Doubting What to Eat: A Computational Model for Food Choice Using Different Valuing Perspectives

Abstract

In this paper a computational model for the decision making process of food choices is presented that takes into account a number of aspects on which a decision can be based, for example, a temptation triggered by the food itself, a desire for food triggered by being hungry, valuing by the expected basic satisfaction feeling, and valuing by the expected goal satisfaction feeling.

Full text

Doubting What to Eat: A Computational Model
for Food Choice Using Different Valuing Perspectives
Altaf H. Abro and Jan Treur
Vrije Universiteit Amsterdam, Behavioural Informatics Group
De Boelelaan 1081, 1081 HV Amsterdam, The Netherlands
{a.h.abro, j.treur}@vu.nl
In this paper a computational model for the decision making process of food
choices is presented that takes into account a number of aspects on which a
decision can be based, for example, a temptation triggered by the food itself, a
desire for food triggered by being hungry, valuing by the expected basic
satisfaction feeling, and valuing by the expected goal satisfaction feeling.
Keywords: Computational model, food choice, Hebbian learning, desire.
1 Introduction
The effect of overweight and obesity has been studied extensively in recent years, as
all over the world it is becoming a substantial problem that increases the risk of health
problems, such as diabetes, high blood pressure, cardiovascular diseases and other [1,
2]. Moreover, obesity was found to be associated with an increased risk of morbidity
and mortality [3, 4]. One of the recent studies, published in The Lancet [5] has
described the current trends in obesity and made future predictions about continuation
of these trends: 18% of men and 21% of women will be obese by 2025. Researchers
from different fields such as neuroscience focus on eating behaviour:
‘What we eat, when and how much, all are influenced by brain reward mechanisms
that generate ‘liking’ and ‘wanting’ for foods. As a corollary, dysfunction in reward
circuits might contribute to the recent rise of obesity and eating disorders’ [6], p. 43
Also from a computational perspective attempts have been made to get an
understanding of the decision making process on food choice; e.g., [7]. Computational
models can play a vital role in providing support to lifestyle change. To develop
smart, human-aware applications, modelling the dynamics and interaction of mental
states underlying behavioural choices is becoming more and more important.
In this paper the aspects of eating behaviour and particularly making choices of
food are explored computationally. For that purpose it has been taken into account:
 How is desire for food generated in relation to a body unbalance interpreted as
being hungry and/or by any kind of sensed external stimulus. For example,
certain types of food may be sensed and considered attractive by their cues.
 How can a person learn to adopt a rational behaviour on the basis of his past
experiences; for example, this can involve Hebbian learning [8, 9].

 How does experienced satisfaction play a role, and how can the choice of food
be altered through the satisfaction level.
 How influential is the role of goals; for example, a person wants to lose weight
(long term goal) and adopts as a short term goal to eat healthy and light food.
2 Background
In the past decade a number of theories and models have addressed different aspects
that may cause overweight and obesity. This involves different fields including,
psychology, social science, health science, and neuroscience [6, 10]. Researchers
attempt to develop interventions which help people change their lifestyle to avoid
obesity. The current paper focuses on a neurologically inspired computational model,
for which in this section some neurological background is discussed.
From a neurological perspective a large number of studies have been conducted on
food-related behaviour, and how food addiction is formed; for example, discussing
how highly palatable food activates reward pathways lead to obesity [11, 12, 13].
Studies suggest that dysregulation of brain reward pathways may contribute to
increased consumption of highly palatable foods, which leads to weight gain and
obesity [14]. Another reason for addictive behaviour formation is excessive eating
over long period, which involves reward circuitry, and persons become habitual in
that behaviour [13]. Different brain regions are involved in food reward circuitry
which leads addiction behaviour: a prefrontal region and the amygdale, and also the
limbic system integrating amygdala with hypothalamus and septal nuclei [15,16, 17].
Changing the eating behaviour is a quite challenging task, as persons often become
habitual in it. Decision making about food choices usually involves considering
various options and comparing them in order to make a reasonable choice out of them
that can satisfy hunger or taste. Every available food option is coupled with an
associated feeling related to prediction whether that option will provide satisfaction or
not. It depends on reward mechanisms, as a form of valuing of the options. In making
a choice for an option such valuing processes play a significant role.
A desire on the one hand triggers preparations for responses and on other hand
generates the associated feelings via recursive body and as-if body loops [18, 19, 20],
which can predict the consequences and satisfaction of particular options before
taking action. This is done by evaluating the options using loops involving
interaction between feeling and preparation. The connections from feelings to
preparations can be either assumed static or adaptive. Adaptive connection strengths
depend on earlier experiences [21, 22]. They can be learnt over time as a form of
neural plasticity by changing the connections strengths [9, 23].
Neuroscientific literature shows that goals guiding an individual’s behaviour are
linked with activations in the prefrontal cortex. Such activations can inhibit the
activation of subcortical structures (e.g., cerebellum and basal ganglia), associated
with habitual behaviour [24, 25] and have the ability to change habitual behaviour
[26, 27, 28]. Such control mechanisms can help persons to change their preferences or
choice of food by keeping goals in mind; for example, they may start to look for low
calories food instead of high calories food.

