ThesisPDF Available

The Non-Player Character: Exploring the believability of NPC presentation and behavior


Abstract and Figures

Over the last few decades there has been immense growth in the video game industry, and we have seen great improvements in both graphics and audio. Unfortunately, the development of artificial intelligence (AI) and non-player characters (NPCs) has not proceeded at the same pace. Although there have undoubtedly been improvements, the field as a whole has lagged behind its siblings. Many of the problems with NPCs stem from the fact that they do not achieve a sufficient level of believability, particularly in the social arena. This is primarily related to the fact that the NPCs do not behave in ways that align with the expectations of the player. This can lead to the player misunderstanding the role and purpose of the NPC, which damages the believability of the game. By extension, this lessens the enjoyment the player can derive from the game. Hence, it is imperative that the design of the NPC be in line with player expectations. This thesis takes a holistic view of NPCs, encompassing their design, evaluation, and player perceptions. It uses a design science methodology, and primarily uses qualitative and interpretative methods. It will provide a description of the various types of NPCs found in games, what their design elements are, and how they are interpreted by players.
Content may be subject to copyright.
The Non-Player Character Exploring the believability of NPC presen-
tation and behavior
Henrik Warpefelt
DSV Report Series No. 16-003
The Non-Player Character
Exploring the believability of NPC presentation and behavior
Henrik Warpefelt
Henrik Warpefelt, Stockholm University 2016
ISBN 978-91-7649-379-3
ISSN 1101-8526
DSV Report Series No. 16-003
Printed in Sweden by Publit, 2016
Distributor: Department of Computer and Systems Sciences, Stockholm University
Cover image by Aware Mustafa
Over the last few decades there has been immense growth in the video game
industry, and we have seen great improvements in both graphics and audio.
Unfortunately, the development of artificial intelligence (AI) and non-player
characters (NPCs) has not proceeded at the same pace. Although there have
undoubtedly been improvements, the field as a whole has lagged behind its
Many of the problems with NPCs stem from the fact that they do not
achieve a sufficient level of believability, particularly in the social arena. This
is primarily related to the fact that the NPCs do not behave in ways that align
with the expectations of the player. This can lead to the player misunderstand-
ing the role and purpose of the NPC, which damages the believability of the
game. By extension, this lessens the enjoyment the player can derive from the
game. Hence, it is imperative that the design of the NPC be in line with player
This thesis takes a holistic view of NPCs, encompassing their design, eval-
uation, and player perceptions. It uses a design science methodology, and
primarily uses qualitative and interpretative methods. It will provide a descrip-
tion of the various types of NPCs found in games, what their design elements
are, and how they are interpreted by players.
Sammanfattning / Swedish abstract
Spelindustrin har växt enormt under det senaste årtiondet, och vi har sett stora
förbättringar vad gäller både grafik och ljud. Tyvärr har inte den artificiella
intelligensen i spel utvecklats i samma takt, i synnerhet vad gäller utvecklingen
av icke-spelarkaraktärer (eng. Non-Player Characters (NPCs), NPC:er). Det
har oförnekligen skett förbättringar, men dessa har inte varit av riktigt samma
magnitud som sina syskonområden.
Många av problemen med NPC:er härstammar från det faktum att de inte
uppnår en tillräckligt hög nivå av trovärdighet, främst inom den sociala sfä-
ren. Det här beror framförallt att NPC:er inte beter sig sätt som ligger i
linje med spelarens förväntningar, något som leder till att spelaren missförstår
NPC:ernas roller och deras syfte i spelet. Detta kan tyvärr innebära att spelaren
uppfattar spelet som mindre trovärdigt, som i sin tur leder till att spelupplevel-
sen blir mindre intressant och underhållande. Det där därför av yttersta vikt
att NPC:er designas ett sådant sätt att deras beteende, utseende och andra
attribut ligger i linje med spelarnas förväntningar.
Den här avhandlingen tar ett helhetsgrepp NPC:er, och omfattar bå-
de design och utvärdering, väl som vilka förväntningar som spelarna har
NPC:er. Avhandlingensarbetet har främst gjorts med en design science-
metodologi, och använder primärt kvalitativa och interpretiva metoder. Av-
handlingen tillhandahåller en beskrivningar ett antal olika typer av NPC:er
som finns i spel, vilka designelement som är viktiga för dessa typer, samt hur
typerna tolkas av spelarna.
No work is truly done alone and although I had help and support from many
people (for which I am eternally grateful) some deserve special mention.
Firstly, my mother, father, and sister for their undying support and love.
Thank you for believing in me, and for supporting me through this sometimes
difficult process.
My main supervisor, Harko Verhagen, deserves no end of thanks for his
contributions to this work, his guidance, and for his support. Without him this
thesis would definitely not have happened.
My assistant supervisor Lars Asker deserves thanks for prompting me to
create one of the concepts that would become central to my understanding of
non-player characters, namely characterhood.
Björn Strååt and Magnus Johansson for being my coauthors and general
partners in crime, as well as purveyors of sage advice.
Mirjam Eladhari, Georgios Yannakakis, and the rest of the research group
at the Institute of Digital Games at the University of Malta. Thank you for
asking the right questions and for providing me with the right advice at the
right time.
Aware Mustafa, for always providing batch after batch of wonderful graph-
ics, posters, and art direction on all too short notice.
No one stays sane without friends. Thank you all for graciously suffering
my withdrawal from social life.
This thesis is dedicated to my mother, father, and
sister. Thank you for your love and support.
List of Papers
The following papers, referred to in the text by their Roman numerals, are
included in this thesis.
PAPER I: Analyzing the social dynamics of non-player characters*
Johansson, M., Strååt, B., Warpefelt, H., and Verhagen, H. (2014). Fron-
tiers in Gaming Simulation. Eds: Sebastiaan A. Meijer and Riitta Smeds.
PAPER II: Analyzing the believability of game character behavior using the
Game Agent Matrix*
Warpefelt, H., Johansson, M., and Verhagen, H. (2013). Proceedings of
the Sixth Bi-Annual Conference of the Digital Games Research Associ-
ation: Defragging Game Studies (DiGRA 2013).
PAPER III: Cues and insinuations: Indicating affordances of non-player char-
acters using visual indicators
Warpefelt, H. (2015). Proceedings of DiGRA 2015: Diversity of Play:
PAPER IV: Towards an updated typology of non-player character roles
Warpefelt, H., and Verhagen, H. (2015). Proceedings of the 8th Inter-
national Conference on Game and Entertainment Technologies (GET
PAPER V: A typology of non-player characters
Warpefelt, H., and Verhagen, H. (2016). First Joint International Con-
ference of DiGRA and FDG. Submitted.
PAPER VI: A model of non-player character believability
Warpefelt, H., and Verhagen, H. (2016). Journal of Gaming and Virtual
Worlds. Submitted.
Reprints were made with permission from the copyright holders.
Papers marked with an asterisk were previously part of my licentiate thesis
Mind the Gap (Warpefelt, 2013)
Other publications by author
A method for comparing NPC social ability*
Warpefelt, H., and Strååt, B. (2012). Proceedings of the 5th Annual Interna-
tional Conference on Computer Games and Allied Technology (CGAT 2012).
