=Paper=
{{Paper
|id=Vol-1680/paper5
|storemode=property
|title=Eliciting Emotions in Design of Games - a Theory Driven Approach
|pdfUrl=https://ceur-ws.org/Vol-1680/paper5.pdf
|volume=Vol-1680
|authors=Alessandro Canossa,Jeremy Badler,Magy Seif El-Nasr,Eric Anderson
|dblpUrl=https://dblp.org/rec/conf/recsys/CanossaBEA16
}}
==Eliciting Emotions in Design of Games - a Theory Driven Approach==
https://ceur-ws.org/Vol-1680/paper5.pdf
Eliciting Emotions in Design of Games – a Theory Driven
Approach
Alessandro Canossa Jeremy Badler Magy Seif El-Nasr
Northeastern University Northeastern University Northeastern University
360 Huntigton Ave. 360 Huntigton Ave. 360 Huntigton Ave.
Boston, MA Boston, MA Boston, MA
a.canossa@northeastern.edu jbadler@ski.org m.seifel-
nasr@northeastern.edu
Eric Anderson
Northeastern University
360 Huntigton Ave.
Boston, MA
ec.anderson@northeastern.edu
ABSTRACT • Human-centered computing → Interaction design →
As technology becomes more powerful, computer software and Interaction design process and methods → Scenario-based design
game designers have ever-expanding tools available to create
immersive, emotional experiences. Until recently, designing
emotional experiences was achieved by veteran designers relying Keywords
on insights from film theory and intuitions developed through Emotion elicitation; Affect; Game design; Personalization;
years of practice. We propose another approach: leveraging Psychology of affect; Conceptual Act Theory.
scientific knowledge of emotions to guide the design process. The
approach can serve as a resource for fledgling designers trying to 1. INTRODUCTION
break into the field, but will hopefully provide a few new insights Creating emotionally engaging experiences is an important
for veterans as well. In addition, it may interest emotion goal of game design. Game designers and developers use many
researchers and psychologists looking to expand their stimulus different design techniques to evoke emotions. The Mechanics,
repertoire. As a necessary underpinning for the design process, we Dynamics, Aesthetics (MDA) model, for example, advocates for
will discuss several theoretical psychological models of emotions. the development of mechanics (game rules) that lead to game
Classical theories largely treat emotions as basic, universal states dynamics (game systems) that achieve aesthetic goals. The goals
that are invariantly evoked by specific stimuli. While intuitive and are defined as states that include: sensation (games as sense-
popular, these theories are not well supported by current evidence. pleasure), fantasy (game as make-believe), narrative (game as
In contrast, a psychological constructionist theory, called the drama), challenge (game as obstacle course), fellowship (game as
Conceptual Act Theory [6] proposes that emotions are constructed social framework), discovery (game as uncharted territory),
when conceptual knowledge is applied to ever changing affective expression (game as self-discovery) and submission (game as
experiences. The CAT proposes that emotional states can exhibit pastime) [74]. Several game design authors have proposed
strong variation across instances and individuals due to principles that describe the role of visual design, environment
differences in situational factors, learning histories and cultural design and other physical properties of games and how they
backgrounds. This theory better fits available data, and also change over time as a way to evoke affect and a general sense of
provides a framework for modeling emotional changes that vary pleasure [75, 76]. The use of writing techniques to develop
by situation, person, and culture. The CAT also fits better with the character and narrative in games that have emotional impact has
game design process, since it treats users holistically as received attention as well [28]. There have also been several
individuals. During the course of a game, similar to real life, works discussing the development of reward systems to encourage
emotions emerge from evaluations of situations and can therefore player achievement, competition or collaboration as a way to
not be deterministically dictated by a single stimulus. Using the evoke emotions and sustain engagement (e.g. [78]). Virtual
CAT framework, we developed a process to create affective environment researchers have also acknowledged the potential
digital game scenarios. Our goal is to give game designers, a and utility of adopting psychological theories of affect and
scientific framework to better guide the design process. emotions. One area where emotion theory has been used is in
developing computational models of emotion elicitation for
CCS Concepts creating believable characters. Examples of this work include the
Oz project, where the research group used scientific "appraisal"
models of emotions [79] to develop expressive believable agents
that can inhabit a virtual narrative world [50].
However, top designers see the game experience holistically.
Thus the process of evoking emotions arises not just from
characters that are expressive or believable, but from the complex
EMPIRE 2016, September 16, 2016, Boston, MA, USA. interaction of all game elements: lighting, movement, sequences
Copyright held by the author(s). of events and user choices, and from the overall feel of the
eenvironment [770].This necessitates a different kind of theoory to witthin a culture have varied emotional livees [8]. Whilee a
gguide the desiign process – an approach that treats thee user commplete review of this researcch is beyond the scope of this t
eexperience as a whole ratherr than as separrate componennts (as chaapter, interestedd readers can coonsult Barrett et al. [9].
cclassical theoriees of emotion ddo). The formal research cited aabove
iis in many waays the exceptioon; more oftenn than not, desiigners Another wayy of characteriziing emotional sttates is in termss of
ddevelop techniqques and makee choices basedd on their experrience theeir underlying affective dimennsions. Two im mportant affecttive
aand intuitions. Experience annd intuition aree difficult to codify. dimmensions are vaalence, the degree of pleasure oor displeasure, and
a
TThus we pose thhe question: can a psychologiccal theory of em motion acttivation, the deggree of arousal [7, 8, 65, 67]. Together, valennce
bbe used in thee game designn process to ennhance the plaayers’ andd activation forrm a unified aff
ffective state (Figure 1). Affecct is
eemotional expeerience in a gaame? If so, hoow, and what is an groounded in the phhysical fluctuattions of the boddy: somatovisceeral,
ooptimal emotioon theory for thhis purpose? We W propose that the kinnesthetic, proprrioceptive, and neurochemical [7, 59]. Affectt is
CConceptual Acct Theory [6]] can be useefully employeed by also a central feature in many psychologgical phenomeena,
ddesigners. Takiing a holistic view, the theoory builds on strong s inccluding emotionn [7, 8, 20, 65], anticipating thhe future [31, 332],
eevidence that em motions are nott hardwired or invariant
i entitiees that psyychopathology [18, 19], andd morality [366, 38]. Affecttive
ccan be triggereed by specific stimuli. Ratherr, the CAT prooposes chaanges are cruccial to the connscious experience of the woorld
tthat emotional instances are nnewly created eeach time they occur aroound us [24]. PPeople in all cuultures around tthe world seem m to
ffrom the sum oof all stimuli, anda vary as a person’s internall (i.e., havve affective exxperiences [53].. Unlike emotioons, affect can be
tthe person’s bbodily state) annd external coontext changes. The cleearly measured in the facial exxpressions [16], in the voice [666],
iinstances also vvary across indiividuals who haave different em motion andd in the peripheeral nervous sysstem [15, 16]. A As a consequennce,
cconcepts, learnning histories aand cultural baackgrounds. Wee first afffect can be thouught of as a neeurophysiologicc barometer of the
ooutline differennt psychologiccal theories off emotion and their inddividual’s relatiionship to an environment
e att a given pointt in
llimitations. W We then descrribe in moree detail the CAT, tim
me, with self-repported feelings as
a the barometeer readings.