3 The Computational Model
In this section the proposed computational model is discussed, based on the literature
described in Section 2. The model was designed by considering a number of aspects
of eating behaviour, such as food desire generation and preparation for actions to
fulfil the desire. As causes of a food desire either internal (feeling less energy,
hungry), or external (stimuli) causes are considered. In relation to a generated desire,
particular actions are considered and the extent to which they will provide a feeling of
satisfaction. Moreover, goals are considered in this model. An overview of the model
is depicted in Fig. 1; the concepts used are explained in Table 1. The cognitive model
reflects the underlying neurological concepts at an abstracted level, and is based on a
network of temporal-causal relationships [29].
srsub
ssub
psbi
dsub
esbi
as-if-body loop
srsw
wsw
ltg
stg
bsbi
bsub
ssw
as-if-body loop
mub
srsbi
ssbi fsbi
srssf
sssf fssat
Figure 1. Conceptual representation of the computational model
The process of a desire generation is modelled in two ways, as shown in Fig.1. First, a
desire can be generated through the metabolic activity (energy level). This process is
modelled by the connections from metabolic state msub that leads to the bodily
unbalance (for being hungry) represented by body state bsub. A person senses the
bodily unbalance via the internal sensor state ssub that enables the person to form the
sensory body representation state srsub representing body unbalance in the brain. By
this body unbalance the desire to eat something to reduce the body unbalance
develops and the person prepares for the available food options. This is modelled by
the connection from the desire state dsub to preparation states psbi.

Not only the desire but also the associated feelings have impact on the preparations;
before performing any action a feeling state fsbi for the (partly prepared) action is
generated by a predictive as-if-body loop via the sensory representation state srsbi.
Table 1. Overview of the states of the proposed model (see also Fig. 1)
Formal
Informal name
Description
wsw
World state w
This characterizes the current world situation which the person is
facing the stimulus, in this case w is the food stimulus
ssw
Sensor state for w
The person senses the world through the sensor state, providing
sensory input
srsw
Sensory representation
state for w
Internal representation of sensory world information on w
mub
Metabolism for ub
Represents metabolic energy level: a low level while being inactive
and a high level while being active
bsub
Body state of ub
A bodily state that represents the body state underlying being hungry;
if a person is becoming hungry, the body unbalance ub increases, and
after eating there will be lower or no unbalance
ssub
Sensor state for ub
The person senses the bodily unbalance state ub, providing sensory
input
srsub
Sensory represent- ation
of ub
Internal sensory representation for sensed bodily unbalance
information on ub
dsub
desire for unbalance ub
Generating a desire based on sensory representation for a body
unbalance ub (e.g., a state of being hungry)
psbi
Preparation for an
action bi
Preparation for an eating response
fsbi
Feeling associated to
body state bi
A feeling state fsbi for the eating action
srsbi
Sensory representation
of bi
Internal sensory representation of body srsb for bi in the brain. In this
case here b represents the associated available food choices
ssbi
Sensor state for bi
The person senses the external body states or environment through the
sensor state, providing sensory input.
sssfbi
Sensor state for
satisfaction sfbi
The person senses the external body states providing sensory input to
the feelings of satisfaction.
srssfbi
Sensory representation
of satisfaction sfbi
Internal representation of the body aspects of feelings of satisfaction
fssfbi
Feeling for satisfaction
sfbi
Feeling of satisfaction; these are the feelings about the considered food
choice, how much satisfactory it is.
esbi
Effector state for action
bi
In the considered scenarios the action bi is performed to eat food of
particular choice for which a person is prepared, to reduce the body
unbalance ub
bsbi
Body state for bi
This represents external body states related to eating that particular
food.
ltg
Long term goal
This represents the long term goal, to lose weight, for example
stg
Short term goal
Short term goal refers to smaller incremental way of achieving long
term goals for example start to eat healthy avoid from fast food etc.
ωfsbi, psbi
Learnt connections
These are the connections which can be learnt by Hebbian learning.
This one models how the generated feeling affects the preparation for
response bi.