Breaking immersion by creating social unbelievabilty*
Warpefelt, H., and Strååt, B. (2013). AISB 2013 Convention. Social Coordi-
nation: Principles, Artifacts and Theories (SOCIAL.PATH).
Anti-heuristics for maintaining immersion through believable non-player
characters [Poster]
Warpefelt, H., and Strååt, B. (2013). Proceedings of the 8th International
Conference on the Foundations of Digital Games (FDG 2013).
Evaluating Game Heuristics for Measuring Player Experience
Strååt, B., Johansson, M., and Warpefelt, H. (2013). Proceedings of GAME-
ON’2013. The 14th International Conference on Intelligent Games and Simu-
Analyzing AI in NPCs: An Analysis of Twelve Games
Johansson, M., Strååt, B., Warpefelt, H., and Verhagen, H. (2014). Multiplayer—
The Social Aspects of Digital Gaming. Eds: Thorsten Quandt and Sonja
Kröger. Routledge.
Applying the Two-Factor-Theory to the PLAY Heuristics
Strååt, B., and Warpefelt, H. (2015). Proceedings of DiGRA 2015: Diversity
of play: Games—Cultures—Identities.
Papers marked with an asterisk were previously part of my licentiate thesis
Mind the Gap (Warpefelt, 2013)
Author’s contribution
The contributions of this thesis are a theoretical framework by which we un-
derstand how NPCs achieve believability and immersion, a typology of NPCs
that describes the impact of NPCs on both the gameplay and the narrative of
the game, as well as the design restrictions and demands on these NPCs. Fur-
thermore, this thesis contributes a models of how NPCs affect believability.
The model provides an overview of how behaviors correlate to different levels
of complexity of NPCs as Artificial Intelligence (AI) constructs, and a descrip-
tion of how these agent types influence the game experience both as a designed
and as an emergent experience.
My contributions to the papers that make up the basis for this thesis have
varied. The first two papers (Papers I–II) were produced in a highly collabo-
rative environment, and as such it is difficult to discern exactly which of the
authors contributed what. For Paper I, the data collection was performed by
three out of the four authors, and the analysis by all four. My contribution to
that paper should thus be roughly 25%. As for Paper II, the data collection
was performed by two of the three authors, and the analysis by all three. My
contribution to the paper should therefore be roughly 35%.
For Paper III I was the sole author, although I was advised on the study
design by my main supervisor. My contribution is 100%.
Papers IV–VI were performed in cooperation with my main supervisor,
who actively contributed to the study design. The data collection and analysis
was performed by me. Overall, my contribution to these papers should be on
average 90%.