eemphasizing inn particular featuures of the theoory that are critiical to
oour game desiggn and iterative tuning process,, and describingg how
tthe theory is different from m others currenntly used by game
rresearchers. Seccond, we review w previous worrk in creating gaaming
eexperiences usiing emotions. T Third, we descrribe a design prrocess
ffrom concept inception to reaalization, througgh the examplee of a
ggame created for research ppurposes. We conclude the paper
bbriefly discussing our evaluaation of the ggame’s usabilityy and
pplayability as well as descrribing an initial study wherre we
ccompare the seelf-reported andd peripheral phyysiological respponses
oof the initial poool of subjects. Last, we discusss our contributtion to
tthe game desiggn process, as well as the efffectiveness of a new
ttheory of emootions that has not previouslyy been used iin the Figgure 1. Circum
mplex model of affect.
ddomain of affecctive computingg. We believe thhe paper will prrovide
ppromising evideence of the utiliity of the approaach, which mayy open
2.22 Concepttual Act Moodel
nnew research diirections in the design of emotiional experiencees. Using affect as its foundatioon, the CAT [6]] hypothesizes tthat
afffective experiennce becomes a ‘real’ emotion (fear, anger, eetc.)
22. THEORRETICAL FOUNDAT
F TION – whhen categorizedd as such using the emotion cooncept knowleddge
of a perceiver. Thhese concepts hhave been learned from languaage,
PPSYCHOL LOGICAL THEORY OF soccialization, and other cultural artifacts withinn the person’s dday-
EEMOTION NS AND AF FFECT to-day experiencee. The process of combining incoming sensory
inpput (from the bbody and from the surroundinngs) with learnned,
22.1 Emotioons and Aff
ffect cattegory knowleddge within the pperceiver’s braiin is a normal ppart
A commonnly held view oof emotions is tthat there existss a set
of what it meanns to be consscious. This cconceptualizing is
oof discrete, innnate and universsal emotional sttates [25, 26, 41, 42,
insstantaneous, onngoing, obligatoory, and autom matic (meaningg, a
445, 63]. This seet of emotions is often referredd to by such Ennglish
perrson will normaally not have a sense of agencyy, effort or conttrol
wwords as angerr, sadness, andd fear, and are viewed as a nnatural
in constructing ann emotion). Coonceptualizing is rarely due too a
kkind [6]. Whenn boiled down to their fundaamental assumpptions,
delliberate, conscioous goal to figuure things out. Thus to a persson,
bbasic emotion m models make uup the dominantt scientific paraadigm
emmotions feel likke they just happpen. The CAT emphasizes the
iin the psychoological studyy of emotion.. Different m models
impportance of sittuations. The coonceptual systeem for emotionn is
eemphasize diffeerent parts of thhe process. For instance, one ffamily
connstituted out off past experiencce, and past expperience is larggely
oof theories calleed "appraisal models", focus onn the set of neceessary
struuctured by peoople within a cultural contexxt. Therefore, the
eevents that triigger emotionss [30, 37]. O Once an emotion is
voccabulary of emootion categoriess that is developped, as well as the
ttriggered, the prresumed result is an automatedd set of synchroonized
poppulation of innstances withinn each categoory are culturaally
cchanges in respponse systems thhat produce thee signature emootional
relative. Such pproperties inteegrate the CAT C with soccial
rresponse. This vview predicts thhat the experiennce and perceptiion of
connstruction apprroaches, positinng that interpersonal situatioons
eemotions are fairly universaal, so little vvariability withhin or
“affford” certain emotions (or certain varietiees of an emottion
bbetween peoplee would be obbserved. While intuitive, the ''basic'
cattegory). As a result, in the CATT emotions (likee all mental stattes)
eemotion view iis not well suppported by the data, variability is the
aree not assumed to be Platonic, physical typees, but instead are
rrule rather than the exception. Quantitative revviews of the ressearch
treated as abstractt, conceptual caategories that aare populated wwith
hhave failed to fi
find signatures oof emotions in thhe body [16] orr brain
varriable instances optimized for a particular situation or contextt.
[48]. Additionnally, evidencee is emergingg that people from
ddifferent culturres perceive em motions differenntly [29] and ppeople
According to the CAT ,there are at least five sources for the models do not implement discrete emotional categories, but rather
variations that occur in emotional episodes: (1) the behavioral treat emotions as continuous variables (i.e. affect; see Figure 1).
adaptations that serve as initial, affective predictions about how to For example, WASABI [11] defines different emotions as ranges
best act in a particular situation (e.g., it is possible to freeze, flee, in arousal-valence-dominance space, appraises the current
fight or faint during fear), (2) the concepts that develop for situation in the same space, and uses the distance between the two
emotion, (3) the vocabulary used for emotions, (4) the variation in to calculate a likelihood of a given facial expression. Anatomical
the types of situations that arise in different cultures, and (5) models [4] are built from the ground up based on neuroanatomical
stochastic processes. As a result, there is variation within emotion data and processes. As such, they tend to be focused on a single
categories, both within individuals and across people and cultures. emotion (e.g., fear) and have received only limited attention from
Not everyone will experience the same emotion to the same the computational community. Finally, rational models are in
stimulus, and even the same stimulus/person pairing can create many ways the opposite of anatomical, eschewing psychology
different emotions at two different times. almost entirely in favor of a pure artificial-intelligence approach.
