See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/309731046

Monitoring the Impact of Negative Events and Deciding About Emotion

Regulation Strategies

Conference Paper in Lecture Notes in Computer Science · December 2016

DOI: 10.1007/978-3-319-59294-7_30

CITATIONS READS

8 216

3 authors:

Adnan Manzoor Altaf Hussain Abro

Quaid-e-Awam University of Engineering, Science and Technology University of Sindh
26 PUBLICATIONS 167 CITATIONS 17 PUBLICATIONS 53 CITATIONS

SEE PROFILE SEE PROFILE

Jan Treur
Vrije Universiteit Amsterdam
1,300 PUBLICATIONS 11,453 CITATIONS

SEE PROFILE

All content following this page was uploaded by Jan Treur on 07 November 2016.

The user has requested enhancement of the downloaded file.

 Monitoring the Impact of Negative Events
and Deciding about Emotion Regulation Strategies

Manzoor, A., Abro, A.H., Treur, J.

Vrije Universiteit Amsterdam, Behavioural Informatics Group

De Boelelaan 1081, 1081 HV Amsterdam, the Netherlands
{a.manzoorrajper, a.h.abro, j.treur}@vu.nl

Abstract. Humans have a number of emotion regulation strategies at their

disposal, from which in a particular situation one or more can be chosen. The
focus of this paper is on the processes behind the choice of these regulation
strategies. The paper presents a neurologically inspired cognitive computational
model of a monitoring and decision mechanism for emotion regulation
incorporating different strategies (expressive suppression, reappraisal or
reinterpretation, and situation modification). It can be tuned to specific
characteristics of persons and events.

Keywords: Cognitive modeling, Emotion, Regulation.

1 Introduction

Emotions play a vital role for a person to function responsibly in society. Proper han-
dling of negative emotions such as stress and anxiety help us to perform our daily life
activities in an efficient manner, and not become vulnerable to stress-related disorders
such as depression or PTDS. It has been found that individuals can apply different
emotion regulation strategies [1]. Several types of emotion regulation strategies exist
which can be effective in particular circumstances. Two of them which have received
much interest of researchers over the years are reappraisal and emotion suppression
[2]. An important but often neglected part of the emotion regulation process is a deci-
sion making process determining under which circumstances different strategies are
selected [3]. Which strategy is applied depends on a number of factors, such as a per-
son’s context, an internal monitoring and assessment concerning her feeling intensity,
and her individual characteristics or preferences. Empirical studies such as [4] show
that individual differences exist when it comes to prefer one strategy over another and
also these differences exist when some individual applies a combination of emotion
regulation strategies.
In this paper the role of monitoring and assessment, and control mechanisms to
recognize a type of negative emotion and to choose for one or more strategies are
explored computationally. The first process acts as an identification stage as described
in [3][4][5] which recognizes and assesses the negative feelings and their intensity.
Based on this assessment one or more control states are activated for specific emotion
regulation strategies. For example, if the intensity of an emotion is very high, then
multiple regulation strategies might be employed at the same time (which also de-
pends on the personality traits). On the other hand if the intensity is very low, then
only emotion suppression could be enough to be applied or if it is of a moderate level,
then it could be the case that only appraisal and emotion suppression are chosen. Sev-
eral simulation experiments that have been realized show how the model can take into
account different kinds of personalities and varying levels of negative stimuli and
feelings.