It should be noted that these percentages do not include time spent writ-
ing the different papers, but rather only the time spent performing the actual
Abstract v
Sammanfattning / Swedish abstract vii
Acknowledgements ix
List of Papers xi
Other publications by author xiii
Author’s contribution xv
Acronyms xxi
List of Figures xxiii
List of Tables xxv
1 Introduction 27
1.1 Purpose, goals and research questions . . . . . . . . . . . . . 27
1.2 Overview of the method . . . . . . . . . . . . . . . . . . . . 28
1.3 Results and contributions . . . . . . . . . . . . . . . . . . . . 29
1.4 Reading this thesis . . . . . . . . . . . . . . . . . . . . . . . 29
2 Non-Player characters 31
2.1 What are non-player characters? . . . . . . . . . . . . . . . . 31
2.2 The concept of characterhood . . . . . . . . . . . . . . . . . . 33
2.3 Classifying of non-player characters . . . . . . . . . . . . . . 36
2.3.1 Typologies of narrative characters . . . . . . . . . . . 36
2.3.2 Ludic support character typologies . . . . . . . . . . . 37
2.3.3 Combining typologies . . . . . . . . . . . . . . . . . 39
3 Theoretical framework 41
3.1 Narrative ............................ 41
3.2 Indexical storytelling . . . . . . . . . . . . . . . . . . . . . . 41
3.3 Immersion............................ 42
3.4 NPCs and the specific forms of immersion . . . . . . . . . . . 44
3.4.1 Imaginative and narrative immersion . . . . . . . . . . 44
3.4.2 Challenge-based immersion . . . . . . . . . . . . . . 45
3.4.3 Sensory-based immersion . . . . . . . . . . . . . . . 46
3.4.4 Social immersion . . . . . . . . . . . . . . . . . . . . 47
3.5 Believability........................... 47
3.5.1 Affordances and believability . . . . . . . . . . . . . 49
3.6 Concluding summary . . . . . . . . . . . . . . . . . . . . . . 51
4 Method 53
4.1 Researchstrategy........................ 53
4.2 Interpretive research . . . . . . . . . . . . . . . . . . . . . . 54
4.3 Evaluatinggames........................ 55
4.4 Studies performed for this thesis . . . . . . . . . . . . . . . . 57
4.4.1 NPCs and believability (S1) . . . . . . . . . . . . . . 57
4.4.2 Types of NPCs (S2) . . . . . . . . . . . . . . . . . . 62
4.4.3 Design elements of NPCs (S3) . . . . . . . . . . . . . 64
4.4.4 Delimitations . . . . . . . . . . . . . . . . . . . . . . 65
4.5 Games included in the studies . . . . . . . . . . . . . . . . . 66
4.6 Research questions answered . . . . . . . . . . . . . . . . . . 66
4.7 Ethical considerations . . . . . . . . . . . . . . . . . . . . . . 68
5 Non-player characters and believability 71
5.1 The importance and mechanisms of the believability of an NPC 71
5.1.1 Findings regarding the mechanisms of believability . . 72
5.2 NPCs and affordances . . . . . . . . . . . . . . . . . . . . . . 73
5.3 Believable behavior . . . . . . . . . . . . . . . . . . . . . . . 73
5.4 Findings regarding NPC believability and behavior . . . . . . 75
5.4.1 Behaviors identified as problematic for believability . 75
5.4.2 Conclusions about NPC behavior and its impact on be-
lievability........................ 79
6 A typology of non-player characters 81
6.1 Thetypology .......................... 81
6.2 Typedescriptions ........................ 82
6.2.1 Description format . . . . . . . . . . . . . . . . . . . 82
6.3 The basic type of all NPCs . . . . . . . . . . . . . . . . . . . 83
6.4 Functions ............................ 83
6.4.1 Vendor ......................... 84
6.4.2 Services......................... 85
6.4.3 Questgiver ....................... 86
6.5 Adversaries ........................... 87
6.5.1 Enemy ......................... 87
6.5.2 Opponent........................ 89
6.6 Friends ............................. 90
6.6.1 Sidekick ........................ 90
6.6.2 Ally........................... 91
6.6.3 Companion....................... 92
6.6.4 Pet ........................... 93
6.6.5 Minion ......................... 93
6.7 Providers ............................ 94
6.7.1 Storyteller........................ 94
6.7.2 Lootprovider...................... 95
6.8 Discussion............................ 96
7 A model of non-player characters 99
7.1 Structure............................. 99
7.1.1 Embedded and emergent behavior . . . . . . . . . . . 100
7.2 Cell descriptions . . . . . . . . . . . . . . . . . . . . . . . . 100
7.2.1 Act ........................... 101
7.2.2 React .......................... 102
7.2.3 Interact ......................... 103
7.3 Discussion............................ 104
8 Conclusions 105
8.1 Research questions . . . . . . . . . . . . . . . . . . . . . . . 105
8.1.1 What types of non-player characters exist within games?
(Q1)........................... 105
8.1.2 What are the design elements of different non-player
character types? (Q2) . . . . . . . . . . . . . . . . . . 105
8.1.3 In what ways do players perceive and determine the
type or types of a non-player character? (Q3) . . . . . 106
8.1.4 What makes a non-player character believable? (Q0) . 106
8.2 Theory versus practice . . . . . . . . . . . . . . . . . . . . . 106
8.3 ThestateofNPCs........................ 107
8.4 Futurework........................... 108
8.5 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . 109
A Value definitions for the Game Agent Matrix 111
A.1 Value definitions per column . . . . . . . . . . . . . . . . . . 111
A.1.1 SingleAgent ...................... 111
A.1.2 Multiple Agents . . . . . . . . . . . . . . . . . . . . 112
A.1.3 Social Structural . . . . . . . . . . . . . . . . . . . . 112
A.1.4 SocialGoals ...................... 112
A.1.5 Cultural Historical . . . . . . . . . . . . . . . . . . . 112
B Games included in the studies 115
B.1 BioshockInnite ........................ 115
B.2 BurnoutParadise ........................ 115
B.3 Company of Heroes . . . . . . . . . . . . . . . . . . . . . . . 116
B.4 CrusaderKingsII........................ 116
B.5 DonkeyKong .......................... 117
B.6 Grand Theft Auto V . . . . . . . . . . . . . . . . . . . . . . . 117
B.7 L.A.Noire............................ 117
B.8 RAGE.............................. 117
B.9 The Elder Scrolls V: Skyrim . . . . . . . . . . . . . . . . . . 118
B.10 Warhammer 40,000: Space Marine . . . . . . . . . . . . . . . 118
References cxix
AI Artificial Intelligence
BA Believable Agent
C&N matrix Carley & Newell fractionation matrix
DM Dungeon Master
DND Dungeons and Dragons
FPS First-Person Shooter
GAM Game Agent Model
HCI Human–Computer Interaction
MMORPG Massively Multiplayer Online Role-Playing Game
MUD Multi-User Dungeon
NPC Non-Player Character
PCG Procedural Content Generation
RPG Role-Playing Game
RTS Real-Time Strategy
SciFi Science Fiction
UI User Interface
List of Figures
2.1 Examples of NPCs. Clockwise from top left: An Ork from
Space Marine, an interview subject from L.A. Noire, a guard
from Skyrim, and two soldiers from Bioshock Infinite. . . . . . 32
2.2 Non-Humanoid NPCs. Clockwise from top: street cars from
L.A. Noire, a news van from Burnout Paradise, and an armored
vehicle from Company of Heroes. ............... 35
3.1 An example of indexical storytelling from Bioshock Infinite . . 43
3.2 An example of sensory-based immersion from Grand Theft
Auto V .............................. 47
4.1 The adapted version of the Carley & Newell fractionation ma-
trix, as seen in (Warpefelt and Strååt, 2013) . . . . . . . . . . 60
4.2 The Game Agent Matrix as presented in Paper II (Warpefelt
etal.,2013) ........................... 61
5.1 An example of an NPC that provides services. In this case, a
barber from Grand Theft Auto V ................ 74
6.1 TheNPCtypology ....................... 82
6.2 The Game Agent Model (GAM) with the different areas of ca-
pability marked. Green = high believability, yellow = medium
believability, orange = limited believability, red = only occa-
sional believability . . . . . . . . . . . . . . . . . . . . . . . 97
7.1 TheNPCmodel......................... 101
List of Tables
4.1 Studies in relation to included papers . . . . . . . . . . . . . . 58
4.2 Games included in the different studies, sorted by title. . . . . 67
4.3 Research questions in relation to the papers . . . . . . . . . . 68
B.1 Games used as examples in the thesis . . . . . . . . . . . . . 116
1. Introduction
Over the last couple of decades, the gaming industry has grown from a small
cottage industry to a multi-billion euro industry. As the industry has grown,
we have seen a rocket-like increase in graphic and auditory fidelity, but the
field of Artificial Intelligence (AI) in games has not quite had the same rate
of progress. Although there has been some improvement over the last decade,
many fundamental problems still persist (Johansson, 2013; Warpefelt, 2013).
As explained by Johansson (2013) and Warpefelt (2013), many of the prob-
lems with the behavior of Non-Player Characters (NPCs) stem from a lack of
believability, primarily in their social aspects. The behavior of NPCs must
align with the expectations of the player (Loyall, 1997; Warpefelt, 2013) in
order to be perceived as believable. For this alignment to take effect, however,
the game must convince the player of what is real within the game world, and
provide convincing environments, worlds, and situations in which the player
can immerse themselves (Ermi and Mäyrä, 2005; Murray, 1997).
Convincing the player, however, is more easily said than done. In many
cases, the player’s interpretations of an NPC can and will differ from the de-
sign intent of the developer (Warpefelt, 2015). This can lead to the player’s
misunderstanding the role of the NPC, thus misleading the player and poten-
tially damaging the believability of the NPC and the player’s feeling of immer-
sion. To better understand in what ways these problems arise, and how they
can be remedied, we need to further study NPCs, and establish in what ways
they affect the game experience. Because of this, there is a need for an expla-
nation of what NPCs are, and how we can design them to achieve believability
and maintain the player’s feeling of immersion.
1.1 Purpose, goals and research questions
The purpose of this thesis is to provide a holistic description of NPCs, encom-
passing design, evaluation, and player perceptions. To achieve this purpose, we
must first construct a framework that can be used to better understand NPCs.
In this thesis, that framework takes the form of a typology and a model
of NPCs. The typology will then be used to structure our knowledge about
NPCs. Each of the types in this typology can then be given a thick description
of what it is, and what functions it provides within the game and its narrative.
Furthermore, there will be suggestions as to how each type should be designed
in order to signal their type to players. The model describes in what ways these
types contribute to the believability of NPCs.
The main research question for this thesis has been What makes a non-
player character believable? The answer to this, however, is not something
that is readily apparent. Thus, we needed to explore believability and its as-
sociated phenomena from many different angles. As a result, the research
questions answered in this thesis are:
Q0. What makes a non-player character believable?
Q1. What types of non-player characters exist within games?
Q2. What are the design elements of different non-player character types?
Q3. In what ways do players perceive and determine the type or types of a
non-player character?
1.2 Overview of the method
This section is only aimed at providing an overview of the methods used for
this thesis, as well as the method for the research as a whole. The descriptions
here will therefore be very superficial, and are aimed only at providing a basic
understanding of the research performed. For a more in-depth explanation,
please see Chapter 4.