In the considered scenario, the desire state dsub has an effect on a number of
preparation states psbi for responses bi which lead to sensory body representation srsbi
and to the feeling states fsbi. Subsequently, the states fsbi have strengthening impacts
on the preparation state psbi, which in turn has an impact on feeling state, fsbi, through
srsbi which makes the process recursive: an as-if body loop (e.g., [30]).
A desire, generated by body unbalance or by environmental factors or both, needs
to be fulfilled by eating a particular selected food. A person takes action for a selected
food option to reduce or eliminate the body unbalance. In this model the effect of an
executed action bi (by esbi) is effectuated on the one hand through the connection from
effector states esbi to body states bsbi that are bodily representations of direct
valuations or satisfaction for the selected options. On the other hand the action effect
is effectuated through the connection from effector states esbi to body states bsub, this
action effect represents reduction of the bodily unbalance bsub after eating.
Another process involves the expected satisfaction feeling of goal fulfilment after
the particular action execution esbi (eating particular food). Expected satisfaction
feelings also have effect on making choices. For example, if a person takes fast food
initially, after eating he may not feel satisfied as it was not healthy, so next time (s)he
may prefer to go for healthy food, in order to feel more satisfaction.
The model as depicted in Fig.1 shows the connection from body state bsbi sensed
by the person via sensor state sssf and represented by sensory representation state for
goal fulfilment satisfaction srssf. This representation state srssf receives impact from
the preparation state psbi via an as-if-body loop as well. The sensory information srssf
then leads to a level of expected feelings of goal fulfilment satisfaction fssf. So on the
basis of this satisfaction a person may continue to choose the same option (eat the
same kind of food) again and again or he or she may change the option, e.g., from fast
food to healthy food. In this way it will have impact on preparation states psbi.
Setting goals in life either long term or short term is also a common practice . Goals
may relate to a healthy life style, behaviour change or other ambition. In this case
long term goals can be considered to weight loss and short term goals can be
considered in relation to taking a healthy diet. If a person does not have any goal
about his or her health he or she may eat anything based on various personal
characteristics/ preference of wanting and liking. On the other hand, if a person wants
to aim at being healthy then this may affect what is considered suitable to eat. In the
considered scenario as a long term goal ltg it is taken that a person wants to lose
weight and that leads to short term goal stg which is to take a healthy diet to achieve
long term goal. A short term goal has a direct influence on the preparation states psbi.
So by keeping goals in mind a person can change his eating behaviour and start giving
preference to healthy food.
Another factor that is included in the model, is how the effect of the feeling on the
preparation can be learnt over time based on past experiences. For this a Hebbian
learning mechanism has been included by which such connections may automatically
emerge or strengthen. The connections from feeling fsbi to preparation psbi with
weights ωfsbi, psbi have been made adaptive in this way. The strengths ωfsbi, psbi are
adapted by the Hebbian learning principle, that connected neurons that are frequently
activated simultaneously strengthen their connections over time [8, 9, 31].