2 Neurological Background

When emotional responses compete with important goals or with socially more ap-
propriate responses, often regulation of them takes place [6][7]. Emotion regulation
can make use of a variety of specific strategies to affect the emotion response levels
[8]. Emotion regulation uses control functions in order to activate one or more of the
different strategies to generate, maintain and adjust the emotional responses [9]. By
such emotion regulation mechanisms, persons have the ability to suppress negative
influences from the environment and maintain a form of emotional homeostasis
[10][11]. Emotions can be regulated in different stages of the emotion generation
process[10][11][12] distinguish antecedent-focused strategies (those that address pro-
cesses before an emotion has an effect on the behavior) from response-focused strate-
gies (those that are used when the emotional response is already coming as expression
or behavior). Note that the different types of emotion regulation share a common
effect on the level of emotion, but may differ much in the path followed to achieve
this effect. Moreover, multiple strategies can be used at the same time, so that multi-
ple paths are followed in parallel with a combined effect on the emotion level.
The current paper focuses on the monitoring and control for three different emotion
regulation strategies: (1) situation modification (2) reinterpretation, and (3) expressive
suppression[11][13]. Here the first two are antecedent-focused strategies and the third
is a response-focused strategy. Situation modification [11] addresses the very first
part of the causal chain from trigger to emotion, namely the external trigger itself by
performing actions that change the external situation in such so that the trigger be-
comes more harmless. Reinterpretation works by changing the assigned meaning or
interpretation of an emotional stimulus in a way that changes its emotional impact
[14]. Expressive suppression is a form of response modulation that involves inhibiting
ongoing emotion-expressive behavior [11].
The model presented here was inspired by a number of neurological theories relat-
ing to fMRI experiments. Much emphasis has been put in the literature on the role
that is played by a bidirectional interaction between the amygdala and the prefrontal
cortex (PFC). In experiments often fMRI measurements have been made focusing on
activity in these brain areas, and anatomically their connections have been analysed.
For both, correlations have been found with (the extent of success in) actual emotion
regulation; e.g., [15][16][17]. For example, it has been found that less interaction or
weak connections between amygdala and prefrontal cortex lead to less adequate emo-
tion regulation [7]. The general idea is that upward interaction from amygdala to PFC
can have the function of monitoring, in order to get an internal representation of the
level of emotion within the prefrontal cortex, which is used to achieve a form of as-
sessment of this level of emotion within the prefrontal cortex, whereas the downward
interaction from PFC to amygdala makes it possible to control and modify amygdala
activation. In the process of monitoring and assessing the level of emotion, leading to
PFC activity, interaction with some areas other than the amygdala may occur as well,
as these areas can also play an important role in developing emotions and feelings.
So, upward interactions can be considered from multiple areas. Also in relation to
the control function of the PFC and connections from there to other areas some differ-
entiation is needed. For different regulation strategies different brain areas need to be
affected. For a response-focused strategy such as expressive suppression, maybe a
main effect can be to suppress amygdala activation in a more direct manner, but may-
be also other areas involved in actual expression of the emotion have to be sup-
pressed. Furthermore, for an antecedent-focused strategy such as reinterpretation it is
quite plausible that the control from the PFC has to affect the interpretation, and not
the amygdala in a more direct manner. For example, in this case the PFC may affect
(working) memory in order to achieve the reinterpretation. After this reinterpretation
has been accomplished, in turn the renewed emotion generation process (based on the
new interpretation) will affect the emotion level, including amygdala activity. In such
a case a more direct suppression of amygdala activation might still take place as well,
but then that effect may have to be attributed to a different regulation strategy which
occurs in parallel, for example, expressive suppression.
After all, it is also a matter of clear definition to distinguish the different strate-
gies. For example, it may be tempting to define the reinterpretation strategy in such a
manner that it also includes the expressive suppression strategy, given empirical data
that may have difficulty to distinguish the two. However, from a conceptual perspec-
tive it is more useful to define the two strategies as exclusive so that different paths
can be attributed to different strategies, even if they occur in the same experiment.
The latter choice is made in this paper. To control different pathways in order to
achieve emotion regulation according to different strategies the PFC has to involve
different areas within the brain. In some recent studies such as [18][15] attempts are
made to relate different regulation strategies to activity in different brain areas. See,
for example, [15] which describes that expressive suppression relates to an increase of
brain activation in a right prefronto-parietal regulation network, and reinterpretation
engages a different control network comprising left ventrolateral prefrontal cortex and
orbitofrontal cortex.

3 The Cognitive Model

The computational model was designed as a temporal-causal network model; see [19].
An overview of the states and causal relations of the proposed model is depicted in
Fig. 1. A description of each state is available in Table 1. The states of this model can
be classified in six groups: the environment, emotion generation, emotion regulation
selection strategy 1, 2 and 3, and, last but not least, an internal monitoring and selec-
tion mechanism for the decision making. The monitoring process is modelled by the
connections from the feeling state fsb to a number of monitoring states msi (which can
be any number but in the simulations has been chosen as 3), and the selection process
is modelled by the connections of the monitoring states to the control states. The up-
ward connections model the connections from amygdala to PFC that are used for
monitoring the lower level processes in the brain (see Section 2). If the feeling inten-
sity reaches at a certain threshold (which may differ for different kinds of persons),
the monitoring system reflects this by activating some of the monitoring states. In
addition, by some inhibiting connections between them the monitoring states are
made mutually exclusive and recognize specific types of stressful situations: monitor-
ing state ms1 recognizes low intensity feeling, ms2 recognizes moderate level feeling
and ms3 high intensity negative feeling. This inhibition-based process between the
monitoring states can be considered as a form of assessment, leading to one unique
indication of the situation concerning the stress level.

