=Paper=
{{Paper
|id=Vol-1315/paper1
|storemode=property
|title=An Interdisciplinary Approach for a Holistic and Embodied Emotion Model in Humanoid Agents
|pdfUrl=https://ceur-ws.org/Vol-1315/paper1.pdf
|volume=Vol-1315
|dblpUrl=https://dblp.org/rec/conf/aic/SchaatHDD14
}}
==An Interdisciplinary Approach for a Holistic and Embodied Emotion Model in Humanoid Agents==
https://ceur-ws.org/Vol-1315/paper1.pdf
An Interdisciplinary Approach for a Holistic and
Embodied Emotion Model in Humanoid Agents
Samer Schaat1, Matthias Huber2, Klaus Doblhammer1, Dietmar Dietrich1
1
Institute for Computer Technology, Vienna University of Technology, A – 1040 Vienna
{schaat, doblhammer, dietrich}@ict.tuwien.ac.at
2
Department of Education, University of Vienna, A – 1010 Vienna
matthias.huber@univie.ac.at
Abstract. Computational models of the human decision making process have
already enabled several insights and applications. Nevertheless, such models
have only recently begun to consider that simply modelling rational decision
making was insufficient to represent human behavior. Recent efforts have start-
ed to consider this factor by utilizing computational models of emotions. One of
the most significant challenges to be faced here is the interdisciplinary coopera-
tion required in order to develop a holistic and integrated model, which reflects
aspects of embodiment and is integrated in a holistic cognitive architecture. In
this paper we will analyse how models from relevant disciplines can support us
in outlining an overview model which considers the mentioned criteria.
Keywords: Emotions · Humanoid Agents · Evaluation Models · Decision Mak-
ing · Simulation · Artificial General Intelligence · Neuroscience · Neuropsy-
choanalysis.
1 Introduction
If one considers humans as being the archetypes of intelligent information processing,
the fundamental mechanisms of human decision making must be considered. This is
especially relevant for a holistic model of decision making. Over the last years the
focus in this area was on rational decision making, with little consideration for the
fact that this is actually based upon unconscious processes. In particular the consider-
ation of affective processes is needed in a computational representation of human
capabilities that are used to cope with every-day problems, such as “intuition”, “gut
feeling” etc. A further tendency in humanoid agents is the reduction of decision mak-
ing to a set of universal rules. Even when the abstract principles of decision making
are generally valid, the definite rules followed by humans cannot be generalized, and
therefore must be generated according to a subjective approach, based on memory
and also in an agent specific manner. Besides, conventional decision making models
are mostly abstracted from the body. Following an embodied approach, human cogni-
tion and particularly decision making is however based upon the interaction of the
brain with the rest of the body.
Page 16 of 171
These topics can only be tackled with an interdisciplinary approach, where com-
puter science is not used as a tool, but its methodologies are applied on the content of
other disciplines. Such approach not only enables to develop computer models for
technical applications, but also serves precision, concretization and sharpening of
concepts from different disciplines. In order to develop a deterministic model of an
agent decision making process according to human-based principles, we will leverage
insights from different disciplines, considering embodiment, subjective approaches
and unconscious processes. In particular we consider neuroscience, psychoanalysis,
neuropsychoanalysis, and current computation models of emotions. Although the
unconscious is a key concept in psychoanalysis, as far as we know, it is not analyzed
yet for computational models of affective processes, and neuroscientific theories are
not reflected in a holistic and embodied computational model of human decision mak-
ing yet. In order to further construct a model transformation upon this and to integrate
neuroscientific and (neuro)psychoanalytic models into a technical model, it is neces-
sary that they are described consistently and holistically. Even though this paper shall
focus on a model of the basic mechanisms of decision making, the holistic perspective
may not be neglected. This represents the integration of cognition and emotion, which
are thereby not modelled separately. Upon highlighting how each discipline (by an
expert of neuroscientific emotion theories, a psychoanalyst and a computer scientist)
perceives the fundamental mechanisms of subjective decision making in terms of the
criteria described, we shall sketch an evaluation model for subjective decision making
in humanoid agents which integrates the insights of the various disciplines.
2 Emotions in Neuroscience
Currently several epistemologically divergent neuroscientific research movements are
endeavoring to describe human emotionality in a holistic context. In particular the
work of Jaak Panksepp is of note, specifically his theory of mutually linked (neuronal
interacting) basic emotion systems (seeking, fear, panic, rage) [1]. Also of note is the
work of Gerald M. Edelman and his theory of Neuronal Darwinism in which emo-
tions, in the context of values and value systems, are introduced in a holistic theory of
consciousness [2], a precursor to a modern integrative comprehension of emotions.