A good example is Scheutz and Sloman [68], who use the simple
2.3 Utility of the Conceptual Act Model to the affect “hunger” to modify the behavior of intelligent, sensing
game design process agents in a world populated by other agents, food and various
Modern games contain complex, dynamic worlds that are lethal entities.
well-suited to the application of the CAT for creating emotional Computational models of emotions are often put to use in the
experiences. Several key features of the CAT model are broader context of believable characters. Indeed, if computer-
particularly relevant for game design: controlled agents are ever to appear “human”, their ability to
1) There is variability in how people will respond to stimuli. realistically express emotion is almost a requirement [10]. Once
This can be due to participants’ past experiences, or contextual an agent has selected the appropriate emotion via an affective
elements present in the situation that can be interpreted differently model, the agent needs to behave accordingly. An agent's
by different participants. emotional state can be conveyed visually by head position and
facial expression [5, 22, 23] as well as body posture and
2) The context is critical for the experience users will have. movement [1, 3, 17, 60, 61]. The link to cognitively-driven
A snake may elicit fear in one context, but amusement in another. behavior was recognized and exploited early on by the Oz project
3) The sequence of events that lead to a specific situation is [72, 73], which developed an expressive artificial intelligence
important when developing an emotional scenario. informed by emotional state. More recently, Hudlicka and
colleagues [39, 40] modeled affect-induced changes in cognition,
A user’s response is not solely determined by current such as an increased threat response if the agent is anxious. Many
conditions; it is also influenced by the preceding sequence of of the researchers computationally modeling emotions use
events. This is an important element of game design. Designers appraisal theory as a theoretical foundation for good reason.
often use a “beat chart" to signify the sequencing of events or Appraisal theories focus on emotion elicitation - exactly what the
beats (single units of action) and their effect on the participant as researchers are attempting to model. For designers, such projects
they go through an experience. Using these three constructs, we are interesting but leave out an important element: the actual
will discuss in Section 4 a framework to guide the design process experience of an emotion. Games seek to provide a holistic
and show how these ideas can aid in developing and designing experience to the player, and since the above models do not
emotionally engaging scenarios. First, however, we briefly review include subjective experience they are of limited use to designers.
current theory on designing for emotions. Many game designers have therefore abandoned the use of
emotion theory and instead adopted an alternative approach, either
3. PREVIOUS WORK ON EMOTIONS IN (a) creative methods that borrow techniques from other disciplines
DESIGN (e.g., film theory) and rely heavily on intuition, or (b) a more
scientific approach where the design is still creative, but is tuned
3.1 Computational Models of Emotions through the iterative process of testing, evaluating outcomes and
Computer scientists have attempted to model how emotions modifying game variables as needed [2, 55, 54].
are elicited by modeling them using digital environments.
Marsella, Gratch and Petta [51] summarized several 3.2 Creating Emotions in Interactive
computational models of emotions. Most of these models use
classical appraisal theories as the theoretical foundation, with the
Experiences
Artists, designers, directors and other content creators often
goal of developing 'emotional' virtual characters used in games
seek to evoke or manipulate the emotions of those who experience
and simulation. Computationally-based appraisal models assess
their work. They are interested in the holistic experience of the
events in the surrounding environment, compare them to an
user. Many design techniques were documented in the 1960s and
internal belief system, and change their emotional state
70s, with the rise of film theory as an academic discipline. In
accordingly. For example, the EDA model [33, 35] parameterizes
films and television [13] as well as advertising [64], visual
external events in terms of desirability and likelihood of
scenery and ambient light and color play a particularly important
happening, which are then used to map to specific emotions. For
role. For example, according to Western cultural norms the color
example, positive desirability with likelihood < 1 yields hope,
red often evokes violence or passion, while blue is methodical and
while negative desirability with likelihood = 1 yields distress. A
cold [12]. Games are no exception [62], and may be even more
good example of customized internal beliefs is the bully agent in
effective conductors of emotion since they provide levels of
the FearNot! system [21], which interprets as desirable another
control and immersion that are impossible using classic
agent having fallen on the floor and crying (having been pushed
techniques [34, 56, 69]. One study [27] asked participants to
by the bully); accordingly, a gloating response is produced.
navigate through versions of a virtual environment that differed
Besides appraisal models, three other categories of affective
only in some visual dimension (color, saturation, brightness or
modeling are dimensional, anatomical and rational. Dimensional
ccontrast). It fouund a measuraable effect on pphysiological siignals earrthquakes. Playyers were instrructed to retrieeve a shovel and
a
ssuch as heart raate and body temmperature. Asidde from visuals, other esccape the cave.
ssensory stimuli such as music and sound [58] and even scennt and • The Callm Valley (Figgure 2B) was a peaceful, natuural
vvibration [56, 669] can also enhhance the gamepplay experience.. areea with trees, fflowers and a lake. Players w were instructedd to
Optimizingg a user’s sennsory experiencce is not suffiicient, retrrieve a flower aand place it in thhe middle of thee pond.
hhowever – theree are also the unnderlying storyy and gameplay itself. • The Excciting Casino (F Figure 2C) had upbeat music anda
TThere are manny narrative tecchniques that increase the plaayer’s maany lively charaacters. Players w
were instructed to pick up a luccky
eemotional connnection to thee story, such as creating ddeeper chiip and play a sslot machine thhat caused prizees to fall from the
rrelationships wwith one or moore non-playerr characters (N NPCs), ceiiling and non-pllayer characterss to cheer.