In the example scenarios it has been shown that if a person starts changing eating
behaviour than the particular connection gets more strength over time due to learning.
Each time the person experiences a certain feeling for a prepared action, through
Hebbian learning this strengthens the association between feeling and preparation. So
the recursive process of predictive as-if body loops via feeling states fsbi to the
preparation states psbi, involves Hebbian learning in order to enable adaptivity.
The model description presented above (Fig.1 and Table1) represents a network
of cognitive and affective states in a conceptual manner. The following elements are
also considered to be given as part of such a conceptual representation:
 For each connection from state X to state Y a weight X.,Y (a number between -1
and 1), for strength of impact; a negative weight is used for suppression
 For each state Y a speed factor Y (a positive value) for timing of impact
 For each state Y (a reference to) a 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. The conceptual representation of the model
(as shown in Fig. 1 and in Table 1) can be transformed in a systematic or even
automated manner into a numerical representation as follows [29]:
 At each time point t state X connected to state Y has an impact on Y defined as
impactX,Y(t) = X,Y X(t)
where X,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
combination function cY(..):
aggimpactY(t) = cY(impactX1,Y(t), …, impactXk,Y(t))
= cY(X1,YX1(t), …, Xk,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 Y:
Y(t + t) = Y(t) + Y [aggimpactY(t) - Y(t)] t
or dY(t)/dt = Y [aggimpactY(t) - Y(t)]
 Thus the following difference and differential equation for Y are obtained:
Y(t + t) = Y(t) + Y [cY(X1,YX1(t), …, Xk,YXk(t)) - Y(t)] t
dY(t)/dt = Y [cY(X1,YX1(t), …, Xk,YXk(t)) - Y(t)]
In the model considered here, for all states for the standard combination function the
advanced logistic sum combination function alogistic,(…) is used [10]:
cY(V1, …Vk) = alogistic,(V1, …, Vk) = (
󰇛󰇜 - 
 ) (   )
Here  is a steepness parameter and  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. For example, for the prepration state psbi the model is
numerically represented in difference equation form as
aggimpactpsbi(t) =
alogistic,(srsub, psbi srsub(t), fsbi, psbi fsbi(t), fssat, psbi fssat(t), stg, psbi stg,(t))
psbi(t+t) = psbi(t) + psbi [aggimpactpsbi(t) - psbi(t)] t
The numerical representation of Hebbian learning is:
ωfsbi,psbi(t+t) = ωfsbi,psbi(t) + [η fsbi(t) psbi(t) (1 - ωfsbi,psbi(t)) - ζ ωfsbi,psbi(t)] Δt

4 Simulation Results
The computational model has been used to conduct a number of simulation
experiments according to different scenarios performed using the Matlab
environment. In this section one is described in detail. The following scenario is used:
James is facing overweight leading to an obesity problem, and has difficulties in
everyday life routine regarding his diet (dietary pattern). To overcome this
problem, he set goals to lose weight by eating healthy food. Here food is
categorized in three categories, light food (very low calories), healthy food (low
calories), and fast food (high calories). So, he has three available food options to
choose from, each of which gives some extent of satisfaction as expected from that
food. As he wants to lose weight, he will want to go for healthy or light food. In
daily routine, when he feels hungry he looks around to have healthy food to eat.
The parameter values used are shown in Tables 2 and 3.
Table 2. Values of threshhold, steepness and update speed
state


state


state


bsub
0.5
8
fsb1
0.16
2
sssf1
0.14
2.4
ssub
0.14
2.4
fsb2
0.16
2
sssf2
0.14
2.4
srsub
0.14
2.4
fsb3
0.16
2
sssf3
0.14
2.4
dsub
0.5
7
esb1
0.14
2.4
srssf1
0.14
2.4
psb1
0.5
8
esb2
0.14
2.4
srssf2
0.14
2.4
psb2
0.5
8
esb3
0.14
2.4
srssf3
0.14
2.4
psb3
0.5
8
bsb1
0.14
2.4
fssat1
0.16
2.4
srsb1
0.14
2.4
bsb2
0.14
2.4
fssat2
0.16
2.4
srsb2
0.14
2.4
bsb3
0.14
2.4
fssat3
0.16
2.4
srsb3
0.14
2.4
ssb2
0.14
2.4
stg
0.14
2.4
ssb1
0.14
2.4
ssb3
0.14
2.4
Table 3. Values of parameters used: connection weights
weight
weight
weight
weight
mub,bsub
1
srsb1,fsb1
1
esb1,bsub
-0.15
bsb1,sssf1
1
bsub,ssub
1
srsb2,fsb2
1
esb2,bsub
-0.15
bsb2,sssf2
1
ssub,srsub
1
srsb3,fsb3
1
esb3,bsub
-0.15
bsb3,sssf3
1
srsub,dsub
1
fsb1,psb1
0.4
esb1,bsb1
1
sssf1,srssf1
1
wsw,ssw
1
fsb2,psb2
0.4
esb2,bsb2
1
sssf2,srssf2
1
ssw,srsw
1
fsb3,psb3
0.4
esb3,bsb3
1
sssf3,srssf3
1
srsw,dsub
0.8
psb1,esb1
1
bsb1,ssb1
1
psb1,srssf1
1
dsub,psb1
1
psb2,esb2
1
bsb2,ssb2
1
psb2,srssf2
1
dsub,psb2
1
psb3,esb3
1
bsb3,ssb3
1
psb3,srssf3
1
dsub,psb3
1
stg,psb1
0
ssb1,srsb1
1
srssf1,fssat1
0
psb1,srsb1
1
stg,psb2
0
ssb2,srsb2
1
srssf2,fssat2
0
psb2,srsb2
1
stg,psb3
0
ssb3,srsb3
1
srssf3,fssat3
0
psb3,srsb3
1
fssat1,psb1
0.3
fssat2,psb2
0.3
fssat3,psb3
0.3
ltg,stg
0.8