This thesis uses a design science methdology (Hevner et al., 2004) and
consists of three studies: NPCs and believability, the types of NPCs, and the
design elements of NPCs. These studies are made up of 6 papers, each aimed
at exploring different aspects of the studies. The interrelation between the
studies and specific papers is described in Chapter 4. The research presented
in this thesis is largely empirical, with each study being iterated over a number
of papers in order to achieve triangulation not only on a researcher basis, but
also by examining the object of study from multiple angles.
The methodology used in this thesis is interpretive, using methods such
as various forms of observation and thematic analysis. The data used for this
thesis have been collected partly as video recordings of gameplay, but also
as online surveys centered around long-form answers. Papers I–II used video
recordings that were processed using structured observation, and the obser-
vations were correlated in order to elicit commonalities within each type of
observation. Papers III–V used online surveys which were analyzed using the-
matic analysis. In addition, Paper IV partially uses video data and analyzes it
using thematic analysis. Paper VI uses the data and results from the previous
papers to construct a model of in what ways NPCs affect and create believabil-
1.3 Results and contributions
This thesis provides a holistic description of what NPCs are, including how
they look, how they behave, and how their different aspects are interpreted and
how players identify their functions within the game and the narrative. This
partly takes the form of a typology of NPC roles which, as described in Section
1.1 above, takes the role of a framework in which we can understand NPCs.
This framework was developed primarily in Papers III–VI, but uses theories
from Papers I and II. The need for a new typology was discovered during
the work for Paper III, and the typology itself emerged in Paper IV, and was
verified in its current form in Paper V. The other part of the holistic description
is the model of NPCs found in Chapter 7. This model was created based on
the findings and results from Papers I–V, and was developed in Paper VI.
The results found in this thesis are intended to provide game designers
and game researchers with deeper insights into what types of NPCs are found
in games, and in what ways these NPCs take form. The conceptual model
described in Chapter 7 also provides researchers with a framework in which
they can describe the interrelation of the different parts of NPCs: AI, narrative
anchoring, believability, and the functions provided within the game.
1.4 Reading this thesis
The disposition of the thesis is as follows:
Chapter 2 discusses what NPCs are, what is encompassed by the term, and
what different shapes NPCs may take in games.
Chapter 3 provides the theoretical framework for these studies, and intro-
duce the concepts used to describe the different aspects of NPCs.
Chapter 4 goes into detail about the methodology used for this thesis and
its component parts. It discusses the methodological choices made, and their
Chapter 5 describes how NPCs achieve believability.
Chapter 6 describes a typology of NPCs found in games.
Chapter 7 describes a theoretical model that describes NPCs as a believability-
creating game component.
Lastly, Chapter 8 contains the conclusions of this thesis, as well as avenues
for future research.
2. Non-Player characters
This chapter provides an explanation of exactly what constitutes the primary
object of analysis of this thesis, and what different forms it may take. This
chapter also aims to provide the foundations for a framework by which the
theory from Chapter 3 can be applied to NPCs in particular.
2.1 What are non-player characters?
Non-Player Characters (NPCs), or NPCs, are characters within a computer
game that are controlled by the computer, rather than the player. The term pre-
dates digital games. NPCs are commonly found in, for example, tabletop Role-
Playing Games (RPGs), where they are characters controlled by the Dungeon
Master (DM)1(Mackay, 2001). In computer games, NPCs are found in many
types of games and are not just limited to the genre of RPGs. The role of
the DM, however, has been subsumed into the general system of the computer
game. Despite the transition to a different medium, many of the fundamental
concepts in analog games are still present in their digital cousins, and thus
some of the terminology has been carried over from the analog to the digital.
In digital games, NPCs come in many different shapes and sizes, and as
illustrated in Figure 2.1, they range from the mundane, to the fantastic, to the
distinctly non-human. What exactly makes an NPC an NPC, however, is still a
matter of some debate. Bartle (2004) divided the various computer-controlled
inhabitants of virtual worlds into NPCs and monsters, where NPCs are entities
that look like the player and who would “think they were the same if given
the AI. (Bartle, 2004). Conversely, monsters are entities that neither look
like nor would think that they are player characters. This distinction, however,
has largely become moot with the advance of game genres. Bartle’s division,
based on similarity, is flawed. In many games, notably with fantasy or Science
Fiction (SciFi) narratives, there are numerous different NPCs from different
races and species with which the player interacts both as enemies and as allies.
Because of this, the degree to which the entity looks like the player is not
feasible as a criterion for categorising the entities in a game.
1The person controlling the game and narrative
Figure 2.1: Examples of NPCs. Clockwise from top left: An Ork from Space
Marine, an interview subject from L.A. Noire, a guard from Skyrim, and two
soldiers from Bioshock Infinite.
Because of the lack of proper criteria, there is little point in making a dis-
tinction between NPCs and monsters. Instead, it would be more productive to
discuss the type of NPC, depending on the functions they provide in the game.
This way, we have a common term for the entities found in the game, but can
also separate them into different categories if need be.
This approach may of course in itself lead to some problems, since some
functions (for example, buying and selling items) could potentially be fulfilled
both by seemingly sentient beings and by non-sentient machines. Thus, to be
able to distinguish sentient beings from simple machines, we need some sort
of criterion with which to make this distinction.
2.2 The concept of characterhood
The concept of characterhood was first introduced in (Warpefelt, 2013). Char-
acterhood basically means that the NPC is actively involved in the portrayal of
its role, and will act in ways that are conducive to convincing the player that
it is indeed in that role. This basically means that the NPC must be able to
enact parts of what Dennett (1981) calls personhood. According to Dennett,
personhood is a ladder of six themes:
A NPC does not, however, have to be able to portray the full battery of themes
described by Dennett. Rather, it must match the first two of Dennett’s six
themes: it must be seen as rational (Dennett’s rationality) and it must act in
such a way that to it is ascribed a conciousness of sorts in that it seemingly
performs actions with intentions (intentionality). Matching the further four
themes is of course desirable, and doing so would involve among other things
being able to cause the player to interact with them as if they were persons
(Dennett’s stance), being able to reciprocate social cues (reciprocity), fluently
and dynamically using language (communication), and expressing a concious-
ness of themselves as a thinking entity (conciousness). These are not, however,
part of the fundamental basis that needs to be fulfilled in order to be able to
act as the most basic type of character. As will be seen in later sections, some
types of NPCs are more advanced and others less: although some of these
will need to be able to portray behaviors that match some of Dennett’s more
advanced themes, this is not universally required of all NPCs.
In the previous example of the vendor and the vending machine, a ven-
dor would ask the player what they would like to buy, and provide some sort
of “theatrics” to the purchasing process, whereas the vending machine would
simply accept the player’s money and then dispense the selected goods. For
technical reasons, the distinction may be less clear in older games than in more
modern ones, since playing non-essential sounds and animations may have had
to take the back seat to other aspects of the game. In modern games, however,
this has become less of a concern, as the availability of processing power and
other system resources has increased.
Similarly, NPCs are distinct from entities that are mere scenery. A tree
would be unlikely to achieve characterhood, even if it were equipped with the
sound of wind rushing through its branches, since it is not actively involved
in the process of being a tree. Rather, it merely exists as a tree in the world.