iincluding intereesting and multiilayered plot eleements, and alloowing
tthe player to innfluence the stoory arc [28]. EEven simple, scripted • The Sadd Hotel (Figure 2D) was a run--down and som mber
pplot elements aare sufficient too evoke emotioons like joy or anger interior. Players were
w instructed to fulfill the last wish of a dyying
[71]. NPCs witth emotional deepth can be im mplemented usinng the maan.
aaffective compputing methodds surveyed eaarlier (section 3.1). The rest areaa was the Hub (Figure 2E), aappearing betweeen
TTechnical agenncy, such as giving players controlc of the game eacch scenario as well as at thee beginning andd end of the pplay
ccamera, is critical for avoidinng frustration inn certain gamess [52]. sesssion. The hub wasw virtually em mpty and featurred a character iin a
EEven subtle diistinctions are important: Leiino [46] arguess that labb coat (“Doc”) tthat interviewedd the player aftter they compleeted
pplayers are morre likely to expeerience emotionns from game coontent eacch area. Doc served
s as an iin-game surveyy, questioning the
tthat is integral to play (“undeeniable”) than ppurely superficcial or plaayer on their affective
a state. He also led thhe player throuugh
aaesthetic (“denniable”). Finallyy, the experiencce of players ccan be sim
mple psychologgical tasks, such as countingg the number of
aaltered even beffore they start tthe game, by priiming them to eexpect vow wels in a senteence or identifyying the item in i a picture. Suuch
a fun or seriouss simulation for example [49]. tasks are frequenttly used in phyysiology experim ments to bring the
subbject's signal levvels back to basseline.
Overall, itt’s necessary to view a gam me as a gestalt,, with
vvisuals and othher stimuli, naarrative, mechaanics, characterrs and
ccontext all worrking in synerggy to maximizee the intensity of o the
uuser experiencce [56, 57, 69]. Much of the previous work
aadmirably attemmpts to codify thhe intuitions off designers, but iis still
nnot driven by ppsychological thheory. This is ppartly due to thhe fact
tthat for designners, most theorries have focussed on stimuluus and
rresponse whilee omitting useer experience. Furthermore, many
ppsychological theories have assumed that a specific stim mulus
iinvariantly causses a specific eemotion in all ppeople. This isnn't the
ccase - as designners intuitively know. Becausee CAT does nott have
tthe same limittations it can be used to infform design, aas we
ddemonstrate in the next sectionn.
44. NEW DESIGN
D AP
PPROACH
H USING CA
AT
We now describe a genneral method fo for applying innsights
ffrom the CAT to the creativve design proceess. To illustratte the
mmethod, we conncurrently descrribe how we appplied it to deveelop a
sshort video gaame that had the explicit purpose p of evvoking
ddifferent, robusst affective expperiences in plaayers. The gamee was Figgure 2. The fivee scenarios of the
t emotion-evvoking game.
ppart of a larger project to studdy individual diffferences in affe
fective We developed the game by implementiing the follow wing
eexperience andd was developedd jointly by afffective scientistts and meethodology for ddesigning affecttive experiencess. It is an iteratiive,
vvideo game reesearchers. Thee game was coonstructed usinng the theeory-driven proocess consistingg of two phasses: A) Affecttive
eengine, assets and
a editing toools provided byy Fallout New Vegas V Staates Definition aand B) Scenarioo Design and Im mplementation. At
((Bethesda Softw works). It conssisted of four sccenarios designned to eacch step the dessign is evaluateed and tuned based
b on multiiple
eelicit different affective statees, along with a recurring nneutral iterrations of inteernal experts’ feedback and external testeers’
sspace designed to allow playerrs to return to a relatively quieescent vallidation (Figuree 3). The proceess draws upon the three prim mary
sstate. The four affective states were chosen too sample the diffferent dessign implicatioons of the CA AT (Section 2..3): 1) individdual
qquadrants of thhe affective ciircumplex (see Figure 1). Sppecific varriability, 2) environmental conttext and 3) sequuence of events..
eemotions withinn each quadrantt were chosen as a target emotions for
eelicitation. Eachh scenario incluuded a task for the
t player to peerform
aand included timing constrainnts to make the game suitable under
rrestrictive expeerimental condittions (e.g., fMRRI). Navigation paths
aand player speed were tuneed so that eacch scenario toook a
mminimum of 900 seconds to coomplete, and a three t minute timmeout
pprovided an uppper bound in case the player ddid not compleete the
ttask. The four sscenarios were:
• The Fear Cave (F Figure 2A) waas a dark, omminous
eenvironment with
w threateninng giant inseects and rum
mbling
FFigure 3. Iterative process for affective sstate definition
n and afffective science research: IAPS S (International Affective Pictture
sscenario design
n, implementattion and tuningg. System) [44] andd IADS (Internaational Affectivve Digital Sounnds)
[144]. These are large sets of o standardized, internationaally
AA) Affecttive State Defi finition: In thee first phase oof the acccessible photoggraphs and sounnds that includde contents from m a
pprocess, the C Circumplex moodel (see Secttion 2) is useed to widde range of semmantic categoriees, developed tto provide a sett of
uunderstand andd define the affeective design sppace for the gam me. At norrmative emotioonal stimuli foor experimentall investigation of
tthe simplest leevel, the desireed emotions too evoke are simply em
motion. Since em motional imagerry is a continuum m of experiencee to
rrepresented as rregions in activvation-valence space,
s which neeed to whhich players willl adapt over timme, it is importtant to incorporrate
bbe sufficientlyy large to acccommodate inndividual variabbility. narrrative and audiiovisual variabillity.
