=Paper=
{{Paper
|id=Vol-2609/AfCAI2019_paper_1
|storemode=property
|title=Influence of the Agent Personality on Its Mood
|pdfUrl=https://ceur-ws.org/Vol-2609/AfCAI2019_paper_1.pdf
|volume=Vol-2609
|authors=Joaquin Taverner, Emilio Vivancos, Vicente Botti, Bexy Alfonso
|dblpUrl=https://dblp.org/rec/conf/afcai/TavernerVBA19
}}
==Influence of the Agent Personality on Its Mood==
https://ceur-ws.org/Vol-2609/AfCAI2019_paper_1.pdf
Influence of the agent personality on its mood
Joaquin Taverner, Emilio Vivancos, Vicente Botti, and Bexy Alfonso
Valencian Research Institute for Artificial Intelligence (VRAIN)
Universitat Politècnica de València, Valencia, Spain
Abstract. When simulating emotions in an affective agent we must
include factors that can affect the agent emotional behavior such as mood
or personality. Personality can make a person more prone to feel certain
types of emotions and mood. In this work we present the preliminary
results obtained in an experiment carried out to find the factors that
affect human emotional behavior in order to simulate this behaviour in
an affective agent. We have focused on the effect of personality on the
differences between agents mood. Preliminary results show that there is a
correlation between the participants’ personality and their initial mood.
Keywords: personality, agent, behavior, emotion modelling.
1 Introduction
There are several theories that try to explain the human behavior. One of the
most relevant is the practical reasoning that is based on the idea that humans use
the reason to decide how to act [7]. But human behavior is not always rational.
This is due in part to emotions, mood and personality [12]. Many psychological
studies have shown that personality influences cognitive processes (e.g. decision
making process) as well and affective processes (e.g. making a person more or less
prone to feel certain moods) [2, 20, 23]. Therefore, in order to simulate human
behavior through an affective agent, it is necessary to study the influence of its
personality on the agent cognitive and affective processes [15, 16]. In this article
we show the preliminary results obtained in an experiment with 300 participants
conducted to develop an affective agent model in the GenIA3 architecture [1].
We have used a regression model to analyze the effect of personality on mood
represented using the PANAS model [21].
2 Background
When analyzing human behaviour from a rational point of view, we can observe
certain inconsistencies when facing situations like a decision-making problem.
These inconsistencies are due in part to the influence of emotions, mood and
personality [4, 6, 23]. From a psychological perspective, emotions can be defined
Copyright c 2019 for this paper by its authors. Use permitted under Creative Com-
mons License Attribution 4.0 International (CC BY 4.0).
as a rapid response to a stimulus while mood has a longer duration in time
than emotions (e.g. hours or days) [3]. In addition, mood is not related to any
particular stimulus, but rather it is the result of a succession of stimuli and
emotions. On the other hand, personality is an affective characteristic that does
not vary over time. Although the debate about what is personality is still open,
today it is accepted that personality is a set of psychic characteristics that involve
different cognitive processes and that produces differences in behavior between
individuals when facing a given situation. Personality influences the way in which
emotional responses to certain stimuli occur and the processes associated with
the emotion elicitation. Personality also refers to the characteristic way in which
a person thinks, feels, behaves, and relates to others [12].
Over the years different models have been proposed to represent different af-
fective characteristics such as personality, emotions and mood. One of the most
used models in psychology to define affection and personality is the PANAS
model [13]. This model uses two dimensions to represent personality: Positive
Affect (PA) and Negative Affect (NA) [21]. According to this model PA and NA
can be considered as two independent variables. For example, the state of relax-
ation corresponds to a low level of negative affect, while calm is conditioned to a
low positive affect. The positive affect reflects how enthusiastic, active, alert, and
energetic a person is. While the negative affect reflects how disgusted, disinter-
ested, angry, guilty, cowardly, and nervous a person is. Some moods are the result
of the combination of both variables. Happiness for example, is related to a high
level of positive affect and a low level of negative affect, while sadness is related
to low levels of positive affect and high levels of negative affect [19]. The PANAS
model can be used to analyze human personality using questionnaires in which
the questions are conducted generically [11]. For example, in the question “Are
you usually feel sad?”, the PA and NA factors can be interpreted as personality
values instead of mood because the question refers to the ”usual” emotional state
of the person. The simplicity of this model make it appropriate to be used in
modelling agents personality. There are other psychological models to represent
personality using quantitative variables such as the Five Factor Model [10] that
uses five variables to define personality: Openness, Consciousness, Extraversion,
Agreebleness, and Neuroticism.