Fig. 1. Conceptual representation of the computational model

This single monitoring state obtained is the basis for a form of decision, by acti-
vating one or more control states for specific regulation strategies. A person’s charac-
teristics for these monitoring and decision processes are represented by the weights of
the connections to the monitoring states and from the monitoring states to the control
states, respectively. The selection process involves the three emotion regulation strat-
egies covered here. Depending on the situation and personality of an individual, one,
two or all of these regulation strategies are selected. For example, if the feeling is
intense then situation modification may be chosen by the person, depending on her
characteristics.
The main states representing the environment are wsw and wse. Here wsw indicates
the person’s environment state and wse covers external events which may affect the
environment of the person. The state of the world is sensed by the person via sensor
state ssw and represented by state srsw.

Table 1. Overview of the states of the proposed model (see also Fig. 1)

Domain Formal Informal name Description

wsw World state w This characterizes the current world situation which
Environ-
ment

the person is facing

wse World event e Circumstances in the world affecting the world situa-
tion in a stress-inducing way
ssw Sensor state for w The person senses the world through the sensor state,
providing sensory input
srsw Sensory representa- Internal representation of sensory world information
tion of w on w
srsb Sensory representa- The person maintains a body representation srsb for b
tion of b in the brain. Here b is embodying the associated
Generation
Emotion

fsb Feeling associated emotion, in the considered scenarios a negative emo-

to body state b tion. Before performing an action, a feeling state fsb
for the action is generated by a predictive as-if body
loop, via the sensory representation state srsb.
bsc Belief state for c Interpretation of the world information; in the case of
different, exclusive interpretations for the same world
information, they may suppress each other
psb Preparation for b Preparation for a response involving body state b
Reappraisal Re-interpretation of world information by belief change: changing the
assigned meaning to a stimulus with negative emotional effects (e.g., by believing
Regulation
Strategy 1
Emotion

that a noisy restaurant will become more quiet soon).

cs1b,c Control state for reap- By becoming activated this control state sup-
praisal of belief c to presses the belief for c, which gives the oppor-
avoid feeling b tunity for alternative beliefs to become dominant.
Suppression of emotion-expressive behaviour Inhibition of the expression, for
example, hide one’s true feelings from another person (e.g., hiding one’s fear when
standing up to a bully).
cs2b Control state for ex- By becoming activated this control state sup-
Regulation
Strategy 2
Emotion

pressive suppression to presses the execution state for b.

avoid feeling b
ssb Sensing body state b To maintain the body representation srsb for b, the
person senses the body state b.
esb Execution state for b Body expression of b, for example a fear expres-
sion
 Situation modification For this strategy the person performs an action in the exter-
nal world to change a situation which triggers negative emotions into a better one
(e.g., leaving a noisy restaurant and enter a quiet place).
Regulation
Strategy 3
Emotion

cs3b,a Control state for situa- By becoming activated this control state activates
tion modification a to the preparation and execution of action a to
avoid feeling b change the situation.
psa Preparation for action a Preparation to modify the situation by action a
esa Execution state for The action a is changing the situation (decreasing
action a the level of world state w)
ms1 Recognizes low feeling The monitoring states are involved in two
and Selection processes

level processes, one which is responsible for monitor-

Monitoring

ms2 Recognizes moderate ing of the feeling (connections to the monitoring

feeling level states) and reaching a form of assessment (by
ms3 Recognizes high feel- some inhibiting links between them), and the
ing level second process is concerned with the selection of
the appropriate regulation strategies (connections
from the monitoring states to the control states).

