AI-Based Game Design Patterns

Treanor, Mike and Zook, Alexander and Eladhari, Mirjam P and Togelius,
Julian and Smith, Gillian and Cook, Michael and Thompson, Tommy and
Magerko, Brian and Levine, John and Smith, Adam (2015) AI-based game
design patterns. In: Proceedings of the 10th International Conference on
the Foundations of Digital Games 2015 (FDG 2015). Society for the
Advancement of Digital Games, Santa Cruz, CA. ISBN 9780991398249 ,
This version is available at https://strathprints.strath.ac.uk/57219/
Strathprints is designed to allow users to access the research output of the University of
Strathclyde. Unless otherwise explicitly stated on the manuscript, Copyright © and Moral Rights
for the papers on this site are retained by the individual authors and/or other copyright owners.
Please check the manuscript for details of any other licences that may have been applied. You
may not engage in further distribution of the material for any profitmaking activities or any
commercial gain. You may freely distribute both the url (https://strathprints.strath.ac.uk/) and the
content of this paper for research or private study, educational, or not-for-profit purposes without
prior permission or charge.
Any correspondence concerning this service should be sent to the Strathprints administrator:
[e-mail]
The Strathprints institutional repository (https://strathprints.strath.ac.uk) is a digital archive of University of Strathclyde research
outputs. It has been developed to disseminate open access research outputs, expose data about those outputs, and enable the
management and persistent access to Strathclyde's intellectual output.

AI-Based Game Design Patterns
Mike Treanor
American University
[e-mail]
Alexander Zook
Georgia Institute of Technology
[e-mail]
Mirjam P Eladhari
Otter Play
[e-mail]
Julian Togelius
NYU Polytechnic School of
Engineering
[e-mail]
Gillian Smith
Northeastern University
[e-mail]
Michael Cook
Goldsmiths College
[e-mail]
Tommy Thompson
University of Derby
[e-mail]
Brian Magerko
Georgia Institute of Technology
[e-mail]
John Levine
University of Strathclyde
[e-mail]
Adam Smith
University of Washington
[e-mail]
ABSTRACT
This paper proposes a model for designing games around
Artificial Intelligence (AI). AI-based games put AI in the
foreground of the player experience rather than in a supporting
role as is often the case in many commercial games. We analyze
the use of AI in a number of existing games and identify design
patterns for AI in games. We propose a generative ideation
technique to combine a design pattern with an AI technique or
capacity to make new AI-based games. Finally, we demonstrate
this technique through two examples of AI-based game prototypes
created using these patterns.
Categories and Subject Descriptors
I.2.1 [Artificial Intelligence] Applications and Expert Systems –
Games. K.8.0 [Personal Computing] General – Games.
General Terms
Design.
Keywords
Game design, Artificial Intelligence, Machine Learning
1. INTRODUCTION
Almost every game features some kind of Artificial intelligence
(AI). The most common role for AI in a game is controlling the
non-player characters (NPCs), usually adversaries to the player
character. Yet this opposing AI is often rudimentary because the
design of the game does not need more complex AI. The
perception of AI as controlling adversaries in turn results in games
designed to not need richer and more varied AI. However, there
are more roles AI can play. Using AI for controlling an
adversarial NPC is one of many design patterns for how AI can be
used in games. This paper proposes a model and ideation
technique for designing games around AI, or AI-based games. A
primary goal of this work is to aid discovering new types and
potential genres of games.
This paper focuses on AI that is foregrounded in the game, as
opposed to AI that operates in the background. We define
foreground AI as agents the player notices and can reason about.
For example, AI that controls a NPC the player either interacts
with or observes for sufficient time to learn its behavior is
considered foregrounded. Meanwhile, AI that supports gameplay
such that the specifics of its behavior is not relevant to the player
is considered background AI. An example of background AI is the
NPC car behavior in Grand Theft Auto V that enables the player to
quickly speed down the road without crashing too often. Other
examples include the “fairer” random number generator in
Civilization IV that skews probabilities in the player’s favor, or
the NPC pathfinding systems in first person shooters. While such
background AI systems are important to gameplay and smooth the
player experience, their operation is not intended to be evident to
the player. This paper strives to advance the idea that putting AI
in the foreground can enable new types of gameplay experiences.
Accepting a broad definition of AI, games based on simulations of
physics can be considered AI-based games. For example, in Super
Mario Bros, the player must reason about how the system is going
to place the character in 2D space based on their input. While the
player may not know all of the specifics of how the game
simulates 2D physics, they build an approximate model and are
able to apply this model to predict how their input will affect the
game state while pursuing intentional acts. The physics simulation
is central to the gameplay experience.