Different studies have been carried out to relate emotions and personality.
For example, the trait of Extraversion is related to positive emotions and moods
while Neuroticism is related to negative emotions and moods [5, 14, 15, 20, 17].
Similarly, in the PANAS model the PA and NA variables related to personality
have a relationship with the PA and NA variables related to mood [22].
In recent years, different models have been proposed to simulate affective
human behavior [8, 9]. One of the most innovative proposals is GenIA3 , which
is a general purpose architecture for affective agents based on the BDI model
(beliefs, desires and intentions) [1]. This architecture allows to create agents
with affective characteristics in a simple way, providing an environment that
allow to develop agents that simulate human behavior. However, in order to
simulate this behavior, it is necessary to analyze the effect that factors such as
Table 1. Summary of the results obtained in the experiment.
Mean Sd
P PA 23.08 4.9256
P NA 11.14 6.2885
M PA 5.798 2.3522
M NA 2.281 1.8860
personality or mood have on it. The design of experiments to determine the
relationships between the different affective categories becomes necessary.
3 Proposal
Our main goal is to develop an affective agent capable of simulating human be-
havior through emotions, personality, and mood. However, as we argued before,
in order to develop this agent model it is necessary to study the relationships
between the different affective categories. As part of the development of our
affective agent model, in this work we analyze the relationship between person-
ality and mood. We have used the data obtained from an experiment [1] where
300 participants (159 female and 141 male) of different ages (between 18 and
50) played an adapted version of the Black Jack game. Each participant played
independently against the bank. The main goal of this experiment was to an-
alyze the variations of mood during the game and the impact of the mood on
decision making according to participants personality. At the beginning of the
experiment, participants were asked about their mood and personality using the
questionnaire proposed in [13]. This questionnaire is a Spanish adaptation of
the one proposed by Watson [18] using scales of positive and negative affect.
The questionnaire for the personality is composed by twenty questions: ten for
the personality positive affect (P PA) and ten for the personality negative affect
(P NA). Each question has five possible answers using a likert-scale where 1
means totally disagree and 5 means totally agree. In the same way, the question-
naire for the mood is composed of six questions using the same likert-scale: three
questions for the mood positive affect (M PA) and three for the mood negative
affect (M NA). Participants selected the response moving a slider. For this study
we have used a composite sample of those participants whose M PA and M NA
was greater than 1, that corresponds to the default value of the slider.
4 Results
Table 1 summarizes the statistics obtained in the experiment. The domain of
the P PA and P NA is [0, 40] corresponding to ten questions using a likert-scale
of five answers. As we can see the average of P PA is higher than the average of
P NA and the same happens with the variables M PA and M NA. In addition,
Table 2. Correlation matrix of PANAS values for personality and mood. *P-value less
than 0.05 significant.
P PA P NA M PA M NA
Correlation 1.0000
P PA
P-value 0.0000*
Correlation -0.0843 1.0000
P NA
P-value 0.1451 0.0000*
Correlation 0.6525 -0.14822 1.0000
M PA
P-value 0.0000* 0.0101* 0.0000*
Correlation -0.3192 0.3008 -0.3837 1.0000
M NA
P-value 0.0000* 0.0000* 0.0000* 0.0000*
Table 3. Linear regression model for M PA variable. *P-value less than 0.05 significant
Estimate Std. Error t value Pr(>|t|)
Intercept 1.0535 1.4063 0.749 0.455
P PA 0.2620 0.0387 6.767 7.48e-10*
P NA -0.0437 0.0303 -1.443 0.152
the standard deviation of the M NA variable is very small, indicating that the
data may be concentrated near the mean. This is better shown in Figure 1, in
which we can see that the data of the variable M NA is concentrated in low
values, while the variable M PA has a more normal distribution.