This sensory information can be interpreted by both a positive belief bsc1 and a
negative belief bsc2, which represent two different interpretations of the same world
condition. These conflicting beliefs compete with each other by mutual inhibiting
connections. In the considered scenario, the negative belief bsc1 has an effect on the
state of preparation for negative emotional response psb which leads to sensory body
representation srsb and to the negative feeling fsb. Subsequently, fsb has an impact on
the preparation state psb, which in turn has an impact on feeling state, fsb, through srsb
which makes the process recursive; this is often called an as-if body loop in the litera-
ture (e.g. [20]). Other states, depicted in Fig. 1, are control states related to three emo-
tion regulation strategies described below.
As described in Section 2, emotions can be controlled in different phases of the
process during which emotions are generated [11]. The first strategy discussed focus-
es on reinterpretation of the world information by changing bad beliefs about the
situation into more positive ones; this is done as follows. Suppose two beliefs bsc1 and
bsc2 are two different, exclusive interpretations of the world state, where bsc2 associ-
ates to bad feelings fsb. The exclusiveness is modelled by mutual inhibiting connec-
tions. Suppose the person has generated belief state bsc2 as dominant, and by her
monitoring and decision mechanism she decides for activation of control state cs1b,c.
Consequently this control state weakens the belief bsc2 and due to this, the positive
belief bsc1 can become dominant, which provides an alternative, more positive inter-
pretation of the world. Also expressive suppression can be used to decrease negative
emotions. In the model, when it is decided to activate control state cs2b for this second
strategy, this suppresses the expression of the emotional response esb. This esb is
sensed by the person him or herself through the body loop, and through that it has a
decreasing effect on the emotion level. The third emotion regulation strategy consid-
ered is situation modification. Leaving an annoying place or person is an example of
this strategy. In the model the control state for this kind of emotion regulation is
cs3b,a. A decision to activate this control state leads to preparing and performing an
action a (i.e., states psa and esa) which can change the situation (characterized by
wsw), for example walking away from a noisy place to a quiet place.
The conceptual representation of the model is represented as a number of states and
connections between them, shown in Fig. 1 and verbally in Table 1, with in addition:

• for each connection from state X to state Y a weight ωX.,Y (a number between -1
and 1), for the strength of the impact through this connection; a negative weight is
used for suppression
• For each state Y a speed factor ηY (a positive value) and (a reference to) a standard
combination function cY(…) used to aggregate multiple impacts from different
states on one state Y
For a numerical representation of the model the states Y get activation values
indicated by Y(t): real numbers between 0 and 1 over time points t, where the time
variable t ranges over the real numbers. More specifically, the conceptual
representation of the model (as shown graphically in Fig. 1 and verbally in Table 1)
can be transformed in a systematic or even automated manner into a numerical
representation as follows [19]:

• At each time point t each state X connected to state Y has an impact on Y defined as
impactX,Y(t) = ωX,Y X(t) where ωX,Y is the weight of the connection from X to Y
• The aggregated impact of multiple states Xi on Y at t is determined using a combi-
nation function cY(..):
aggimpactY(t) = cY(impactX1,Y(t), …, impactXk,Y(t)) (1)

= cY(ωX1,YX1(t), …, ωXk,YXk(t))

where Xi are the states with connections to state Y

• The effect of aggimpactY(t) on Y is exerted over time gradually, depending on

speed factor ηY:

Y(t + ∆t) = Y(t) + ηY [aggimpactY(t) - Y(t)] ∆t (2)

or dY(t)/dt = ηY [aggimpactY(t) - Y(t)]

• Thus the following difference and differential equation for Y are obtained:

Y(t + ∆t) = Y(t) + ηY [cY(ωX1,YX1(t), …, ωXk,YXk(t)) - Y(t)] ∆t (3)

or dY(t)/dt = ηY [cY(ωX1,YX1(t), …, ωXk,YXk(t)) - Y(t)]

For all states for the standard combination function either the identity function id(..) or
the advanced logistic sum combination function alogisticσ,τ(…) is used [19]:
cY(V) = id(V) = V
1 1
cY(V1, …Vk) = alogisticσ,τ(V1, …, Vk) = ( - ) (1 + e−στ ) (4)
1+ e−σ( 𝑉1+ …+ 𝑉k − τ) 1+ eστ

Here σ is a steepness parameter and τ a threshold parameter. The advanced logistic

sum combination function has the property that activation levels 0 are mapped to 0
and it keeps values below 1. The identity function id(..) is used for the states with a
single impact: ssw, ssb. For all other states the advanced logistic sum combination
function is used. For example, for the feeling state fsb the model is numerically
represented in difference equation form as

aggimpactfsb(t) = alogisticσ,τ(ωsrsb,fsb srsb(t), ωcs1b,c,fsb cs1b,c(t), ωcs2b,fsb cs2b(t) ωcs3b,a,fsb

cs3b,a(t))

fsb(t+∆t) = fsb(t) + ηfsb [aggimpactfsb(t) - fsb(t)] ∆t (5)

In this way the model represented conceptually in Fig. 1 is transformed into a

numerical representation of the model in terms of difference or differential equations.
The simulations are performed by applying a computational simulation method to this
numerical model representation, in a dedicated software environment. All the
simulations were performed within the MATLAB™ environment.

