[Thoughts on Romance in System-Driven Games](/blog/articles/romance)

A couple of days ago, I had an interesting chat with Alexis Kennedy about a lot of things. One of the topics that we went over was the idea of a Cultist Simulator spin-off about exploring a new relationship. An idea core to Cultist Simulator is that of the system’s opacity reflecting the experience of unknown knowledge. It seemed to me as though replacing the unknown knowledge of demonic rituals with that of learnings about another person would work well. Additionally, this works well with the text-heavy structure that a Cultist Simulator spin-off would demand. Also, I absolutely love the idea of replacing the eldritch lore in Cultist Simulator with learnings about the person you are in a relationship with.

The question of a system-driven game built around romance made for an interesting discussion and so I wanted to put down a couple of points about it here.

Common Approaches and Their Failings

Vending Machines

The idea of romance as a vending machine is one that’s both deeply toxic and unfortunately common. This is what you see in a game like Stardew Valley where you give gifts to the person you are romantically interested in and increase the number that represents your relationship with that person. Eventually, the relationship becomes strong enough for marriage or sex or whatever else the game chooses for a win-state in this mini-game. This idea of just buying affection is obviously a toxic statement and dehumanizing as well.

For a game like Stardew Valley though, the issue is that the verbs and even the nouns make this kind of approach to representing relationships a natural choice. The game is about material goods and converting them into other material goods and sometimes, incidentally, cash. The vast majority of the game revolves around this in one way or another. It only makes sense that the game attempts to handle relationships in a similar manner.

From this, you can see how the same structure could feel very different if done with a Cultist Simulator spin-off or with a story in The Quiet Sleep. The use of material goods as gifts in Stardew Valley feels ugly, but giving a person something like reassurance or kindness or empathy has a very different feel to it. Additionally, the fact that it is just a sub-theme in Stardew Valley adds to the banality, but making it the focus of a spin-off gives the represention of the relationship the space that it needs if it is to be elevated.

Dialogue Trees

Dialogue trees are another common approach to romance in games. This is not quite germane to this article as dialogue trees are not what I consider to be system-driven (although my exact thoughts on that are not quite so rigid, but that’s a topic for a different article). However, it’s still worth a quick look at.

These are a common approach for games in which most of the narrative autonomy comes from dialogue trees, such as RPGs or games which are almost entirely dialogue trees, such as IF or Telltale games. Again, the choice of structure is logical, if a dialogue tree is the standard tool for similar problems in this game, it makes sense to use it for romance as well.

However, there are a couple of issues with common implementations of this. The first is the idea that if you just say the right thing at the right time, you will end up with the person. This is far too close to a pick-up artist ethos for comfort and is not even particularly accurate. The other issue is that due to the natural opacity of wholly designed decision points, the difficulty of these paths tends to be kept very low. This results in the feeling that you’re guaranteed to get the person desired, which is also a dangerous statement to make. Again, both of these concerns are implementation dependent, but the context of most games is conducive to these issues.

Examining Core Issues and Their Solutions

Breadth and Depth of Mechanics in Video Games

Key to both of the above approaches is that systems that aren’t explicitly designed to represent romance are expanded and retrofitted until they do. A fact of video game design is that it is necessary to stretch mechanics to cover a range of situations simply because it is substantially cheaper than building a new mechanic for each case. This is especially true for sub-themes as it is easy to deprioritize them in favor of major parts of the game.

This would not be an issue for our hypothetical spin-off game as it would be built around the romance and so those parts of the game naturally become the highest priority work.

Non-Player Autonomy

The other, more important, concern common to the approaches above is a lack of autonomy for the romantic interest. When the game makes the other person into a subject to be acted upon, it is inevitable that the romance will be at best shallow and at worst disturbing. This is in fundamental conflict with the fact that a single-player game (which is all that we are considering here) can only offer true autonomy to the player.

This is where the structure of Cultist Simulator works well however. It uses the system opacity to give you the feeling of unknown knowledge, which works well for the eldritch, but should also work well for relationships. Additionally, this helps prevent gaming the systems of the game, although Cultist Simulator itself is a little amenable to that.

Additionally, the text-heavy structure of Cultist Simulator fits well with romance as a whole and allows the designer to make the act of exploring the relationship seem more human.

Another approach to take is to clearly separate the protagonist from the player. Cultist Simulator does this well, first person or not, and The Quiet Sleep is also notable for this. Making the protagonist someone that you have less direct control over reduces the distance between the protagonist and the romantic interest in terms of autonomy.

Finally, the obvious thing to do is to give the person some amount of AI. I used this in the third story of The Quiet Sleep, but made sure that the character that the AI was controlling was meant to be oblivious, which justified the AI’s incompetence. Having an AI feel like a person is a non-trivial task.