’s a huge benefit for them, because they don’t have to worry about not looking photorealistic – they don’t go for the most beautiful possible screenshot ever.

There’s another element to it as well: As somebody who has been in development for a long time, I have seen that developers have a very vested interest in handcrafting content. That’s what they get paid to do. Things are getting better in that regard, but in the late 90s and early 2000s, you had those incredible, beautiful screenshots. It was a huge selling point to have a game that looked realistic. So all of us developers grew up in that environment, where the cool thing was to have a game that looked amazing or looked like a movie.

People developped skills focused on making photorealistic content. Suddenly having a computer do it is scary for them.

A lot of people developed skills particularly focused on making photorealistic content. So it’s very scary to have those skills and have some designer come and say: “Well, we’re just going to have the computer do it.” You know that it’s not going to look as good, but if it does, it’s a little threatening: If you can write a system that can produce something that is not transparent for the player anymore, so that they don’t know whether it was made by a human or a computer… at the end of the day, as developers it's hard to thing the players don’t actually care about that. If you could make something that plays just as well and looked reasonable, the players don’t care how it was made.

But here is the problem: Content is going to be a problem going forward – it already is. We have got teams that are massive, and the team-sizes keep growing. Now we have those new consoles coming that have huge power and huge amounts of memory. That means that people expect not to have loading screens anymore and that everything streams at once. But it also means that people expect massive worlds. Look at something like Skyrim, for example. It’s a fantastic game, which – even though the terrain itself is not procedurally generated – lives from those procedural moments when you are encountering people in the wild, fighting monsters, and experiencing incredible, unexpected moments. But it takes a lot of people to make that kind of content, and I just don’t see that stopping any time soon.

At the same time, it’s not reasonable to assume that team sizes are going to continue to grow in such a way, and that we can support that growth. I think that procedural generation of content is absolutely part of the future for nearly all developers. I truly, truly believe that. That’s should not be something that is scary, it’s something that is hugely exciting. Once we find a way to really procedurally generate environment and content, monsters and whatever else there might be – I think that will be the next frontier for games.

We are living in interesting times. Unreal Engine 4 is a good example. Previous versions of that engine were really static. But the next version is going to be very procedural and allows a lot of dynamic content. That’s obviously a good sign for the industry.

So, to sum it up: there is a bit of a vested interest in developers to continue to make handcrafted content, but it’s much more efficient to create content procedurally than to handcraft everything. It’s just incredibly expensive to handcraft content, and as the cost rises, we as developers have to look at all options carefully.

Reading across forums, you get the impression that fans sometimes have the attitude that anything produced my man is intrinsically more valuable than something the computer comes up with – which seems to include, for them, procedurally generated or arranged content. What are your thoughts on that?

I think we have seen players respond in a way that shows us they are ok with procedural content. We cannot prodedurally generate design yet, thank god. But it is definitely going to be a big part of our future as an industry.

Actually, there is a guy, Michael Cook, working on a PhD project dealing with that very question: Is it possible to create an artificial intelligence that is capable of designing whole games?

I think I have seen that, actually. It is pretty fascinating,  but I think one day, we are going to look back and realize  that it was this guy who created the Terminator.

Well, it looks like we are still a long way from AIs creating really compelling games. Still, there are obviously more and more games relying on some sort of procedural generation.

This is true. Honestly, if I could get some more system design work out of a computer, I probably would use it. But design being so subtle and so related to psychology, hopefully, it won’t put us out of work. Again, the really important thing to remember about procedural generation is that it is what makes us special.

It’s funny, but when I was a kid and I played Ultima V, gaming was more in its infancy and I didn’t know any players. So I just played it on my own. There was this big world and I felt like I was the only person experiencing it. I remember this incredible feeling of interacting with this world. There were all these people, and all the different concerns, and I think that this sense of wonder is the sort of thing that only procedural generation can really bring you. Now that gaming is such a big medium, you can just go on the internet and find out about all those different games, and how to beat a level. Procedural generation is a remedy to that.

Procedural generation is a remedy to having mysteries spoilt by the internet.

If you look at Minecraft, people share new experiences, they post videos of things that have happened there. Even DayZ, which has no procedurally generated environments, but gives you a similar feeling of a unique experience through the interaction with other people. And people post stories and videos resulting from that online. I think that this sort of thing, allowing people to either create their own experiences or at least experience their own unique game experiences, is what makes us special as a medium.

Arcen Games' Chris Park said that so far, the AAA sector mostly relied on multiplayer to increase replay value – even in games in which it does not make too much sense. He thinks that procedural generation and improved AI might, in the long term, be better options. Would you agree?

There is no question about it. At Firaxis, that’s one of our core tenets – replayability. In turn-based strategy, some people like multiplayer. But I do not think that it’s our big hook in something like Civilzation or even XCOM. Still, they are incredibly replayable, for years. And that is 100% due to the procedural elements in the game. If you go on Steam right now and look at how many people are playing Civ V, it’s going to be something like 20-30’000 people. Civ V is over 2 years old at this point. But people still play the game because it has incredible replay value, thanks to the procedurally generated content in that game. That’s incredibly valuable. It keeps the game alive, it increases the attach rate of people who buy expansion packs and additional content. It also sells more games, because people keep talking about the game, and the game keeps coming up. And of course, it decreases used game sales.

It’s true – people put in multiplayer because the want the players to play the game longer. But a replayable singleplayer modus is just as important. Skyrim would be a good example of that, with events that are very procedural, in terms of what you can do and how thing react to you in the world. That’s a game that has sold massive amounts, even without multiplayer. People are still playing that, they have put in hundreds and hundreds of hours. So yeah, I think that Chris Park is absolutely right.
The biggest value of procedural generation, apart from the fact that players feel a sense of ownership over their experience, is the fact that from a development standpoint, it makes the game really replayable and adds a lot of value.

A lot of the guys I interviewed were certain that in the future, procedural generation will find its way into virtually every genre, in the indie- and the AAA-sector. Would you agree? What do you think does the future hold in store for procedural generation?

I think that the exciting future of procedural generation is not in the experiences that we give to the players, but rather that which the players create and share among themselves. The real future power of procedural generation is the ability of the player to almost craft game experiences of their own, Minecraft being an early forerunner of this.

And yeah, I do think that it’s hard to imagine that a genre won’t benefit from some sort of procedural content or gameplay in the future. As I said, it’s what helps us combat all the rising costs of development.  Well, I don’t think that strictly linear, narrative experiences are going to go away, because those can be incredibly powerful and are some of our best-selling and best games. There is always going to be a place for that. But I do think that the future is going to be procedural – if it’s not the environment, it is going to be procedural interaction between actors and the environment.

One thing I should say is that for me, one of the first really big procedural moments was when I was playing GTA. That might sound weird, but it’s the sort of game where the procedurality comes from the interactions between the people and city, and the things you are doing. It’s the sort of game where you do all sort of things just to see what the hell is going to happen, and you get your own crazy experiences that people then share online. Games like that will just continue to become bigger and bigger. Whether or not it is the actual environment that is procedural or it's the freedom in how you can interact with NPCs or different actors within the game worlds itself doesn’t matter. What is really fascinating about procedural generation is giving players the ability to shape the experience and generate content on their own that other players can share in, in whatever way.

Thank you for your time!

You can find the first interview in the series here, where I talked in length with Arcen Games' Chris Park. The second interview was with Jim Rossignol, game journalist and head designer at BIG ROBOT. Check back later for interviews with Pwnee Studios (Cloudberry Kingdom) and Michael Cook, who makes all other interviewees tremble in fear.