Trees with characterhood would instead be something similar to “Old Man
Willow” from the Lord of the Rings. In essence, characterhood requires that
the NPC has the agency required to take action to fulfill the role they have been
Thus, an NPC that has achieved characterhood has actively acted in such a
way that it can be said to fulfill its role within the game. The extent to which
an NPC could be said to have achieved characterhood is thus very much de-
pendent on what role an NPC is said to have, and the specific attributes and
propensities of the NPC. In some cases, the NPC may not even be recogniz-
ably humanoid, or even portrayed as a “living” thing. As with the example of
Old Man Willow above, some NPCs may portray what are usually considered
inanimate objects. However, the player will still interpret them as being ac-
tors in the game due to their perceived volition. In some cases, an NPC may
achieve characterhood even though its representation is that of an inanimate
object, if through its behavior the inanimate object becomes infused with a de-
rived agency of sorts. In a racing game, a car in motion would be considered
to have achieved at least some level of characterhood, in that it will actively
behave and act in relation to the player (for example as an opponent in a race).
A car in itself would be an inanimate object, but given that it is in motion and
seemingly under control, the player will assume that there is some type of en-
tity in control of the car, even if that entity is wholly unseen. Thus, the car
becomes the avatar of this unseen entity, and, by its design, a car that has been
imbued with characterhood. Examples of NPCs of this type can be found in
Figure 2.2.
Figure 2.2: Non-Humanoid NPCs. Clockwise from top: street cars from L.A.
Noire, a news van from Burnout Paradise, and an armored vehicle from Company
of Heroes.
2.3 Classifying of non-player characters
Characters have long been the subject of media studies, where the different
roles of characters in literature, film, and other arts have been described time
and again (for example by Propp (1968)). To some extent, this work has been
adapted to the field of digital games, notably through the work of Aarseth
(2012). As stated by Juul (2001), however, the study of games cannot be per-
formed only using the lenses of other disciplines, but we must also understand
what the unique characteristics of games are:
“Using other media as starting points, we may learn many
things about the construction of fictive worlds, characters ... but
relying too heavily on existing theories will make us forget what
makes games games: Such as rules, goals, player activity, the pro-
jection of the player’s actions into the game world, the way the
game defines the possible actions of the player. (Juul, 2001)
Thus, the analysis of NPCs in games cannot be made only from fields outside
of game studies, and it cannot only come from the perspective of the narrative.
It is, however, important to remember that NPCs are essentially incompre-
hensible if they are not framed according to the narrative (Warpefelt, 2015).
Because of this, we cannot view NPCs as entities existing solely in either nar-
rative or ludic space, but must instead consider them as entities existing in an
intertwined world of both. However, in previous research, much of the focus
has been on the ways NPCs tie into the narrative, rather than on how they help
support the ludic nature of the game.
2.3.1 Typologies of narrative characters
There are a number of typologies that can be said to exist in the narrative
space. In his work on traditional folk tales, Propp (1968) identified a list of
prototypical characters and what he calls their action spheres. These are:
Villain the opponent of the hero
Donor provides the hero with a magic agent (possibly as a reward)
Helper transports, rescues, or otherwise assists the hero
Princess the person the hero is trying to “acquire” through marriage or
something similar
The princess’s father gatekeeper for the marriage, provides a quest
Dispatcher sends the hero on a quest
Hero Two types:
Seeker hero goes on a quest or completes a task to fulfill the
requirements of the donor and/or the father. Marries the princess.
Victim hero As above, but does not go on a quest
False hero tries to steal the glory from the hero
These characters each appear in different combinations in the folk tales studied
by Propp, and can be combined to form a number of different narratives. Propp
introduces a grammar that can be used to formalize the basic concept of the
stories he has studied.
Characters, and specifically NPCs, have also been studied in the field of
game studies. Aarseth (2012) presented a theory of narrative in games, where
characters play a role as one of the analytical dimensions. Aarseth divides
characters into deep characters,shallow characters, and bots.Deep charac-
ters have believable and seemingly well-rounded personalities, comprehensive
backstories, and evolve with the story. Shallow characters, by comparison,
may have fairly comprehensive backstories, but will rarely exhibit the same
level of personalities as their deep brethren. They will also rarely develop with
the story, but rather remain static over the course of the narrative. Bots are the
proverbial faceless enemies (or allies) of the world: they lack an individual
identity (Aarseth, 2012).
Aarseth’s typology is of course vastly different from Propp’s. They do,
however, represent two different ways of looking at how characters fit into a
narrative. Where Aarseth examines each type of character based on its com-
plexity, Propp instead examines how the story is supported by the type of char-
acter. This illustrates two antipodes on the scale of how characters can be
examined from a narrative point of view. There is, however, more to NPCs
than just the narrative. One must also consider their ludic aspects.
2.3.2 Ludic support character typologies
Unfortunately, fairly little work has been done on NPCs as ludic support char-
acters of the game. Bartle (2004) provided a typology of NPC functional roles
in Designing Virtual Worlds, but the work there has been shown to have been
quite over-fitted to games like Multi-User Dungeons (MUDs), as we have de-
scribed in Paper IV (Warpefelt and Verhagen, 2015). Nevertheless, Bartle’s
work represents the most well-developed work on how NPCs can provide sup-
porting functions in a game. The typology contains the following types:
Buy, sell and make stuff.
Provide services.
Guard places.
Get killed for loot.
Dispense quests (or clues of other NPCs’ quests).
Supply background information (history, lore, cultural atti-
Do stuff for players.
Make the place look busy.
(Bartle, 2004)
Bartle’s types are fairly straight-forward, but some explanation is necessary to
describe the specific aspects of each type. Bartle describes the Buy, sell and
make stuff type as “fictional conveniences; they may as well be vending ma-
chines”. He then states that the same is to some extent true for the Provide
services type. Essentially, these two types are interface access functions that
have been given a diegetic form. Compared to User Interface (UI) elements,
these NPCs are more conducive to an immersive experience, and Bartle de-
scribes them as being player-friendly.
The Guard places type is similar in the way that it allows a designer to less
intrusively introduce restrictions on a player’s freedom by punishing undesir-
able and antisocial player behavior by having someone enfore the “laws” of
the virtual world. It should be noted that in Bartle’s work this type does not
include NPCs that are essentially only placed as guards to prevent the player
from progressing in a dungeon or something similar: such NPCs should in-
stead fall under the Get killed for loot type. Bartle considers NPCs that are
there to be killed for loot are simple monsters, and thus distinct from NPCs.
However, NPCs of the Get killed for loot type are described as “basically just
regular monsters that look like player characters. Bartle also states that these
NPCs rarely even have names, and by extension they are therefore likely to
belong to Aarseth’s narrative type Bots.
NPCs of the Dispense quests type, much like their vendor and service
provider cousins, often are there to provide the player with a diegetic inter-
face to quest management, to provide the player with the mission, and then
to reward the player when they have completed the mission. However, Bartle
states that although this type can co-occur with other types, it is most often
found separately. The inverse of this NPC type is the Do stuff for players
type. Whereas quest givers give the player a task to complete, the NPCs of the
do stuff for players type instead complete missions given to them by players.
These are usually simple missions, for example, carrying the player’s loot or
healing them in combat.
