Systematic Thinking

In a nutshell

After Ludum Dare I realized that there is a certain kind of game I enjoy the most: The semi-automated ones. Since then I am thinking about new game ideas that would work based on the same principles. To be precise, because I don't know whether this kind of game has a name at all, and as such I also have no idea whether someone already categorized them, my focus was laying down some principles. 

This post gives a rough overview of what I came up with. In the close future I will work on games which illustrate the main ideas. If I have something, I'll post them here. Don't expect anything really fancy, just some simple web based stuff.

Definition

Semi-automated games are those where the actors are controlled by extensive, environment aware AIs, and the player's task is to indirectly influence these actors towards a given goal. 

This might not be the most precise definition that includes every possible similar games and excludes all of those which are not semi automated, but this definition is easy to understand.

There are several notions in this sentence which can/should be further expanded: the actors, environment, influence and goal. There is also an implicit one, which needs to be mentioned for further reference: rules of acting. I will explain these things below.

Actors and Environment:

Actors are those objects in the environment which can interact with other environmental objects, including other actors. Often, this is very simple: A human or animal is an actor, while a chair or a rock is an environmental object. But there are some less obvious examples: In a space game, a planet with gravity might be a non-actor (it's not actively interacting with anything) but on the other hand it can totally change the way the player have to interact with the environment, and may enforce the player to directly act against the planet. The fine line of distinction is probably drawn at the context awareness: if the object interacts with others because it is aware of its local or global context, then it is an actor. Otherwise, it's just an environmental object.
The number of actors is a very interesting factor: Having 0, 1 or more actors can fundamentally change the whole game. Semi-automated games might actually come in all of these 3 flavours.

Rules:

This one is tricky. It's possible that what I want to compress into this one definition should actually be separated into two. On the other hand, these two things are in a sense very similar.
One of them is the environmental trigger for actions. If you want to hammer a nail in a piece of wood, you will need a hammer, a nail, and a piece of wood. This is not a trigger, it's context. On the other hand, if it is too hot, you will take some clothes off. This one is an environmental trigger.
The other one is the behavioural configuration of the AI: Certain actions will only happen if the intenseness of some external trigger is above a certain, pre-configured level. Say, you will start removing your clothes above 30 degrees Celsius, while your neighbour can take the heat up until 50 degrees. (Some guy!) The same might be true for some internal state: you will want to eat only when you are hungry, but the required level of "hungriness" might be different for each individual.
All of these are enabling factors that will be taken into consideration before acting. But it's even more complex than making a decision: environmental objects might have triggers that will change their state without explicit actions, for example ice will melt if the temperature is above 0 Celsius.
Changeable rules are a very common feature in semi-automated games.

Influence:

This is the biggest and most complex of the above mentioned notions. This is what the player "does" when playing, and obviously this can be a lot of different things.
The first question is what the player can interact on/with: It might be actors, environmental objects or rules, or any combination of these three.
Let's say that the player can control actors: it can tell an actor to cut down a tree, for example. This is a pure Actor Interaction, because the player do not directly interact with the tree - it is only felled if the actor is capable of executing the order.
For a pure Environmental Object Interaction, a good example is when the player is moving around objects in a game level so that some autonomic actors can finish some actions. In this case the player can not force the actor to do something, or prevent it from doing - he or she is only able to control the environments, which indirectly enables or disables the action.
Pure Rules Interaction is where the player can not affect nor the actors, nor the objects, but it can change some properties which will have an effect on how the actors behave, for example the player can raise or lower the current temperature, which will force the actors to take of their clothes, or will melt the ice, etc.

Scope of interaction is related to what with interact on: Is it only one actor, or a few actors, or all of them? It is only a certain type of objects, or some different kinds, or any object on the map? Just one rule, or some, or all? (Nota Bene: Defining all rules is not an easy thing to do.) Obviously, this applies to everything that we can control in the game, and there are a lot of different possible combinations.

When we know exactly what we can control, we can also talk about the extent of interactions. If we can control, say, one actor, what can we do with it? Everything, or just some things? For example, in case of a combatant, we might be able to force him to attack an enemy, but we might have no control over when he will flee. Or, in case of environmental objects, can we control all, or some, or even just one of location, speed, orientation, weight, colour, etc? And obviously, the extent of control over rules is trivial: can it be any value, or just one of a set of values?