The research of Joseph LeDoux concerning the “Amygdala Fear System” [3] and also
the work of James McGough on the significance of emotions in learning and memory
[4] are also worthy of mention. The Portuguese Antonio Rosa Damasio is recognized
as one of the most significant and influential contemporary neuroscientists. His “The-
ory of Emotion” is highly regarded far beyond neurobiology circles, as his compre-
hension of emotions and feelings possesses great explanatory potential, confirmed
through diverse studies [5], [6], and a plethora of neurophysiological data [7], [3], [8]:
On the one hand Damasio is able to justify the holistic claim as his comprehensive
body of work includes, along with the Theory of Emotion, a “Theory of Self”, a
“Theory of Consciousness”, a “Theory of Mind” along with a theory governing the
automatic regulation of life. On the other hand, Damasio also took on the challenge of
creating a model for the dualistic juxtaposition of body and mind, based on the inevi-
Page 17 of 171
table interdependence of reason and feelings, thereby transporting an anthropological
comprehension which satisfies both the psychoanalytical developmental theories and
phenomenological traditions.
The starting point of Antonio Damasio’s deliberations is the organism, as a holis-
tic, open system, that progressively interacts with the environment and is organized
according to the operating principle of homeostasis (automatic life regulation),
whereby emotions form the ultimate level of these regulative (permanently active)
mechanisms. This also clarifies why emotion and feeling take on a biological (and
also evolutionary) function as necessary survival regulators for the protection of an
organism. Damasio differentiates in this context between three processing stages: (1)
an emotional status, which can be unconsciously triggered and executed, (2) a feeling
state, which can be unconsciously represented along with (3) a consciously generated
feeling state, in which the organism knows that it possesses emotions but also feelings
[6, p. 51]. Furthermore three types of emotions are differentiated: (a) the primary
emotions (fear, anger, sorrow, happiness, disgust, surprise) which are congenital (pre-
organized in terms of classical “Jamesist” feelings) and can be thought of as being
genetically determined, universal and culturally dependent basic emotions. (b) The
secondary or social emotions which in contrast develop over the course of ontogeny
and emerge as soon as systematic connections of objects and/or situations with prima-
ry emotions are formed. Social or secondary emotions (e.g. compassion, embarrass-
ment, shame, guilt, pride, jealously, awe, envy etc.) are thereby acquired and respec-
tively triggered through mental registration with respect to the cognitive processing of
situations and/or objects. (c) Background emotions are seen by Damasio as being the
consequence of certain combinations of simple regulative (homeostatic) reactions
(e.g. desire) [9, p. 56]. They are ever-present but are seldom consciously perceived
and may be interpreted as being an expression of well-being or discomfort. Emotions
thereby fulfill a double biological function: as already mentioned they must continu-
ously regulate the inner status of the organism. Above and beyond that they must also
trigger a specific reaction to a particular stimulus or situation. Two mechanisms are
available for this: emotions are formed either when our sensory organs process certain
objects and/or situations (the body-loop) or when the organism retrieves certain ob-
jects and/or situations from memory and represents them as imaginary images in the
thought process (the as-if-loop). It has already been stated that emotions and feelings
are temporally and structurally predetermined: the key content of a feeling is the illus-
tration of a particular bodily state. A feeling is thereby a projection of the body under
certain conditions and the feeling of an emotion is the projection of the body while
under the influence of a particular emotional process. Additionally, along with the
bodily-related projections, in certain situations specific projections of the thought
process are also relevant. Therefore a (conscious) feeling consists of the perception of
a certain bodily state along with the perception of a certain associated mental state
(the cognition of the emotion). As before, Damasio also differentiates here between
three types of feelings: (a) feelings of basic universal emotions (in terms of the prima-
ry emotions), (b) feelings of differentiated universal emotions (as a connection of
cognitive states with emotional bodily states; depending on experience, in terms of the
Page 18 of 171
social emotions), along with (c) background sensations (in terms of background emo-
tions, although not formed by emotions in the strictest sense) [5, p. 208].