HHowever, recallling the imporrtance of event sequence, emootions
aare best plottedd as trajectories across the spacce. In a normal ggame, External evaluation: As the game components and a
tthe story arc will
w define the trajectory (Figgure 4A) and can be levvels are transformed from storryboards to fullly-realized diggital
mmodified as apppropriate during this step. Forr our applicatioon, we envvironments, thrree types of evaaluation need too take place. Fiirst,
wwanted emotioons that were as unambigguous as posssible. thee designers andd/or affect psycchologists need to verify that the
TTherefore, for each scenario we w chose a single emotion tow wards connstructed enviroonments actuallly convey the ccorrect emotionn, at
tthe center of onne quadrant (Figgure 4B). Each scenario beginss after leaast for the maajority of obseervers. Only tthe full, dynam mic
aan interval in the neutral cennter space, trannsitions to the target envvironment conntains the conntext and evennt sequence tthat
eemotion, gradually increases inn intensity to coounteract habituuation crittically contribuute to emotionall impact. As an example from our
eeffects and provide variation, then transitions back to the nneutral owwn developmentt process, the team t agreed onn an initial cassino
ccenter to allow a “cool-down” establish a basseline before thee next moodel based on static screenshhots, but an eaarly demonstrattion
sscenario and collecting
c subjeective affective assessment ussing a revvealed certain ssubtle textures that detracted from the overrall
ffive point Likerrt scale for valennce and arousall. exccitement. The teextures were noot feasible to repplace, so we endded
up changing to a ddifferent casinoo model for the final version. T The
seccond and thirdd types of evaaluation are booth conducted by
extternal testers. Pllayability testinng is already fam
miliar to designeers,
butt is particularlyy important for an emotion-baased game becaause
tecchnical issues (e.g., unclear tasks or difficcult controls) ccan
fruustrate players aand completely obliterate any desired emotioonal
imppact. In additionn, given individdual variability, it’s vital to asssess
em
motional impactt on a numberr of people as well. Surveys or
phyysiological meaasurements cann be used; for our scenarios we
em
mployed both. More details oon our formattive and affecttive
evaaluations will bee given in Sectiion 4.1.
Previous worrk on emotionaal design (Sectioon 3.2), as welll as
FFigure 4. Trajjectories in acctivation-valence space developed
thee CAT theory itself, undersccore the necessity of a holisstic
dduring the Afffective State D Definition phasee. A, a typical story
appproach. That is,, a game and eaach contiguous segment withinn it
aarc: 1-expositiion, 2-crisis, 3--climax followeed by resolutioon, 4-
muust be seen as a whole: ambiennt sounds, music, light propertiies,
ddenouement. B,B the first halff of our affect ggame: 1-hub, 33-fear
collors, tasks to bbe performed, navigation pattterns and shap apes
ccave, 6-calm valley, 2,4,5,7-trransitions. Reppresentations oof our
colllectively contriibute to emotioon elicitation. T The iterative steps
ffour affect targgets are also in
ndicated.
of our design process help enssure that as many of the aboove
Experrt evaluation: tthe evaluation sstage assures thhat all prooperties as posssible are incorpoorated, and cann act in synergy,, as
iindividuals invoolved in the deesign process arre on the samee page thee game is built up from basicc emotional stattes. As mentionned
wwith regard to eemotional definnitions. Given inndividual variabbility, preeviously, meeeting the reqquirements im mposed by our
iit’s important tthat (e.g.) lead designers and level
l designers share expperimental ressearch setup meant a ceertain degree of
tthe same goalss. Technical lim mitations of thhe game enginee also com mpromise, in paarticular the sellf-contained sceenarios designedd to
nneed to be aaccounted for at this stagge, since a flawed fl alloow for randoom shuffling and isolate pparticular affeccts.
iimplementationn of an affectivee state is unlikelly to achieve itss goal. Hoowever, within each
e scenario w
we utilized the C CAT-based moddel,
IIn our case, we
w initially seleected eight emootional states sspread carrefully composiing the visuals and audio stim muli over time anda
oover the four quadrants
q of acttivation-valencee space, then ddrafted integrating them with story elements to achhieve the targeeted
sscenario outlinnes that targeteed each one. T The game desiigners afffective state results. Visual eleements can be gglimpsed in Figgure
eevaluated the scenarios
s for ddevelopment feaasibility in term ms of 2: tthe Fear Cave iis dark and sinisster, the Excitinng Casino is briight
ttechnology annd resources available, while w the affefective andd active, etc. AAudio elements also matched;; the Calm Vallley
ppsychologists ffocused on em motional impacct and focus. After feaatured slow-pacced, relaxing music
m while thee Sad Hotel hadd a
sseveral iteratioons, four final states were sselected: Exciteement muuch darker souundtrack and ppattering rain sounds. Dynam mic
((positive valennce, high aroussal), Calm (poositive valence, low feaatures were useed to counteract adaptation annd habituation; for
aarousal), Sadnness (negative valence, low arousal) and Fear exaample, the lighhting in the Feear Cave becom mes progressivvely
((negative valennce, high arousall). moore saturated annd shifts from a yellow orangee tint to red, whhile
BB) Scenaario Design annd Implementaation: in the seecond thee insects progreess from large ants to giant scorpions and the
pphase of the ovverall design proocess, the gamee is fleshed out based earrth progressively shakes and rumbles moree and more. T Two
oon the states prreviously identiified. The prelim
minary outliness from narrrative examplees are the Casinno, with the stoory arc Enter, Get G
tthe previous phhase are expannded into storybboards, then ussed to chiip, Play slot maachine, Win jackkpot, Prizes falll from ceiling, and
a
cconstruct fullyy-elaborated sscenarios. Apppropriate emootional thee Hotel, with arrc Enter, See sad people, Findd sick man, Brring
iimagery is speccified and incorrporated into thee narrative and game sicck man his diarry, Discover sick man is deadd before he haas a
eenvironment with
w the aid ofo audiovisual elements useed by chaance to see it.
5. FORMATIVE AND AFFECTIVE lab. During the post-play reports, all of the subjects gave
descriptive adjectives that almost exactly matched the
EVALUATION development targets. Thus "scary" or "creepy" were used to
In order to validate our four scenarios, we performed two describe the cave, "calm" or "relaxing" were used for the valley,
rounds of testing. The first battery of tests (with 3 subjects, two "exciting" or "fun" the casino, and "depressing" or "sad" the hotel.
females, aged 20 and 21 and one male aged 20) was intended as a Our goal is to use these scenarios in future research, modulating
formative evaluation to assess usability and playability of the both the targeted affect and its intensity to explore individual
experience. Testers were recruited among students enrolled in the differences in affective reactivity.