In the scenario the strength of the weights fsbi,psbi of the connections from feelings
to the considered preparations change over time through the Hebbian learning
mechanism. Initial values of the activation levels for all states have been chosen 0
except the metabolic activity which depends on the scenario, and activation values for
goals either are 0 or higher, also depending on the scenario. The simulation is
executed for some scenarios for 180 time points and for other scenarios simulations
for 1500 time points; the time step t = 0.1. Learning rate η = 0.016 and extinction
rate ζ = 0.000015 have been used when Hebbian learning is used. Details of the
values for parameters used in the simulation are given in Table 2 (threshold  and
steepness ) and in Table 3 (connection weights). In the scenario depicted in Figure 2,
the simulation shows five times hunger.
Figure 2. Simulation results of five times body unbalance (hunger), with Hebbian learning
The role of the learning mechanism is illustrated, so the person can change his eating
behaviour by learning through past experiences. When the person notices a body
unbalance or hunger that triggers the desire to eat then initially the person may have
equal tendency for all available options Figure 2(b). But through the associated
feelings valuation of the options takes place and the person can learn Figure 2(e): the
eating behaviour can be changed by adapting it slowly due to learning from his

experiences. In this simulation Hebbian learning is used on the weights fsbi,psbi of the
connections between associated feelings and preparations so these links are dynamic
in this scenario; the weights esbi,bsbi of the connections from execution states to body
state are 0.3, 0.9, 0.6 respectively, these values are only for this simulation, change in
these connections will make a change in the learning connections as well.
5 Conclusion
The model presented in this paper is a neurologically inspired computational model
for making food choices by involving a number of internal and external factors. The
focus of this paper is to formalise the dynamics and interaction of internal states
which are involved in decision of making food choice. This model will help to
understand the complex process from food desire generation to food choices that
evolve based on internal prediction in combination with associated feelings for
valuation of the available food options. The model also incorporates the role of goals
and their influence on decision making on food choices. The simulation results
suggest that this model is capable of learning of making food choices on the one hand
through associated feelings to fulfil the desire, and on the other hand through levels of
satisfaction in relation with various types of food and goals. The proposed
computational model for making food choices can be used to develop human-aware
intelligent systems that can help and support persons with overweight and obesity.
In future work, the model will be extended with food desire regulation strategies,
and further focus will be on the social and environmental factors; in addition it may
be further extended with more personal characteristics.
References
1. WHO.: Global status report on noncommunicable diseases 2014. World Health
organization (2014)
2. WHO.: Obesity: Preventing and managing the global epidemic . Geneva: World Health
Organization (2000)
3. Stroebe, W.: Dieting, overweight, and obesity: Self-regulation in a food-rich
environment . Washington, DC: American Psychological Association (2008)
4. McGee, D.L.: Body mass index and mortality: a meta-analysis based on person -level data
from twenty-six observational studies. Annals of Epidemiology, 15, 87—97 (2005)
5. Vos, T., Barber, RM., Bell, B., et al.: Global, regional, and national incidence, prevalence,
and years lived with disability for 301 acute and chronic diseases and injuries in 188
countries, 1990-2013: a systematic analysis for the Global Burden of Disease Study 2013.
The Lancet. 386(9995), 743--800 (2015)
6. Berridge, K.C., Chao-Yi, Ho., Jocelyn, M.R., Alexandra, G.D.: The Tempted Brain Eats:
Pleasure and Desire Circuits in Obesity and Eating Disorders. Brain Research, 1350, 43--
64 (2010)
7. Bosse, T., Hoogendoorn, M., Memon, Z.A., Treur, J., Umair, M.: A computational model
for dynamics of desiring and feeling. Cognitive Systems Research, 19-20, 39--61 (2012)
8. Hebb, D O.: The Organisation of Behaviour. New York: Wiley (1949)