The final two types are directly related to conveying the narrative to the
player. In essence, NPCs of the type Supply background information will act as
dispensers of lore, either after being queried by the player or by forcibly telling
the player the snippet of lore. They will often have some other function, for
example, dispensing quests. The NPCs of the type Make the place look busy
are there to simply make the place look busy. Bartle claims that these NPCs
are uncommon, but in more modern games they are in fact extremely common.
Although Bartle’s typology is, with the exception of our typology2, the
only one focused on the actual in-game functions of NPCs, it has some draw-
backs. It seems to suffer from over-fitting to the specific case it was devel-
oped from, namely, MUDs. Furthermore, Bartle’s typology divides NPCs into
somewhat strictly separated categories, whereas in more modern games, NPCs
are more likely to portray more than one role at a time. Bartle does acknowl-
edge that NPCs can belong to several types, but unfortunately some of the
types in his typology are overly restrictive for the purposes of modern games.
(Warpefelt, 2015; Warpefelt and Verhagen, 2015)
2.3.3 Combining typologies
NPCs play a central role in creating and upholding the player’s feeling of en-
joyment in the game, and in order to do so, they must be perceived as believ-
able. This believability, however, does not arise from how well the NPCs are
portrayed as either narrative actors or ludic support constructs: instead, they
must be a fusion of both. It is a tempting notion to unify the disparate ap-
proaches into one grand typology, capable of describing both the narrative and
the ludic roles of NPCs, but one must also acknowledge that these typologies
are very dissimilar, and may not directly map in a meaningful way. Thus, a
better approach is to add a layer which encompasses the previous typologies.
Each of the typologies presented in this chapter highlights different aspects
of NPCs, and allows us to categorize NPCs according to different criteria.
They also provide us with some of the components we need to create holistic
descriptions of NPCs. That said, there are still some components that are miss-
ing. As mentioned above, a holistic typology will probably be best served by
being constructed as a layer on top of these existing typologies. Unfortunately,
the typologies we have so far discussed are not easily combined, and there is
a need for a theory that can be used as a glue between the typologies. The
following chapters will describe this theory.
2Our typology is described in Chapter 6 and in Papers IV (Warpefelt and Verha-
gen, 2015) and V (Warpefelt and Verhagen, 2016b).
3. Theoretical framework
This chapter describes the theoretical framework for the thesis. This is in-
tended to serve as an introduction to the concepts used throughout the thesis,
and to provide the reader with an understanding of some of the central concepts
used in game studies.
3.1 Narrative
The narrative of the game is essentially the story told as the player progresses
through the game. It is delivered in different ways and from different sources.
Jenkins (2004) divides the narrative into two parts: the embedded and the
emergent narratives. The embedded narratives are narratives that are part of
the designed game experience; for example, the quest lines or in-game envi-
ronments. The emergent narrative is the part of the narrative that comes into
being as the player plays the game. Examples of this would be how NPCs in
the game behave over time.
However, the narrative as an all-encompassing entity connected to all parts
of the game has been criticised by Calleja (2009). He argues that the current
use of the narrative as a concept makes it useless as an analytical tool since
it effectively encompasses the entirety of the game. Instead, Calleja proposes
that we differentiate what he calls the scripted narrative from the alterbiog-
raphy of the player. The scripted narrative is akin to Jenkins’s concept of the
embedded narrative in that it encompasses the pre-defined parts of the game
narrative, e.g., the quest lines from the previous example. Similarly, Calleja’s
alterbiography can be likened to Jenkins’s emergent narrative, with the impor-
tant difference that Calleja does not consider it to be a part of the narrative, but
rather a separate process which uses the scripted narrative to generate a user
experience. In essence, Jenkins’s emergent narrative is the story of how the
player has played the game as told by a narrator, whereas the alterbiography is
the story of how the player has played the game as described by the player.
3.2 Indexical storytelling
The framework for the alterbiography is constructed by providing the play-
ers with clues as to what to expect from the world, which Fernández-Vara
(2011) calls indexical storytelling. In essence, the environment and actors of
the game will have small indicators, called wieners by Fernández-Vara, that
clue the player into what to expect from this environment and its denizens. To-
gether, these indicators, through Fernández-Vara’s indexical storytelling, form
part of the embedded narrative described by Jenkins (2004). Combinations
of indicators can be used to create certain types of environments, which the
player will identify as signaling certain things. Basically, the indicators de-
scribed by Fernández-Vara (2011) are used to provide the player with clues as
to what type of environment or NPC they are encountering, and thus invoke
certain preconceived notions. These indicators can be found in many different
elements of the game, including the architecture of the buildings, environmen-
tal effects such as fog, the items worn by NPCs, and the items placed in the
game world. For example, an environment with chimneys, transportation in-
frastructure, and a lot of various goods will likely be interpreted as belonging
to an industrial sequence. The game environment essentially becomes a simu-
lacrum (Baudrillard, 1994), i.e., an instantiation of something that only exists
in the narrative context of the game. In essence, it is the realization of a con-
cept, such as the aforementioned industrial environment. This is illustrated by
Figure 3.1, where we see many indicators that tell us about the environment
and narrative. Waggoner (2009) calls this kind of characteristic simulacratude.
The simulacratude of an environment provides the player with clues as to what
to expect from the game, and in what ways the narrative will be communicated.
In essence, this forms an implicit contract between the designer and the player;
breaking this contract may adversely affect the believability of the game, and
by extension lessen the player’s enjoyment of the game.
In order for these indicators to actually strengthen the player’s sense of
immersion, however, they must form a coherent picture. If the indicators are
unclear or conflicting, they can instead cause the player to misunderstand what
the game is trying to convey. This can lead to the player’s perceiving the
game as non-believable, thus shattering the suspension of disbelief (Coleridge,
1817) and negatively affecting the player’s sense of immersion. In addition,
this can also cause the player to become frustrated with the game since they
can’t interact with it in the manner they expected.
Figure 3.1: An example of indexical storytelling from Bioshock Infinite
3.3 Immersion
The concept of immersion is defined by Bartle (2004) as the feeling the player
has of losing themselves in the game world. Murray (1997) describes immer-
sion as “the sensation of being surrounded by a completely other reality, as
different as water is from air, that takes over all of our attention. In essence,
achieving immersion means that the player starts accepting the game world as
a real world, and acts in it as if it were a real world. This effect is akin to the
concept of suspension of disbelief, and can be seen as an effect akin to that
described by the Thomas Theorem: “If men define situations as real, they are
real in their consequences” (Thomas and Thomas, 1928). Thus, to achieve im-
mersion, the player must accept that the game world has a set of rules that may
be different from the real world, and that such a world can be “real” despite its
working differently.
Immersion has, however, long been criticised for being what McMahan
(2003) calls “an excessivly vague, all-inclusive concept.” In her 2003 article,
McMahan describes two different phenomena: when “the player is caught up
in the world of the game’s story”, and their “love of the game and the strategy
that goes into it”. McMahan claims that these define immersion on a diegetic
and non-diegetic level, respectively. In later research, Ermi and Mäyrä (2005)
categorized immersion into certain subtypes, which independently and in co-
operation cause the player to achieve an immersed state. Ermi and Mäyrä
divided immersion into imaginative,challenge-based, and sensory-based im-
mersion. Other authors have also discussed other aspects of immersion, for ex-
ample, the narrative,tactical,strategic (Adams, 2010), and social immersion
(Johansson, 2013). These types of immersion all highlight different aspects of
the overall concept, and in combination they help address some of McMahan’s
critique of the concept.