Game Design program at Northeastern University. After playing,
they were assessed as to how well they could form a mental map 6. CONCLUSIONS
of the locations and orient themselves in the designed worlds, as In this paper we discussed a theory driven approach to
well as whether they could perform the actions required by the develop interactive experiences, especially games, which evoke
scenario. They were also informally queried on their emotional affective responses from users. In particular, we argue for the
impressions. After testing, the design team made several holistic nature of designing emotional experiences, and thus
modifications. For example, the Fear Cave had proved particularly propose using the CAT as a psychological model of emotions.
difficult to navigate due to the dim lighting and slightly non- The CAT acknowledges that individual differences, situational
intuitive layout, so to avoid disorientation the floor plan was context, past experiences, mindset, and sequence of stimuli jointly
adjusted and unique light emitters were placed at key junctions. influence participants’ affect and behavior. Based on the model
For the second testing phase, our goal was to assess the we developed a generalized, systematic process for designing
effectiveness of the game at evoking affective responses. As game scenarios to evoke emotional experiences, and used it to
mentioned above, there is no biological signature of emotions, develop our own research tool. We hope this novel approach
though affect is much more reliably measured. Thus we used a facilitates a new perspective on theory-driven design and leads to
multi-measurement approach, with both psychophysiology and interactive experiences with more varied and vivid emotions
retrospective verbal reports. While playing the four scenarios within.
electrodermal activity (EDA) was measured as an index of
arousal. EDA and other physiological signals are often used to
7. REFERENCES
evaluate some aspect of a user’s experience during gameplay [43, [1] Allbeck, J. and Badler, N. (2002). Toward representing agent
47, 58]. The testers were three subjects (one female/ two male) behaviors modified by personality and emotion. Embodied
from the Interdisciplinary Affect Science Lab at Northeastern Conversational Agents at AAMAS, 2, 15-19.
University who were not familiar with the project and had a [2] Ambinder, M. (2011). Biofeedback in gameplay, GDC vault
variety of experience with video games. The subjects played the [3] Amaya, K., Bruderlin, A., and Calvert, T. (1996). Emotion
game in the physiology experiment space while their EDA was from Motion, Graphics Interface '96, pp. 222-229
recorded. After completing play, they were asked to report their
affective state during the game, questioned about their actions in- [4] Armony, J. L., Servan-Schreiber, D., Cohen, J. D., and
game and asked to describe the atmosphere of each scenario. LeDoux, J. E. (1997). Computational modeling of emotion:
Explorations through the anatomy and physiology of fear
Although three subjects is too few to provide a reliable conditioning. Trends in Cognitive Sciences, 1(1), 28-34.
analysis of EDA data, the qualitative trends were nonetheless
[5] Arya, A., Jefferies, L., Enns, J., and DiPaola S. (2006).
promising: the Fear Cave signals were consistently above
Facial Actions as Visual Cues for Personality, Computer
baseline, while the Calm Valley and Sad Hotel were consistently
Animation and Virtual Worlds (CAVW) Journal.
below. These findings are consistent with our goal of the Fear
Cave eliciting higher arousal and the latter scenario eliciting low [6] Barrett, L. F. (2014). The Conceptual Act Theory: A Précis.
arousal. The Exciting Casino was more variable, probably due to Emotion Review, 1–20.
the fact that two of the three subjects had trouble executing the [7] Barrett, L. F. (2006a). Emotions as natural kinds?
scenario task (we have since revised the in-game instructions) Perspectives on Psychological Science, 1, 28-58.
causing the onset of frustration that took precedence over any
other affective state. Individual differences in game play were also [8] Barrett, L. F. (2006b). Solving the emotion paradox:
visible in the data; the most striking example was a subject who Categorization and the experience of emotion. Personality
accidentally removed the clothes from the dead refugee in the Sad and Social Psychology Review, 10, 20–46.
Hotel and began laughing uncontrollably, yielding an EDA spike [9] Barrett, L. F., Lindquist, K., Bliss-Moreau, E., Duncan, S.,
that persisted even after the scenario ended. Gendron, M., Mize, J., and Brennan, L. (2007). Of mice and
men: Natural kinds of emotion in the mammalian brain?
Such examples illustrate that subjects who do not
Perspectives on Psychological Science, 2, 297-312.
successfully complete the task at hand are likely to report very
different emotional experiences. They also highlight the fact that [10] Bates, J. (1994). The role of emotion in believable agents.
aesthetic features alone are not sufficient to guarantee the desired Communications of the ACM, 37(7), 122-125.
affective state. The tasks to be performed and the action [11] Becker-Asano, C. and Wachsmuth, I. (2010). Affective
possibilities in each scenario are not just additional elements of computing with primary and secondary emotions in a virtual
the design that can be treated separately, but are a fundamental human. Autonomous Agents and Multi-Agent Systems,
layer of the whole experience. Additionally, the critical 20(1), 32-49.
importance of players’ interactions with their environment shows
how game scenarios rather than video or audio stimuli can achieve [12] Bellantoni, P. (2005). If it's purple, someone's gonna die: the
deeper emotional impact, an observation of particular interest to power of color in visual storytelling. Taylor and Francis.
affective scientists who wish to study powerful emotions in the
[13] Block, B. (2007). The visual story: creating the visual expressions are not culturally universal: Evidence from a
structure of film, TV and digital media. CRC Press. remote culture. Emotion, 14, 251-262.
[14] Bradley, M. M. and Lang, P. J. (1999). International affective [30] Gendron, M. and L. F. Barrett (2009). "Reconstructing the
digitized sounds (IADS): Stimuli, instruction manual and past: A century of ideas about emotion in psychology."
affective ratings (Tech. Rep. No. B-2). Gainesville, FL: The Emotion Reviews 1(4): 316-339.
Center for Research in Psychophysiology, University of [31] Gilbert, D. T., and Ebert, J. E. J. (2002). Decisions and
Florida. revisions: The affective forecasting of changeable outcomes.
[15] Bradley, M. M. and Lang, P. J. (2000). Measuring emotion: Journal of Personality and Social Psychology, 82, 503–514.