On the other hand, Table 2 shows the correlation between PANAS variables
for both personality and mood. As we can see, P PA and P NA have a moderate
correlation with M NA. While M PA variable has a high correlation with P PA
and a moderate correlation with P NA. We can also see that all correlations
are statistically significant except for P PA and P NA. In order to analyze how
personality affects mood we have performed a regression analysis on M PA and
M NA variables.
The result of the regression analysis for M PA variable is shown in Table
3. The only statistically significant variable in this regression model is P PA
since its p-value is less than 0.05. Figure 2 shows the statistical analysis of
the regression model. As we can see, the residues of the regression model fall
along the straight line at a 45% angle in the Q-Q plot which provide strong
evidence that these numbers come from an uniform distribution. The residual
graph (Figure 2) shows that the variation of the residues is constant, so we can
assume that the assumption of linearity is met. Therefore we can assume that
the regression model is valid for the M PA variable.
The r-squared value is 0.2931 therefore this regression model explains the
29.31% of the variability of the variable M PA. Although it is a low percent-
14
12
10
8
6
4
M_PA
1 M_NA
2
Fig. 1. Boxplot of mood variables.
Residuals vs Fitted Normal Q−Q
4
2
Standardized residuals
2
1
Residuals
0
0
−2
−1
−4
−2
1046 1046
256 256
−6
1271 1271
5 6 7 8 9 10 11 12 −2 −1 0 1 2
Fitted values Theoretical Quantiles
Scale−Location Residuals vs Leverage
1271
256
1.5
2
1046 851
Standardized residuals
Standardized residuals
1
1.0
0
−1
0.5
Cook's
76
distance
−2
−3
0.0
1271
5 6 7 8 9 10 11 12 0.00 0.02 0.04 0.06 0.08 0.10
Fitted values Leverage
Fig. 2. Analysis of the linear regression model for M PA variable.
age, we must consider the variability of the data when performing this type of
psychological experiments. Therefore, both a r-squared of thirty percent can be
a good value if we consider that we have only used personality to explain the
differences between participants’ initial mood. Therefore, each affective agent of
our model will calculate its initial mood considering that thirty percent of its
M PA depend on its P PA value.
For the prediction of M NA value, as we mentioned above, the data is asym-
metric and with low dispersion, which makes it difficult to apply a linear regres-
sion algorithm. We have tried to apply different transformations of the data and
we have not found any linear regression model that met the normality conditions.
5 Conclusion
In this work we have presented a preliminary analysis of the influence of person-
ality on mood using the data obtained in an experiment with 300 participants.
This analysis allow us to obtain a model that partially explain the relation-
ship between mood and personality and emotional behavior. We are using this
model to develop a multi-agent system capable of simulating human behavior in
a realistic way.
We have shown that the P PA variable from the PANAS model has a posi-
tive correlation with the variable M PA, which indicates that there is influence
of personality on mood. However, the variable M NA has an asymmetric distri-
bution, which has made it difficult to find a linear regression model to determine
the influence of personality on this variable. Nevertheless, in this work we have
shown the importance of personality in mood, being able to explain almost thirty
percent of the variability of the variable M PA.
We are currently analyzing in deep all the results of the experiment in order
to find a model that allows us to identify the proportion of variance of the NA
variable that can be explained by personality. Once these models are obtained,
we want to implement a multi-agent system within the GenIA3 architecture that
allows us to simulate human behavior taking into account different personalities.
Acknowledgements
This work is partially supported by the Spanish Government project TIN2017-
89156-R, FPI grant ACIF/2017/085, and GVA-CEICE project PROMETEO/
2018/002.
References
1. Bexy Alfonso. Agents with Affective Traits for Decision-Making in Com-
plex Environments. PhD thesis, Universitat Politécnica de Valencia, Spain.
http://hdl.handle.net/10251/90497, 2017.
2. Mark S Allen, Sylvain Laborde, and Emma E Walter. Health-related behavior
mediates the association between personality and memory performance in older
adults. Journal of Applied Gerontology, First publication March 2017, 2017.