Super Mario Bros makes use of what have been called “graphical
logics” because the player’s understanding of its physics
simulation is achieved via visually represented entities moving
and interacting on a screen [14]. This can be considered
foreground AI, as the player’s understanding of the system is
central to how they make choices. This idea of putting a
visualization of the physics simulation central to a game suggests
that visualizations of other AI systems might make for interesting
games. For example, what types of games could be made when a
social simulation, or a learning algorithm are visualized and made
central to gameplay? We present several such design patterns for
how foreground AI can be used to make new types of games.
As part of this effort, we analyze how AI is used in several
existing games and identify design patterns for AI in games. We
propose a generative ideation technique to combine a design
pattern with an AI technique or capacity to make new AI-based
games. Finally, we demonstrate this technique through two
examples of AI-based game prototypes created using these
patterns.
2. RELATED WORK
Our argument for the value of foregrounding AI in game and our
design pattern-based taxonomy builds upon prior research in
design patterns, theoretical frameworks for understanding games,
and analysis of existing AI-based games.
A design pattern approach to describing games and game content
allows us to build a common vocabulary for discussing games,
identify common elements between games at the mechanical and
player levels, and reason about the structure of games [2, 9, 17,
19]. Design patterns are typically descriptive and informal, drawn
from a close analysis of multiple source games. The patterns
themselves typically have a short name, a description of how the
pattern is abstracted across games, and several motivating
examples to show the capacity of the pattern to describe a variety
of scenarios across multiple games. Our pattern taxonomy follows
the same model: the games we analyzed to extract our patterns
come from diverse developers, including large industry studios,
academic research, and independent development.
Though the primary purpose of design patterns is typically to
provide an analytical lens, they also have the potential to be used
generatively. Hullett and Whitehead’s [9] FPS level patterns were
evaluated via the deliberate design of levels that incorporate those
patterns. Dahlskog and Togelius’s [4] pattern-based platformer
level generator takes this one step further, formalizing the patterns
to the extent that a computer can perform the pattern-based
design. We pose that the patterns we have identified can be used
generatively during the ideation phase of design, to allow us to
consider new kinds of playable experiences.
Our taxonomy also builds on previous work in understanding the
role of AI in games and the potential it holds for the future of
games. Mateas [13] calls for the creation of “expressive AI”:
playable experiences with complex underlying AI systems where
all interaction is framed by the player needing to read meaning
into the AI’s actions. Eladhari et al. [5] describe a process for
designing games where the AI system is an integral part of the
game’s design. They distill a common process followed during the
design of four games: the Pataphysic Institute [6], Prom Week
[15], Mismanor [18], and Endless Web [16]. With Endless Web,
Smith et al. pose that an AI-based game is one where the
mechanics, dynamics, and aesthetics of the game [10] are deeply
linked to the AI system. More recent games designed around their
AI system include Horswill’s MKULTRA [8] and Cook’s A Rogue
Dream [3].
3. DESIGN PATTERNS
Below we discuss several design patterns for AI-based games.
These patterns illustrate ways to develop a game mechanic
starting from an AI technique (e.g., AI is Visualized) or starting
from an intended experience that requires AI (e.g., AI as Role-
model). The design patterns and example games are meant to be a
tool for thinking about creating AI-based games, rather than serve
as a comprehensive taxonomy of methods. Note also that multiple
techniques may apply to a single game: Table 1 provides an
overview of these patterns and game examples.
3.1 AI is Visualized
Pattern: Provide a visual representation of the underlying AI
state, making gameplay revolve around explicit manipulation of
the AI state.
Explanation: Many AI techniques revolve around an estimation
of the value of actions or game states. Typically these values are
hidden from players to promote the sense that an opposing AI
agent possesses an intelligence motivating its actions. Visualizing
the state of a system or agent enables gameplay as the system is
now exposed as a potential obstacle to player progress.
Example: Third Eye Crime [11] is a stealth game that illustrates
this pattern by visualizing the guard AI position tracking and
estimation system. Gameplay involves avoiding guards or
throwing distractions to manipulate the guards’ predictions of
player location. The direct visualization of AI state allows a
designer to build a game around manipulating, understanding, and
mentally modeling how the AI state changes.
3.2 AI as Role-model
Pattern: Provide one or more AI agents for the player to behave
similarly to.
Explanation: AI techniques to date often demonstrate strongly
patterned behavior that players come to predict: e.g., finite state
machines (FSMs) follow fixed routines that can often be easily
noticed. Rather than attempt to make agent behavior more
unpredictable, this pattern leverages the behavioral rigidity of a
technique to set a stage for the player to act on. Gameplay in this
pattern involves acting to mimic the behaviors of AI agents,
leading to an “imitation game” judged by an in-game system or
opposing players.
Example: Spy Party is a game where one player is a spy at a party
populated by FSM agents and the opposing player is a sniper
watching the party with a single shot to kill the spy. Gameplay for
the spy centers on the player attempting to act similarly to the
party agents while discreetly performing tasks in the environment
like planting a bug or reading a code from a book. Gameplay for
the sniper focuses on discerning the human player from AI agents
by looking for behavioral cues that differentiate the two. An
imitation game thus forces players to explicitly reason about the
processes followed by an AI technique.
3.3 AI as Trainee
Pattern: Have player actions train an AI agent to perform tasks
central to gameplay.
Explanation: Machine learning techniques revolve around
learning new behaviors using examples. By using player actions
as a source of examples an AI agent can learn to perform tasks,