Behavior, feeling, and physiology. In R. D. Lane and L. [32] Gilbert, D. T., Pinel, E. C., Wilson, T. D., Blumberg, S. J.,
Nadel (Eds.), Cognitive neuroscience of emotion (pp. 242– and Wheatley, T. (1998). Immune neglect: A source of
276). New York: Oxford University Press. durability bias in affective forecasting. Journal of Personality
[16] Cacioppo, J. T., Berntson, G. G., Larsen, J. T., Poehlmann, and Social Psychology, 75, 617–638.
K. M., and Ito, T. A. (2000). The psychophysiology of [33] Gratch, J. and Marsella, S. (2001). Tears and fears: Modeling
emotion. In R. Lewis and J. M. Haviland-Jones (Eds.), The emotions and emotional behaviors in synthetic agents. In
handbook of emotion (2nd ed., pp. 173–191). New York: Proceedings of the fifth international conference on
Guilford Press. Autonomous agents (pp. 278-285). ACM.
[17] Chi, D., Costa, M., Zhao, L., and Badler, N. (2000). The [34] Gratch, J. (2003). Hollywood meets simulation: creating
EMOTE model for effort and shape, Proceedings of the 27th immersive training environments at the ICT. In Virtual
annual conference on Computer graphics and interactive Reality, 2003. Proceedings. IEEE (p. 303). IEEE.
techniques, p.173-182.
[35] Gratch, J. and Marsella, S. (2004). A domain-independent
[18] Davidson, R. J. (2000). Affective style, psychopathology, framework for modeling emotion. Cognitive Systems
and resilience: Brain mechanisms and plasticity. American Research, 5(4), 269-306.
Psychologist, 55, 1196–1214.
[36] Greene, J. D., Sommerville, R. B., Nystrom, L. E., Darley, J.
[19] Davidson, R. J., Pizzagalli, D., Nitschke, J. B., and Putnam, M., and Cohen, J. D. (2001). An fMRI investigation of
K. (2002). Depression: Perspec- tives from affective emotional engagement in moral judgment. Science, 293,
neuroscience. Annual Review of Psychology, 53, 545–574. 2105–2108.
[20] Diener, E. (1999). Introduction to the special section on the [37] Gross, J. J. and Barrett, L. F. (2011). "Emotion generation
structure of emotion. Journal of Personality and Social and emotion regulation: One or two depends on your point of
Psychology, 76, 803–804. view." Emotion Review 3(1): 8-16.
[21] Dias, J. and Paiva, A. (2005). Feeling and reasoning: A [38] Haidt, J. (2001). The emotional dog and its rational tail: A
computational model for emotional characters. In Progress in social intuitionist approach to moral judgment. Psychological
artificial intelligence (pp. 127-140). Springer Berlin Review, 108, 814–834.
Heidelberg.
[39] Hudlicka, E. (2009). Affective game engines: motivation and
[22] DiPaola, S. and Arya, A. (2006). Emotional Remapping of requirements. In Proceedings of the 4th International
Music to Facial Animation, ACM Siggraph ’06 Video Game Conference on Foundations of Digital Games (pp. 299-306).
Symposium Proceedings, 2006. ACM.
[23] DiPaola, S. (2013). Face, portrait, mask: using a [40] Hudlicka, E. and Broekens, J. (2009). Foundations for
parameterised system to explore synthetic face space. In modelling emotions in game characters: Modelling emotion
Electronic Visualisation in Arts and Culture (pp. 213-227). effects on cognition. In Affective Computing and Intelligent
Springer London. Interaction and Workshops, 2009. ACII 2009. 3rd
[24] Duncan, S. and Barrett, L. F. (2007). Affect as a form of International Conference on (pp. 1-6). IEEE.
cognition: A neurobiological analysis. Cognition and [41] Izard, C. E. (1977). Human emotions. New York: Plenum.
Emotion, 21, 1184–1211.
[42] Izard, C. E. (1993). Four systems for emotion activation:
[25] Ekman, P. (1973). Cross-cultural studies of facial expression. Cognitive and noncognitive processes. Psychological
In P. Ekman (Ed.), Darwin and facial expression: A century Review, 100, 68–90.
of research in review (pp. 169–222). New York: Academic.
[43] Kivikangas, J. M., Chanel, G., Cowley, B., Ekman, I.,
[26] Ekman, P. (1992). An argument for basic emotions. Salminen, M., Järvelä, S., and Ravaja, N. (2011). A review
Cognition and Emotion, 6, 169–200. of the use of psychophysiological methods in game research.
[27] El-Nasr, M. S., Morie, J., and Drachen, A. (2010). A Journal of Gaming and Virtual Worlds, 3(3), 181-199.
scientific look at the design of aesthetically and emotionally [44] Lang, P.J., Bradley, M.M., and Cuthbert, B.N. (2008).
engaging interactive entertainment experiences. Affective International affective picture system (IAPS): Affective
Computing and Interaction: Psychological, Cognitive and ratings of pictures and instruction manual. Technical Report
Neuroscientific Perspectives, 281-307. A-8. University of Florida, Gainesville, FL.
[28] Freeman, D. (2003). Creating Emotion in Games: The Art [45] LeDoux, J. E. (1996). The emotional brain: The mysterious
and Craft of Emotioneering. New Riders Publishing, underpinnings of emotional life. New York: Simon and
Thousand Oaks, CA, USA. Schuster.
[29] Gendron, M., Roberson, D., van der Vyver, J. M., and
Barrett, L. F (2014). Perceptions of emotion from facial
[46] Leino, O. T. (2007). Emotions about the SIGGRAPH/Eurographics symposium on Computer
Deniable/Undeniable: Sketch for a classification of game animation, San Diego, California
content as experienced. In Situated play, proceedings of digra [61] Neff, M. and Fiume, E. (2005). AER: aesthetic exploration
2007 conference. and refinement for expressive character animation,
[47] Lindley, C. A., Nacke, L., and Sennersten, C. C. (2008). Proceedings of the 2005 ACM SIGGRAPH/Eurographics
Dissecting Play - Investigating the Cognitive and Emotional symposium on Computer animation, Los Angeles, California
Motivations and Affects of Computer Gameplay. In [62] Niedenthal, S. (2008). Complicated shadows: The aesthetic
Proceedings of CGAMES. Wolverhampton, UK: University significance of simulated illumination in digital games.
of Wolverhampton. Blekinge Technical University.