In the context of this “Theory of Emotion”, an elementary principle is clearly evi-
dent in Damasio’s work: emotion, feelings and consciousness are continually depend-
ent on the representations of the organism and their common entity is and remains the
body. Human thoughts and actions are therefore dependent on the emotional constitu-
tion and respectively to certain changes in bodily state. The purpose of thought and
the prerequisite for action is however always a decision, whereby the essence of a
decision lies in choosing a certain response (e.g. a course of action) [5, p. 227].
If one considers the decision making process on the basis of an undesirable devel-
opment, one thereby creates an undesirable/negative outcome, that is connected to the
associated response and is consciously perceived, even when short-lived, as an un-
comfortable/negative feeling. As a feeling (from emotion) affects the body, Damasio
choose the term somatic (soma = body), and as the feeling identifies, or marks a pro-
jection, he also chose the term marker. A somatic marker is thereby understood as
being the perception of an automatic reaction of the body to a certain projected image
(a situation or an event) respectively, as a bodily signal marking a particular scenario
as being either good or bad. [5, p. 238]. Accordingly a positive somatic marker func-
tions as a start signal and a negative somatic marker functions as an inhibitor. Somatic
markers are formed throughout the course of upbringing and socialization through
connecting certain classes of stimuli with certain classes of somatic statuses. There-
fore they touch on the process of the secondary emotions. The adaptive function of
the somatic marker (as an assistant with anticipatory skills) orientates itself towards
congenital, regulatory dispositions (internal preference system) which ensure the sur-
vival of the organism and take care of avoiding pain and seeking or increasing desire.
Looking back at the functional mechanisms of emotion, one can differentiate four
forms of the decision making process: (A) in the context of the body-loop, the body is
actually (from the prefrontal cortex and the amygdala) prompted to take on a certain
state profile, the result of which (via the somatosensory cortex) is considered with
attention and perceived. (B) In the context of the as-if-loop the somatosensory cortex
functions (as instructed by the prefrontal cortex and the amygdala) as if the signals
were received from the body. Therefore the body is taken out of the loop, nevertheless
the as-if activity patterns influence the decision making process, as it suggests that
real bodily states are symbolically processed. (C) Additionally somatic markers (e.g.
feelings) can represent very concrete components or triggers for decisions, regardless
of whether they follow a real or representative route. (D) Very often decisions are
made, where it appears that no feeling at all was involved. Therefore it is not – and
that is key here – that it doesn’t come to an evocation of a bodily state or that of a
representative surrogate, but rather just that of the bodily state with which the signal
function is activated, it is simply just not considered and therefore not consciously
perceived [10, p. 84]. By this means somatic markers operate permanently outside of
consciousness and persistently influence conscious thought and decisions. Therefore
one differentiates between somatic markers with respect to the influence of emotion
and feeling on the decision making processes based on their neural route, (A) the real
Page 19 of 171
body-loop versus (B) the representative as-if-loop and on the basis of their influence,
(C) manifest versus (D) covert.
3 Evaluation Models in Psychoanalysis
There are multiple aspects to the description of decision making in the “psychic appa-
ratus” in classical psychoanalytic theory [11, 14]: On the one hand the body delivers
via homeostatic differences drive tensions, so-called “quota of affects” represented in
the psyche, which can consequently cathect1 psychic contents. This allocation of the
“quota of affects” to psychic contents already activates content in the unconscious
whereby these become accessible for mental processing. The level of cathexis is a
measure of the grade of the activation, representing an evaluation, which ultimately is
a key factor in determining, if content shall be processed and ultimately become con-
sciously perceivable and actionable. The psychic contents on the other hand, come
from memory traces which are associated via perceptional data and drive representa-
tives and by this means serve as a basis for cathexis. A cathected association complex
is called thing presentation. This is, during the transition from (unconscious) primary
process to secondary process – upon going through a conflict regulating defense –
linked with a so-called word presentation, which means, that from now on a psychic
content can be treated within general (formal, verbal) logic. This was not yet possible
in the primary process, as this is governed by a pre-logical order of associations. Ca-
thexes, which in the course of the mental processing of the primary process have been
displaced many times, remain intact throughout these transitions. In the secondary
process, the topical description of which encompasses the psychoanalytical precon-
scious and conscious areas, are now fixed to the “quota of affects” of certain contents
and contribute, along with the logical links via word presentations to the evaluation of
the association complexes.