4 Scenarios and Simulation results

The computational model presented above has been used to perform number of
simulation experiments addressing the selection of emotion regulation strategies. This
has been done for different scenarios describing different cases with different levels
of stimulus and negative feeling, and varying from selection of just one of the
regulation strategies to selecting multiple regulation strategies at the same time.
Scenarios also vary on certain characteristics of the person, such as: sensitivity of a
person for negative stimuli, and a person’s preferences for regulation strategies
For example, some persons may have a higher preference for the situation
modification strategy (e.g., they tend to try to escape from a disturbing situation),
maybe in combination with a high sensitivity for disturbing stimuli, whereas other
types of persons may prefer the other regulation strategies while staying in the same
situation. More specifically, some persons are good in suppression of their negative
feelings related to a stimulus and keep the same interpretation and stay in the same
situation, whereas other persons may prefer to try to reinterpret (reappraise) the
situation in a more positive way by changing the negative beliefs about the situation
into positive beliefs, in oreder to reduce the level of negative feelings. Another
category of persons may be quite sensitive to the stimulus and initially try to reduce
their level of negative feelings by suppression and may use the reinterpretation
strategy to make their positive beliefs more stronger against the negative beliefs, and
if they fail to do so they still may try to escape from the bad situation or try to modify
the situation in another way.
 The simulation experiments demonstrate the role of the monitoring, assesment
and decision making with an important role for the monitoring states, which are used
as a basis to select one or more of the three available regulation strategies. The
selection process starts when a monitoring state reflects that a certain type (level) of
negative feeling arises. In a very first stage just a low level of negative feeling fsb
triggers monitoring state ms1 (indicating a low level of feeling), which in turn may
lead to a decision to activate one or more regulation strategies preferred by the person
for such a low level of negative feeling (recall that these preferences are represented
by the weights of the connections from the monitoring state to the three control
states). Then there are two possibilities: these strategies are adequate and limit the
feeling level, or the feeling level still increases so that monitoring state ms2 (indicating
a mediate level of feeling) is triggered. In the latter case this monitoring state ms2 in
turn may lead to a decision to activate another selection of regulation strategies.
Again there are two possibilities: these strategies may limit the feeling level, or the
feeling level still increases to the situation that monitoring state ms3 (indicating a high
level of feeling) is triggered. In the latter case again another selection of regulation
strategies can be decided for. A specific case of such a scenario is shown in Table 2.
In the scenario indicated in this table the first regulation strategy used (after ms1 is
triggered) is suppression of the negative feeling.

Table 2. Regulation selection choices for an example scenario

Feeling Triggered Selected regulation

level monitoring state strategies
low ms1 suppression
median ms2 reappraisal
high ms3 situation modification

This means that based on ms1 it is decided to activate control state cs2b in order
to suppress the negative feeling. In a second stage, when the level of negative feeling
increases further, due to the development of negative beliefs about the situation, this
triggers the next monitoring state ms2, and based on that it is decided to activate con-
trol state cs1b,c for the second regulation strategy: reappraisal (reinterpretation). This
starts to down-regulate the negative feelings in a different way by changing (reinter-
preting) the meaning of the stimulus (switching of a negative belief to a positive be-
lief). The control state cs1b,c is usually slower compared to cs2b, because humans
often take much time to change their beliefs about the environment (stimulus), so it
takes some more time to change beliefs. The third and last monitoring state ms3 trig-
gers when the level of feeling becomes high; then based on this it is decided to acti-
vate the third control state cs3b,a which initiates situation modification by performing
the (physical) action needed to achieve that. As this situation modification strategy
involves movement, it is slower and takes some more time compared to two other
regulation strategies mentioned above which involve mental processes instead of
physical action.
Note that in Table 2 for each monitoring state exactly one regulation strategy is
selected. However, it is also possible that the strategies selected for a lower level of
the feeling are still selected as well for higher levels of the feeling. The more specific
simulation results discussed here are based on the following scenario. The person is in
a restaurant which has become rather noisy, and this triggers negative feelings. First
she suppresses these negative feelings. Moreover, she tries to suppress her negative
belief about being in a noisy restaurant the whole evening to give space for a positive
belief (it will soon become more quiet). However still some negative feeling remains.
Therefore she decides to leave the restaurant. The simulation executes for 120 time
points with ∆t = 0.1. Details of the values for parameters used in the simulation are
given in Table 3 (threshold τ, steepness σ, and update speed η) and in Table 4 (con-
nection weights between all states).