[48] Lindquist, K. A., Wager, T. D., Kober, H., Bliss-Moreau, E., [63] Panksepp, J. (1998). Affective neuroscience: The
and Barrett, L. F. (2012). The brain basis of emotion: A foundations of human and animal emotions. New York:
meta-analytic review. Behavioral and Brain Sciences, 35, Oxford University Press.
121-143.
[64] Poels, K. and Dewitte, S. (2006). How to capture the heart?
[49] Luigi, D. P., Tortell, R., Morie, J., and Dozois, A. (2006). Reviewing 20 years of emotion measurement in advertising.
Effects of priming on behavior in virtual environments. DTEW-MO_0605, 1-47.
Retrieved August, 8, 2010.
[65] Russell, J. A. (2003). Core affect and the psychological
[50] Marsella, S., Gratch, J., and Rickel, J. (2004). Expressive construction of emotion. Psychological Review, 110, 145–
behaviors for virtual worlds. In Life-Like Characters (pp. 172.
317-360). Springer Berlin Heidelberg.
[66] Russell, J. A., Bachorowski, J., and Fernandez-Dols, J. M.
[51] Marsella, S., Gratch, J., and Petta, P. (2010). Computational (2003). Facial and vocal expressions of emotion. Annual
models of emotion. A Blueprint for Affective Computing-A Review of Psychology, 54, 329–349.
sourcebook and manual, 21-46.
[67] Russell, J. A. and Barrett, L. F. (1999). Core affect,
[52] Martinez, H. P., Jhala, A., and Yannakakis, G. N. (2009). prototypical emotional episodes, and other things called
Analyzing the impact of camera viewpoint on player emotion: Dissecting the elephant. Journal of Personality and
psychophysiology. In Affective Computing and Intelligent Social Psychology, 76, 805–819.
Interaction and Workshops, 2009. ACII 2009. 3rd
International Conference on (pp. 1-6). IEEE. [68] Scheutz, M. and Sloman, A. (2001). Affect and agent
control: Experiments with simple affective states.
[53] Mesquita, B. (2003). Emotions as dynamic cultural
phenomena. In H. Goldsmith, R. Davidson, and K. Scherer [69] Swartout, W., Hill, R., Gratch, J., Johnson, W. L.,
(Eds.), Handbook of the affective sciences (pp. 871–890). Kyriakakis, C., LaBore, C., and Moore, B. (2006). Toward
New York: Oxford. the holodeck: Integrating graphics, sound, character and
story. UNIVERSITY OF SOUTHERN CALIFORNIA
[54] Mirza-Babaei, P., Nacke, L. E., Gregory, J., Collins, N., and MARINA DEL REY CA INST FOR CREATIVE
Fitzpatrick, G. (2013). How does it play better?: exploring TECHNOLOGIES.
user testing and biometric storyboards in games user
research. In Proceedings of the SIGCHI Conference on [70] Totten, C. (2014). An Architectural Approach to Level
Human Factors in Computing Systems (pp. 1499-1508). Design. CRC Press.
ACM. [71] Wang, N. and Marsella, S. (2006), Introducing EVG: An
[55] McAllister, G., Mirza-Babaei, P., and Avent, J. (2013). Emotion Evoking Game. In 6th International Conference on
Improving Gameplay with Game Metrics and Player Metrics. Intelligent Virtual Agents, Marina del Rey, CA, 2006.
In M. S. El-Nasr, A. Drachen, and A. Canossa (Eds.), Game [72] Mateas, M. (1999). An Oz-centric review of interactive
analytics: Maximizing the value of player data. (pp. 621- drama and believable agents. In Artificial intelligence today
638). Springer London. (pp. 297-328). Springer Berlin Heidelberg.
[56] Morie, J. F., Iyer, K., Valanejad, K., Sadek, R., Miraglia, D., [73] Mateas, M. (2001). Expressive AI: A hybrid art and science
Milam, D., ... and Leshin, J. (2003). Sensory design for practice. Leonardo, 34(2), 147-153.
virtual environments. In ACM SIGGRAPH 2003 Sketches
[74] Hunicke, R., LeBlanc, M., and Zubek, R. (2004). MDA: A
and Applications (pp. 1-1). ACM.
formal approach to game design and game research, In
[57] Morie, J., Williams, J., Dozois, A., and Luigi, D. (2005). The Proceedings of the Challenges in Games AI Workshop,
Fidelity of “Feel”: Emotional Affordance in Virtual Nineteenth National Conference on Artificial Intelligence,
Environments 11th International Conference on Human- San Jose, CA.
Computer Interaction, Las Vegas, NV; July 2005.
[75] Kahrs, J., Calahan, S., Carson, D., Poster, S. Pixel
[58] Nacke, L. E., Grimshaw, M. N., and Lindley, C. A. (2010). Cinematography. Sigghraph 1996
More than a feeling: Measurement of sonic user experience
[76] Solarski, C. (2013) The Aesthetics of Game Art and Game
and psychophysiology in a first-person shooter game.
Design. Gamasutra, retrieved on 8/8/2016
Interacting with Computers, 22(5), 336-343.
http://www.gamasutra.com/view/feature/185676/the_aestheti
[59] Nauta, W. (1971). The problem of the frontal lobe: A cs_of_game_art_and_.php
reinterpretation. Journal of Psychiatric Research, 8, 167–187.
[77] Adams, E. (2013) Fundamentals of Game Design, New
[60] Neff, M. and Fiume, E. (2003). Aesthetic edits for character Riders Publishing
animation, Proceedings of the 2003 ACM
[78] Rigby, S., Ryan, R. M. (2011) Glued to Games: How Video [79] Ortony, A., Clore, G. L., Collins, A. (1990) The cognitive
Games Draw Us in and Hold Us Spellbound. Praeger structure of emotions. Cambridge university press
Publisher