Frequency of interactions might not seem to be the most important thing, but from the perspective of gameplay it is one of the main factors that the designer needs to carefully balance. In the case of semi-automated games, the frequency is closely related to the stability of the system.
One extreme example is when the players only does a setup (any or all of actors, environment and rules) before he/she starts playing, and then all he/she needs to do is to watch what happens. We could call this a simulation instead of a game, although it may indeed be a game if the goal is to get to a certain winning state. This is about the lowest interaction that is possible, except maybe for running simulations in randomly generated environments, where all you do is a mouse click.
The other extreme is when the stable states of the system are losing conditions, so that the player has to constantly keep the system in a desirable non-stable state. This is a total interaction game, especially when there are a lot of factors that you need to constantly watch and adjust.

Any of the above can be used when designing a semi-automated game.

Goals:

This is what makes the game a game. I won't expand this very far, because this can be anything. In case of semi-automated games, goals are often related to something that the player can not directly control, thus forcing him/her to understand the inherent behaviour of the system.

The nature of semi-automated games 

Multiplayer:

It is equally easy to create solo, cooperative or competitive semi-automated games. There is only one important thing that the designer needs to realize: solo and cooperative games have either only one or several simultaneously achievable goals, while competitive games have to have several goals (say, one for each player) that are mutually exclusive. This is somewhat misleading, though: You can say that the goal of each player is "to kill all the other players". The wording is the same, so you could argue that all of them has the same goal. But formally that is a different goal for every one of them: if we have player A, B and C, then the goal for A is to kill B and C, the goal for B is to kill A and C, and so on.
Multiplayer can also mean that the players have control over different things, say, one of them can influence the actors, while the other one can change the rules. Obviously, this will be a very asymmetric game, but that doesn't invalidate the idea.

Avatar:

In some games, the player is not part of the world, only an outsider who can affect it, either as a non material common mind (pure actor games) or as a god (rules, environment, possibly actors). On the other hand, having an avatar that represents the player in the world is also an option.
The presence of the avatar might change the game in several ways:

• The player may feel more involved in the game
• A mortal avatar might add extra complications to the game (the player may need to nurture it, defend it, teach it, etc.)
• An avatar might place some restrictions on the influence the player have over the world or other actors (actions might only happen in the presence of the avatar, etc)

Example Games:

The Guild:

In this game, the player's avatar is a craftsman in a European country in the middle ages, and the goal is to get rich and to have a political career. The player can directly control one actor, representing him/her, and can "indirectly" control other actors, by giving them money (wages) and telling them what to work on - which is basically influencing the rules. Also, by allowing the player to buy and sell items, build a house and workshop, there is a limited control over the environment.
The reason why this game is a good example for semi-automated games is that the actors are actually part of the world, the player can watch the opponents, the workers, city guards, etc. moving around and having an actual life. Also, the opponents are bound by the same rules as the player, they have to send their carts to the market, and buy materials to be able to create things in their workshops.

Majesty:

Majesty is very similar to a normal RPG/RTS game, like Warcraft, except for the fact that the player doesn't have direct control over the fighting units (heroes). The player doesn't have an avatar, but he/she is able to build buildings, research technologies and "summon" heroes. Then, to motivate them, the player can mark places with flags, offering a reward to any hero that explores the place and eliminates the local wildlife/inhabitants. Then, the reward is used for buying better equipment, and the heroes get some experience to get better - all of this is happening without any player interaction whatsoever.

Ecosphere:

I don't know the actual title of this game, except for that it was working like a simplified ecosystem. The player interaction is actually quite minimal: they create animals with a simple system, and then a certain number of these animals is added to the autonomic game world, and from that point on the player can only watch the statistics. It is very interesting to watch how the ecosystem changes continually until there is a temporary balance.

E-volution:

This is my Ludum Dare entry. Probably not the most interesting game around, but it is a good representation of a simple pure Rules Interaction game. In Challenge Mode a random rule is generated for one of the actors, and you need to change the rules for the other actor to match that of the first one.

Gaic:

Google AI challenges are taking this idea a bit further. The challenge is usually the following: For a simple game, you need to write an AI, which is then tested against the other players' code. It's kind of like a very complex version of a pure Rules Interaction game. Participating in this challenge is good fun, but also, it's very educating for a future game designer / AI developer.