Table 3. Values of threshhold, steepness and update speed

state τ σ η state τ σ η
ms1 0.08 50 6 esb 0.5 4 6
ms2 0.32 50 6 cs2b 2 5 6
ms3 0.6 50 6 ssb 0.5 4 6
psb 0.4 4 6 wsw 0.1 5 0.4
fsb 0.1 4 6 ssw 0.2 4 6
bsc1 0.1 8 6 srsw 0.2 4 6
bsc2 0.36 15 6 psa 0.4 5 6
cs1b,c 1.5 15 0.5 esa 0.5 100 6
srsb 0.2 3 6 cs3b,a 1.2 5 0.1

Table 4. Values of parameters used: connection weights

weight weight weight weight

ωesa,wsw -1 ωsrsb,fsb 0.9 ωcs1b,c,fsb -0.1 ωpsb,cs2b 0.8
ωwsw,ssw 0.5 ωesb,cs2b 0.8 ωcs2b,fsb -0.2 ωcs3b,a,esa 0.8
ωssw,srsw 0.9 ωsrsw,psa 0.1 ωcs3b,a,fsb -0.3 ωpsa,esa 0.7
ωsrsw,bsc1 0.3 ωcs3b,a,psa 1 ωbsc2,psb 0.7 ωsrsw,cs3b,a 0.8
ωcs1b,c,bsc1 0.0 ωfsb,cs3b,a 0.3 ωfsb,psb 0.7 ωms1,cs1b,c 1
ωbsc1,bsc2 -0.2 ωfsb,ms1 0.9 ωcs2b,psb -0.2 ωms1,cs2b 1
ωsrsw,bsc2 0.9 ωfsb,ms2 0.9 ωpsb,srsb 0.6 ωms1,cs3b,a 1
ωcs1b,c,bsc2 -0.25 ωfsb,ms2 0.9 ωssb,srsb 0.8 ωms2,cs1b,c 1
ωbsc1,bsc2 -0.2 ωms2,ms1 -1 ωpsb,esb 0.7 ωms2,cs2b 1
ωfsb,cs1b,c 3 ωms3,ms1 -1 ωcs2b,esb -0.1 ωms2,cs3b,a 1
ωbsc2,cs1b,c 1 ωms3,ms2 -1 ωesb,ssb 0.7 ωms3,cs1b,c 1
ωbsc1,cs1b,c 0.0 ωms3,cs3b,a 1 ωfsb,cs2b 0.1 ωms3,cs2b 1

The personality type concerning sensitivity to a stimulus has been taken into ac-
count by varying the connection strength of the weights ωsrsb,fsb between the sensory
representation of the b and the feeling state fsb. Table 5 shows the variation in person-
ality type from less sensitive to most sensitive. The model has been executed a large
number of times with such scenarios; in Fig. 2 one of them is depicted, the person has
high sensitivity to the stimulus.

Table 5. Personality variation for sensitivity

Person ωsrsb,fsb Personality type

person_ 1 0.3 less sensitive to the stimulus
person _2 0.6 more sensitive to the stimulus
person _3 0.9 most sensitive to the stimulus

1 fs b
0.9 ms1
0.8 ms2
ms3
0.7
cs1 b,c
0.6 cs2 b
0.5 cs3 b,a
0.4
0.3
0.2
0.1
0
0 1 2 3 4 5 6 7 8 9 10 11 12

1
cs2 b
0.8 fs b
es b
0.6 srs w
0.4

0.2

0
0 1 2 3 4 5 6 7 8 9 10 11

1 bs c1
bs c2
0.8
fs b
0.6 cs1 b,c
0.4 srs w

0.2
0
0 1 2 3 4 5 6 7 8 9 10 11

1 fs b
0.8 cs3 b,a
es a
0.6
srs w
0.4 ps a
0.2
0
0 1 2 3 4 5 6 7 8 9 10 11
 Figure 2. Simulation results of scenario 3 for person_3 (most sensitive to the stimulus). Up-
per graph: monitoring and decision process. Three lower graphs: the 3 controlled regulation
strategies

As the upper graph shows, when the simulation starts, first ms1 becomes active,
after a while ms2, and in the last phase ms3. The graph also shows the control states;
first based on ms1 it is decided to activate the emotion suppression control state cs2b.
It suppresses the negative feeling (shown in the second graph) but as the negative
feeling still increases, ms2 is triggered, and based on this it is decided to activate
control state cs1b,c for the reappraisal strategy. This alters the beliefs by suppressing
the negative belief, resulting in strengthening of positive belief and at the same time
the negative feeling decreases (shown in the third graph). After applying two
strategies, the level of negative feeling still is increasing, which triggers ms3, and
based on this it is decided to activate control state cs3b,a for situation modification.
Due to this the person moves away (change of situation) from the stimulus and gets
rid of the negative feelings (shown in the last graph).