9. Bi, G.Q., Poo, M.M.: Synaptic Modification by Correlated Activity: Hebb's Postulate
Revisited. Annu. Rev. Neurosci. Annual Review of Neuroscience, 24(1), 139--66 (2001)
10. Berridge, K.C., Terry, E.R., Aldridge, J.W.: Dissecting Components of Reward: ‘liking’,
‘wanting’, and learning. Current opinion in pharmacology, 9, 65--73 (2009)
11. Finlayson, G., King, N., Blundell, JE.: Liking vs. wanting food: importance for human
appetite control and weight regulation. Neurosci Biobehav Rev., 31(7): 987--1002 (2007)
12. Berner, L.A., Avena, N.M., Hoebel, B.G.: Bingeing, self-restriction, and increased body
weight in rats with limited access to a sweet-fat diet Obesity, 16, 1998—2002 (2008)
13. Davis, C.: From Passive Overeating to ‘Food Addiction’: A Spectrum of Compulsion and
Severity. ISRN Obesity 2013, 1--20 (2013)
14. Berthoud, H.R., Lenard. N.R., Shin A.C.: Food Reward, Hyperphagia, and Obesity. AJP:
Regulatory, Integrative and Comparative Physiology, 300(6), R1266--R1277 (2011)
15. Elliott, R., Friston, K.J., Dolan R.J.: Dissociable neural responses in human reward
systems. J Neurosci., 20(16), 6159--6165 (2000)
16. Barbara, J. R., Erin, L.M., Liane, S.R.: Portion size of food affects energy intake in
normal-weight and overweight men and women. Am J Clin Nutr., 76(6), 1207—13 (2002)
17. Koob,G.F, Nora, D.V.: Neurocircuitry of Addiction. Neuropsychopharmacology, 35(1),
217--238 (2010)
18. Damasio, A.R.: The Feeling of What Happens: Body and Emotion in the Making of
Consciousness. New York: Harcourt Brace (1999)
19. Damasio, A.R.: Looking for Spinoza. Vintage books, London (2004)
20. Damasio, A.R.: Self Comes to Mind: Constructing the Conscious Brain. Pantheon books,
New York (2010)
21. Damasio, A.R.: Descartes' Error: Emotion, Reason and the Human Brain. Papermac,
London (1994)
22. Damasio, A R. The somatic marker hypothesis and the possible functions of the prefrontal
cortex. Philosophical Transactions of the Royal Society of London. Series B, Biological
Sciences, 351 (1346), 1413--1420 (1996)
23. Gerstner, W., Kistler, W.M.: Mathematical formulations of Hebbian learning. Biol.
Cybern, 87, 404--415 (2002)
24. Ashby, F.G., Turner, B.O., Horvitz, J.C.: Cortical and basal ganglia contributions to habit
learning and automaticity. Trends Cogn. Science, 14, 208--215 (2010)
25. De Wit, S., et al.: Differential engagement of the ventromedial prefrontal cortex by goal-
directed and habitual behaviour toward food pictures in humans. J. Neurosci., 29, 11330--
11338 (2009)
26. Mowrer, O.H.: Learning Theory and the Symbolic Processes. Wiley, New York (1960)
27. Watson, J.B.: Psychology as Behaviourist Views It. Psy. Review, 20, 158--177 (1913)
28. Webb, T.L., Paschal, S., Aleksandra, L.: Planning to Break Unwanted Habits: Habit
Strength Moderates Implementation Intention Effects on Behaviour Change. British
Journal of Social Psychology, 48(3), 507--23 (2009)
29. Treur, J.: Dynamic Modeling Based on a Temporal–causal Network Modeling Approach.
Biologically Inspired Cognitive Architectures, 16, 131--168 (2016)
30. Damasio, A.R., “The Feeling of What Happens: “Body Emotion and the Making of
Consciousness, Vintage, London (2000)
31. Gerstner, W., Kistler, W.M.: Mathematical formulations of Hebbian learning. Biol.
Cybern., 87, 404--415 (2002)