However, the different types also have some amount of overlap. In essence,
tactical and strategic immersion (Adams, 2010) are two aspects of the challenge-
based immersion presented by Ermi and Mäyrä (2005). Narrative (Adams,
2010) and imaginative (Ermi and Mäyrä, 2005) both describe how the player
achieves immersion on what McMahan (2003) calls the diegetic level, i.e.,
through immersion in the story of the game.
3.4 NPCs and the specific forms of immersion
NPCs affect the different forms of immersion differently. Although this list
is probably not exhaustive, it provides a broad view of the ways the player’s
sense of immersion can be defined. The sections below will explore in what
ways the impact of NPCs on immersion can be described within the context of
each form of immersion.
3.4.1 Imaginative and narrative immersion
Imaginative immersion (Ermi and Mäyrä, 2005) and narrative immersion (Adams,
2010) describe two very similar concepts. In essence, both describe how the
player is immersed through the telling of a story and the stimulation of the
imagination. They do, however, present different aspects of how the player
is affected by the narrative. Imaginative immersion is more closely tied to
the concept of alterbiography (Calleja, 2009), whereas narrative immersion is
more closely related to the pre-defined (Calleja, 2009) or embedded (Jenkins,
2004) narrative.
NPCs affect these forms of immersion in a variety of ways. As mentioned
in Chapter 3, NPCs can affect immersion on the diegetic level (McMahan,
2003) of the game. This means that through their presence in the game, and
the way they perform their roles as actors in the narrative, they will affect
how the player experiences the story of the game, and how the player there-
fore builds their alterbiography (Calleja, 2009). This means that NPCs must,
to some degree, be designed as a part of the narrative of the game, so that,
through their design as a part of the pre-defined narrative, they can influence
the player’s alterbiography. This need not be in the most complicated way:
the types described by Aarseth (2012) range from deep characters, to shallow
characters, to faceless bots. Each NPC needs to be assigned a role in the nar-
rative and needs to have the capabilities and appearance to fulfill that role. If
they do not, they will negatively affect the player’s sense of immersion.
3.4.2 Challenge-based immersion
Challenge-based immersion (Ermi and Mäyrä, 2005) is essentially the type of
immersion the player derives from being challenged. This is very closely tied
to the concept of flow described by Csikszentmihalyi and Csikszentmihalyi
(1992). Basically, flow is achieved when the player’s skill and the challenge
of the game form an equilibrium, and the player enters a state of optimal con-
centration, and the rest of the world falls away. This is very similar to the
description of immersion. The flow state, however, can be disrupted by the
challenge’s being too hard or too easy relative to the player’s level of skill.
This will also have a jarring effect on the player, where the illusion of a con-
querable yet somewhat difficult challenge is shattered and the situation loses
believability—either through sheer frustration or by the player’s failing and
being launched into a death screen or the like. NPCs can affect this type of im-
mersion by being either too skilled at their task, or not skilled enough to pose
a credible challenge. If one applies the work of Adams (2010), the concept of
challenge-based immersion can be further broken down into different types. In
essence, challenge-based immersion can be said to work on two levels: tactical
and strategic.
Tactical immersion
Tactical immersion (Adams, 2010) is the type of challenge-based immersion
that arises from the immediate need to act in a situation. The player must
make rapid decisions to overcome challenges. This type is often found in fast
paced games, such as First-Person Shooters (FPSs) or timed puzzle games.
Adams calls it tetris trance. This type of immersion can be greatly influenced
by NPCs, especially in games where combat situations and the like are com-
mon, such as FPSs or action RPGs. The NPCs will often meet the player as
the “opposing team, but can also be used to assist the player in overcoming
situations (Tremblay and Verbrugge, 2013). NPCs must behave in ways that
are in accordance with the expectations of the player, and be skilled enough to
not be too easily conquered. They cannot, however, be too skilled: this will
cause the player to become frustrated, and thus drop out of the flow channel
and lose their sense of immersion. This is a difficult balance between difficulty
and stimulus, and designing NPCs for this purpose is a very active field of re-
search (Orkin, 2006; Weber et al., 2011). If the assisting NPCs require a lot
of management, this can potentially be immersion-breaking if it interrupts the
flow of the game, or transform the scenario so that strategic immersion instead
becomes the more prominent type.
Strategic immersion
Strategic immersion (Adams, 2010) is the form of challenge-based immersion
that arises not from the more immediate situation of the game, but from the
overarching strategy of the game. This type of immersion is found notably in
strategy games, especially of the turn-based, grand strategy, variety. Adams
describes this as the immersion of the chess master: the importance of things
such as narrative and game world fall to the side, and instead the overall grand
plan becomes the sole focus. This type of immersion is difficult to achieve
with NPCs, and they will probably only act as pawns in the different tactical
situations that make up the strategic landscape. Although strategy games will
often have one or more opposing AI, this is rarely diegetically represented as
a singular NPC, but rather as an incorporeal entity much like the player them-
selves. In some games it is possible to find NPCs without a diegetic represen-
tation that will still influence immersion, for example, the nobles in Crusader
Kings II (see Section B.4).
Figure 3.2: An example of sensory-based immersion from Grand Theft Auto V
3.4.3 Sensory-based immersion
Sensory-based immersion (Ermi and Mäyrä, 2005) arises from the audiovisual
aspects of the game. Through a combination of interesting and attractive ar-
chitecture and nature, lighting, animations, camera angles, sound, and music,
the game can lull the player into a sense of immersion, as exemplified in Fig-
ure 3.2. This type of immersion needs to be complemented with narrative or
imaginative immersion, which will provide the frame for the appearance of the
game. NPCs will affect this type of immersion through their appearance and
the sounds they make. Again, it is important that these are in accordance with
the narrative. Most of the work for this type of immersion will happen as the
game is designed and implemented, and the situations that strengthen this type
of immersion will often be pre-designed experiences.
3.4.4 Social immersion
Social immersion (Johansson, 2013) is the form of immersion that arises from
social interactions within the game. As Johansson defines it, this exists only
between the players and the NPCs, and arises when their interaction reaches
such a level of complexity that the player becomes immersed in the social
interaction itself. Of the types of immersion described in this thesis, this is the
least developed type.
3.5 Believability
In order for the player to achieve a feeling of immersion, the game must draw
the player in and cause them to become engaged in the game. This requires that
the game persuades the player that it is believable. Furthermore, the narrative
must contextualize the player experience and provide a framework in which
the player can construct their alterbiography (Calleja, 2009). However, many
of the factors that affect believability also affect immersion, and to some ex-
tent immersion also affects believability through the narrative anchoring1that
immersion provides. In essence, the immersion and believability co-exist and
feed into each other. Because of this, a negative feedback loop in either will
probably spread to the other:
“The key to immersion is persuasion. The more persuasive
an environment is, the easier it is to become immersed in it. The
biggest weapon in the designer’s armor of persuation is familiar-
ity. [...] When knowledge and belief coincide, that’s immersion.