5 Discussion

In this paper, a neurologically inspired cognitive computational model for inter-

nal monitoring and decision making about the selection of emotion regulation strate-
gies has been presented. The model covers three emotion regulation strategies (ex-
pressive suppression, reappraisal or reinterpretation, and situation modification),
adopted from [21] which lacks an internal monitoring and decision model as ad-
dressed in the current paper. A number of simulation experiments have been per-
formed according to different scenarios, thereby considering different personality
characteristics and intensities of stimuli. The decision process to select one or more
particular regulation strategies primarily takes the assessed current feeling state into
account, but can easily be extended to involve other elements as well.
The obtained human-like model can be used in different ways. As a first
application it can be a basis for virtual characters showing emotions and applying
emotion regulation strategies in a flexible way depending on the situation. Secondly,
the model can be used as an ingredient to develop human-aware or socially aware
computing applications; e.g., [22][23][24]. More specifically, in [24][25] it is shown
how such applications can be designed with knowledge of human processes as a main
ingredient, represented by a computational model of these processes which is
embedded within the application. Such computational models can have the form, for
example, of qualitative causal models, or of dynamical numerical models. The
computational model for decision making about emotion regulation proposed here can
be used in such a way to design a human-aware software application to support
persons with stress-related problems and professionals supporting them.
In the literature a number of computational emotion regulation models have been
proposed over the years, one of which was presented in [5]. Here a theory of appraisal
was modeled. The presented model is based on the idea that emotions are generated
based on an individual’s interpretation of the situation. In this approach the model is
based on symbolic and numeric representations and appraisal operates on them,
whereas our approach uses a dynamical systems representation. Different coping
strategies are proposed in that paper, e.g., “belief-related coping” which can be related
to reappraisal in the model proposed here. A difference is that the model presented in
the current paper focuses on modeling the decision making process in an explicit
manner, and that the modeling approach here is based on temporal-causal networks as
described in [21, 28].