3. P Becker. Structural and relational analyses of emotions and personality traits.
Zeitschrift fur Differentielle und Diagnostische Psychologie, 22(3):155–172, 2001.
4. John Broome. Practical reasoning. Reason and nature: Essays in the theory of
rationality, pages 85–111, 2002.
5. Douglas Derryberry and Marjorie A Reed. Temperament and attention: Orienting
toward and away from positive and negative signals. Journal of personality and
social psychology, 66(6):1128–1139, 1994.
6. Nico H Frijda. The emotions: Studies in emotion and social interaction. Paris:
Maison de Sciences de l’Homme, 1986.
7. Brian R Gaines. Modeling practical reasoning. International journal of intelligent
systems, 8(1):51–70, 1993.
8. Patrick Gebhard. Alma: a layered model of affect. In Proc. of the fourth inter-
national joint conference on Autonomous agents and multiagent systems, pages
29–36, 2005.
9. Stacy C Marsella and Jonathan Gratch. EMA: A process model of appraisal dy-
namics. Cognitive Systems Research, 10(1):70–90, 2009.
10. Robert R McCrae and Oliver P John. An introduction to the five-factor model
and its applications. Journal of personality, 60(2):175–215, 1992.
11. Albert Mehrabian. Comparison of the PAD and PANAS as models for describing
emotions and for differentiating anxiety from depression. Journal of psychopathol-
ogy and behavioral assessment, 19(4):331–357, 1997.
12. Andrew Ortony, Gerald L Clore, and Allan Collins. The cognitive structure of
emotions. Cambridge University Press, 1990.
13. Bonifacio Sandı́n, Paloma Chorot, Lourdes Lostao, Thomas E Joiner, Miguel A
Santed, and Rosa M Valiente. Escalas PANAS de afecto positivo y negativo:
validación factorial y convergencia transcultural. Psicothema, 11(1):37–51, 1999.
14. Joaquı́n Taverner, Bexy Alfonso, Emilio Vivancos, and Vicent J Botti. Modeling
personality in the affective agent architecture GenIA3. In International Conference
on Agents and Artificial Intelligence (ICAART), pages 236–243, 2018.
15. Joaquı́n Taverner, Bexy Alfonso, Emilio Vivancos, and Vicent J. Botti. Using dif-
ferent agent personalities to obtain different agent behaviours. In Proceedings of the
Workshop on Affective Computing and Context Awareness in Ambient Intelligence
(AfCAI 2018), 2018.
16. Joaquı́n Taverner, Bexy Alfonso, Emilio Vivancos, and Vicente J. Botti. Integrat-
ing expectations into Jason for appraisal in emotion modeling. In Proc. of the
8th International Joint Conference on Computational Intelligence, IJCCI 2016,
Volume 1: ECTA, pages 231–238, 2016.
17. Joaquı́n Taverner, Emilio Vivancos, and Vicente J. Botti. Towards a computational
approach to emotion elicitation in affective agents. In International Conference on
Agents and Artificial Intelligence (ICAART), 2019.
18. David Watson. Intraindividual and interindividual analyses of positive and nega-
tive affect: their relation to health complaints, perceived stress, and daily activities.
Journal of personality and social psychology, 54(6):1020, 1988.
19. David Watson, Lee A Clark, and Auke Tellegen. Development and validation
of brief measures of positive and negative affect: the PANAS scales. Journal of
personality and social psychology, 54(6):1063, 1988.
20. David Watson and Kristin Naragon-Gainey. Personality, emotions, and the emo-
tional disorders. Clinical Psychological Science, 2(4):422–442, 2014.
21. David Watson and Auke Tellegen. Toward a consensual structure of mood. Psy-
chological bulletin, 98(2):219, 1985.
22. Kay Wilson and Eleonora Gullone. The relationship between personality and affect
over the lifespan. Personality and Individual Differences, 27(6):1141–1156, 1999.
23. John M Zelenski. The role of personality in emotion, judgment and decision mak-
ing. In Vohs K, Baumeister R, Loewenstein G, Do Emotions Help or Hurt Decision
Making, pages 117–132, 2007.