(Bartle, 2004)
Loyall (1997) exemplifies this persuasion with the Disney classic Dumbo. The
movie tells the viewer that the “Dumboverse” is a world where animals speak,
and where an elephant can fly using its ears, thus making Dumbo believable as
a character. This builds a narrative that becomes convincing, and provides the
framework for the alterbiography.
However, the problem of contextualizing characters in virtual environ-
ments is not unique to games. In the field of AI there exists a subfield con-
cerned with the development of agents2, which are essentially pieces of soft-
ware that act on their own volition in order to achieve some goal. Some of
the research into agents is applicable to NPCs, particularly in cases where the
agent research is concerned with believability or social capability—most often
called Believable Agent (BA).
Mateas (1999) discusses the difference in research goals between the tra-
ditional AI approach and that of the BA approach. Whereas the goals of AI re-
search are competence, objective assessment, generality, and realism, the goals
of BA research are personality, audience perception, specificity, and charac-
ters. Both of these are of course important to NPCs, but one particular point
1The narrative anchoring of NPC believability and immersion encompasses both
the alterbiography and the pre-defined narrative (Calleja, 2009), as well as the embed-
ded and emergen narratives (Jenkins, 2004).
2Agents share many characteristics with NPCs, and to some extent NPCs can be
said to be a special case of agents. But it is important to note that not all agents are
that Mateas makes is that “[t]he success of a believable agent is determined by
audience perception. If the audience finds the agent believable, the agent is a
success. This approach is very much relevant to NPCs, which are almost en-
tirely judged by how believable they appear to be. It also encapsulates the core
of immersion, in that the one must reach acceptance with the target audience.
A basic level of competence is of course required to achieve believability, but
as stated by Bartle (2004) NPCs, cannot be too competent or they will spoil the
fun of the game. Consider, for example, an NPC that has perfect aim. This is
trivial to achieve in most games (by adding a node to the most efficient point to
hit the target, we can make the NPC a superhuman marksman) but this would
not be perceived as believable since no one will score perfect hits every time.
Instead, in order to maintain believability, the NPC must maintain a level of
marksmanship that falls inside the scope where they are believably accurate,
without being so competent that the player’s level of fun is disrupted. For
highly delimited tasks such as marksmanship this is fairly simple: simply add
some deviation to the aim of the NPC. For less delimited tasks, for example
social interaction, the problem becomes much more complicated.
Loyall (1997) presents a list of the abstract properties that an agent must
possess in order to be perceived as socially believable. These are:
Self Motivation
Social relationships
Illusion of Life
Appearance of Goals
Concurrent pursuit of Goals
Parallel Action
Reactive and Responsive
Resource Bounded (in body and mind)
Exists in a Social Context
Broadly Capable
Well Integrated (capabilities and behaviors)
(Loyall, 1997)
This list contains many attributes that are closely related to the social capa-
bilities of BA. As one can discern from Loyall’s approach, the capability for
social exchange and interchange is extremely important for the believability of
agents. However, these are only part of what makes agents, and by extension
NPCs, believable. In order for the player to understand and interpret the be-
havior, it must also be framed in a narrative in which the behavior makes sense.
This also ties into the indicators described by Fernández-Vara (2011): NPCs
are littered with small clues as to what they are and what they do. Thus, play-
ers must be able to perceive and identify what type of NPC they are interacting
with. In short, the NPCs must afford certain interactions.
3.5.1 Affordances and believability
In the field of Human–Computer Interaction (HCI), the concept of affordances
(Gibson, 1977) exists as a way of describing how users understand in what
ways they can interact with the software. The software must indicate what will
happen when the user performs a certain interaction. When an object in a user
interface is interactable, its appearance must indicate what will happen when it
is interacted with. For example, the printing icon in a word processor will look
like a printer and thus indicate that clicking the icon will send the document to
the printer.
According to Gibson (1977), these affordances exist independently of the
entity that is trying to perform these actions. However, the notion that the ex-
istence of affordances is independent of the acting entity has been criticised
by later research, notably by Norman (2013), who argues that the perception
of affordances is key to being able to use them. This is further expanded
by Mcgrenere (2000), who point out how one must not only be able to per-
ceive the affordance, but also do so correctly. Failure to do this can result
in the non-detection or the false detection of an affordance. Furthermore, as
described by Mcgrenere, Norman’s affordances are also dependent on the ex-
perience, knowledge, and culture of the acting entity. The implication of the
experience-dependency of Norman’s affordances is that they can also to some
extent be taught. Before the demise of the floppy3disk as a conventional stor-
age medium, the use of a floppy disk as a save icon also indicated an affor-
dance. Nowadays this has largely been relegated to convention, where users
who have never actually seen a floppy disk still know that clicking the floppy
disk icon will save the document.
Similarly, the perceived affordances in games must be in line with what the
player expects. According to Linderoth (2013), the perception and manipula-
3The name “floppy disk” is itself a relic of the past, since the later generations (for
example 3.5" disks) were in fact rigid rather than floppy.
tion of affordances is central to the act of playing a game:
“To engage in game-play is to perceive, act on, and transform
the affordances that are related to a game system or to other play-
ers in a game. (Linderoth, 2013)
For players to be able to use these affordances, they must first be able to iden-
tify them (Mcgrenere, 2000). This can become problematic if the player is
unable to discern and interpret the affordances presented by the game. This
transforms the challenge of playing the game from what Linderoth (2013) calls
performative to being what Linderoth calls exploratory. In performative chal-
lenges, the objective of the challenge is perfectly visible to the player, and the
challenge lies in the performance of the steps needed to complete the chal-
lenge. In an exploratory challenge, the difficulty instead lies in discovering
what is to be done, and the action(s) needed to complete the challenge are
fairly simple. If the affordances needed to complete a performative challenge
are poorly designed, the challenge instead becomes both an explorative and a
performative challenge, potentially increasing the difficulty of the challenge.
Thus, identifiable affordances is a primary design concern for any game,
and something that must be taken into account. In addition, these affordances
must, as previously mentioned, be in line with what the player expects. The
items in the world must, if they are to be interactable, be interactable in such a
way that they conform to the expectations a player would have of such an item.
For example, a wooden crate would afford stacking, climbing onto, putting
items on, and putting items in. If it fulfills these affordances, it could poten-
tially be seen as believable4.
3.6 Concluding summary
As described in this chapter, the evaluation of the player experience cannot
begin and end with narrative and immersion. Instead, these are just the first
links in a chain of phenomena that combine to create the player experience. If
we examine this chain of phenomena in reverse order, we can see that game
developers, by having affordances that the player can manipulate in expected
ways, can create believable NPCs which can then act in ways that induce the
player to enter a state of immersion. This then enables the player to use the
indicators found in the game as a basis on which they can build their alterbiog-
raphy, and experience the emergent narrative of the game. In turn, this shapes
4It should be noted that some items may not have affordances that are directly
matched to their real-world counterparts. The crate mentioned above may, for exam-
ple, not be openable, and