References

1. Samson, Andrea C., Antonio Y. Hardan, Rebecca W. Podell, Jennifer M. Phillips, and James
J. Gross, “Emotion Regulation in Children and Adolescents with Autism Spectrum Disor-
der.” Autism Research 8, 2015, 9–18.
2. Balzarotti, Stefania, Oliver P. John, and James J. Gross. “An Italian Adaptation of the Emo-
tion Regulation Questionnaire.” European Journal of Psychological Assessment, 26, 2010,
pp. 61-67
3. Gross, James J., “Emotion Regulation: Current Status and Future Prospects.” Psychological
Inquiry, 26, 2015, 1–26.
4. Marsella, S., Gratch, J., Wang, N., Stankovic, B.: Assessing the validity of a computational
model of emotional coping. In: 2009 3rd International Conference on Affective Computing
and Intelligent Interaction and Workshops. pp. 1–8. IEEE (2009).
5. Marsella, S.C., Gratch, J.: EMA: A process model of appraisal dynamics. Cognitive Systems
Research. 10, 70–90 (2009).
6. Webb, Thomas L., Inge Schweiger Gallo, Eleanor Miles, Peter M. Gollwitzer, and Paschal
Sheeran. 2012. “Effective Regulation of Affect: An Action Control Perspective on Emotion
Regulation.” European Review of Social Psychology, 23, 2012, 143–86.
7. Bonanno, George A., and Charles L. Burton. “Regulatory Flexibility an Individual Differ-
ences Perspective on Coping and Emotion Regulation.” Perspectives on Psychological Sci-
ence, 8, 2013, 591–612.
8. Côté, Stephane, Gyurak, Anett, & Levenson, Robert W., The ability to regulate emotion is
associated with greater well-being, income, and socioeconomic status. Emotion, vol. 10,
2010, pp. 923-933.
9. Van Dillen, Lotte F., Dirk J. Heslenfeld, and Sander L. Koole, “Tuning down the Emotional
Brain: An fMRI Study of the Effects of Cognitive Load on the Processing of Affective Im-
ages.” Neuroimage 45, 2009, 1212–1219.
10. Cuijpers, Pim, Annemieke Van Straten, and Lisanne Warmerdam, “Behavioral Activation
Treatments of Depression: A Meta-Analysis.” Clinical Psychology Review 27, 2007, 318–
26.
11. Summerfield, Christopher, Emily H. Trittschuh, Jim M. Monti, M.-Marsel Mesulam, and
Tobias Egner, “Neural Repetition Suppression Reflects Fulfilled Perceptual Expectations.”
Nature Neuroscience 11, 2008, 1004–1006.
12.Gross, James J.. “Emotion Regulation in Adulthood: Timing Is Everything.” Current Direc-
tions in Psychological Science, 10, 2001, 214–219.
13.Gross, James J., “The Emerging Field of Emotion Regulation: An Integrative Review.”
Review of General Psychology 2, 1998, 271-299.
14.Gross, James J., and Ross A. Thompson. 2007. “Emotion Regulation: Conceptual Founda-
tions.” In: Gross, James J. (ed), (2007). Handbook of emotion regulation. New York, NY,
US: Guilford Press, pp. 3-24.
15.Gross, James J.. “Emotion Regulation: Affective, Cognitive, and Social Consequences.”
Psychophysiology, 39, 2002, 281–91.
16. Lazarus, Richard S., and Elizabeth Alfert, “Short-Circuiting of Threat by Experimentally
Altering Cognitive Appraisal.” The Journal of Abnormal and Social Psychology, 69, 1964,
195-205.
17. Dörfel, Denise, Jan-Peter Lamke, Falk Hummel, Ullrich Wagner, Susanne Erk, and Henrik
Walter, “Common and Differential Neural Networks of Emotion Regulation by Detachment,
Reinterpretation, Distraction, and Expressive Suppression: A Comparative fMRI Investiga-
tion.” Neuroimage, 101, 2014, 298–309.
18. Kim, M. Justin, Rebecca A. Loucks, Amy L. Palmer, Annemarie C. Brown, Kimberly M.
Solomon, Ashley N. Marchante, and Paul J. Whalen.. “The Structural and Functional Con-
nectivity of the Amygdala: From Normal Emotion to Pathological Anxiety.” Behavioural
Brain Research, 223, 2011, 403–410.
19. Phelps, Elizabeth A., Mauricio R. Delgado, Katherine I. Nearing, and Joseph E. LeDoux.
“Extinction Learning in Humans: Role of the Amygdala and vmPFC.” Neuron, 43, 2004,
897–905.
20. Ochsner, Kevin N., and James J. Gross, “The Neural Bases of Emotion and Emotion Regu-
lation: A Valuation Perspective.” In: Gross, James J. (ed), Handbook of Emotional Regula-
tion, 2nd Ed. New York: Guilford, pp. 23–41.
21. Treur, Jan, Dynamic Modeling Based on a Temporal-Causal Network Modeling Approach.
Biologically Inspired Cognitive Architectures, 2016, to appear.
https://www.researchgate.net/publication/289193241_Dynamic_Modeling_Based_on_a_Te
mporal-Causal_Network_Modeling_Approach
22. Damasio, Antonio, “The Feeling of What Happens: Body.” Emotion and the Making of
Consciousness, Vintage, London, 1999.
23. Abro, Altaf H., Adnan Manzoor, Seyed Amin Tabatabaei, and Jan Treur, “A Computational
Cognitive Model Integrating Different Emotion Regulation Strategies.” Proc. BICA’15.
Procedia Computer Science 71, 2015, 157–68.
24. Pentland, Alex, Socially aware computation and communication, IEEE Computer, 38, 2005,
33-40.
25. Pantic, Maja, Pentland, Alex, Nijholt, Anton, and Huang, Thomas S., Human Computing
and Machine Understanding of Human Behavior: A Survey, Proc. of the Int. Conf. on Mul-
timodal Interfaces, 2006, 239-248
26. Treur, Jan, On Human Aspects in Ambient Intelligence. In: Proc. of the First Int. Workshop
on Human Aspects in Ambient Intelligence. In: M. Muehlhauser et al. (eds.), Constructing
Ambient Intelligence: AmI-07 Workshops Proceedings. Comm. in Computer and Infor-
mation Science (CCIS), vol. 11, Springer Verlag, 2008, pp. 262-267.
 27. Bosse, Tibor, Hoogendoorn, Mark, Klein, Michel C.A., and Treur, Jan, A Generic Agent
Architecture for Human-Aware Ambient Computing. In: Mangina, E., et al. (eds.), Agent-
Based Ubiquitous Computing. World Scientific Publishers: Atlantis Press, 2009, pp. 35-62.
28. Treur, Jan: Network-Oriented Modeling: Addressing Complexity of Cognitive, Affective and
Social Interactions. Series on Understanding Complex Systems, pp. 503. Springer Interna-
tional Publishers, 2016, pp. 499.

View publication stats