The overall evaluation of action chains generated in this manner is regulated on the
one hand – in the primary process – by the so called “pleasure principle” and in the
secondary process by the “principle of reality”. The “pleasure principle” states that
the overall goal of all mental activity is to avoid unpleasure and to aim for pleasure
(as in [12, p.321]), the “principle of reality” is a variation upon this, namely the mo-
ments in which the outside world becomes included into these activity designs (as in:
[12, p.378]). Pleasure is created if psychic energy is discharged, unpleasure is equiva-
lent to the “quotas of affect” present within the apparatus. Both of these principles are
ultimately relevant for decision making and choosing a course of action, so much so,
that in total a maximum of expected pleasure less the necessary unpleasure in order to
achieve it is always sought. Primary and secondary process oriented thinking and
evaluation mechanisms complement each other towards taking action. Thinking is
essentially an experimental kind of acting. [13, p.220]. If action is undertaken, the
“quota of affects” is discharged and alongside the physical impulses, consciously
perceptible affects and impulses of feelings in particular shades are formed. Uncon-
1
In psychoanalysis this also known as the economic aspect of psychological operation.
Cathexis ( a psychoanalytic term) is the allocation of quota of affects to psychic contents.
Page 20 of 171
scious affects, feelings and sensations are of no relevance in psychoanalytic theory,
they are, - in contrast to psychic contents –virtual qualities, which with respect to the
occupation conditions of the “quota of affects” can be constructed retrospectively [14,
p.176].
4 Emotions in Neuropsychoanalysis
Neuropsychoanalysis seeks to forge a connection between psychoanalytical models
and related neuroscientific findings [7, 15]. It seeks to assume a neuroscientific per-
spective of every mental function and thereby wishes to reassess Freud’s description
of the functional organization [15, p. 830]. The results of these comparisons may
remain patchy in accordance with the method applied and usually only the most sig-
nificant theses and statements of these disciplines are studied.
Considering evaluation models for decision making, neuropsychoanalysis holds
that the conscious decisions for actions, in reality follow their unconscious initiation
[cf. 15, p. 384]. Feelings of pleasure and unpleasure with respect to an object or a
situation represent the most elementary evaluation of a consciousness, according to
Solms, Panksepp and Damasio [15, p. 836]. Likewise the basic units of these evalua-
tions, of the driving forces, can be illustrated neurologically and yield that: An amaz-
ingly large consensus emerges between Panksepp’s SEEKING-System and Freud’s
Libido-System and the highest priority of a regulating function of consciousness is to
generate feelings of pleasure and unpleasure, which in turn are then associated with
the objects which are best suited to their generation [cf. 15, p. 848].
5 Emotions in Agent-based Systems
In recent years several computational models of emotions have been developed and
integrated in the decision making process of artificial agents. These models differ in
various aspects. Generally they differ in the components which are considered as
being intrinsic to emotions (e.g. bodily processes, behavioral tendencies), in their
relationship to cognitive processes and in their representation [16]. The most im-
portant difference lies in the supporting theory, which in most cases originates from
psychology. This in turn influences the distinction if emotions are generated dynami-
cally (emergent emotions) or are designed explicitly (discrete emotions). These as-
pects of distinction are mirrored in the division of “appraisal, dimensional and ana-
tomical” computational models of emotion [16], whereby the former is the most
widespread due to its aptness in linking emotion with cognitive processes.
In appraisal models (e.g. OCC Modell [17]), emotions are formed through the
evaluation of external events regarding the agent’s beliefs, desires and intentions,
whereby coping strategies (e.g. planning, delaying) are triggered. A computational
model of this, which offers the chance to adjust the appraisal process, is FAtiMA
Modular [18]. The agent architecture consists of an extensible core architecture which
offers the framework for various implementations of appraisal models, enabling easier
comparison. The appraisal process is split into the appraisal derivation and the affect
Page 21 of 171
derivation. The former evaluates the relevance of an event and creates appraisal vari-
ables (e.g. unexpectedness, appealingness and desirability). The latter builds upon
these variables by creating the associated emotions, determined by a specific appraisal
theory. According to a set of rules, emotions then influence the choice of action in
either a reactive or deliberate form.
In dimensional models, emotions are located in a dimensional space instead of be-
ing formulated as discrete entities. A typical model is the PAD model [19], with the
dimensions pleasure, arousal and dominance. Some computational models such as
ALMA [20] and WASABI [21] utilize both, appraisal models to model appraisal pro-
cesses, and dimensional models to model mood processes. Other models such as Mi-
croPsi [22, p.143-155] describe emotions implicitly as regions of a multi-dimensional
space, defined by the parameters which determine the behavior of the agent. These
are: arousal, resolution level, dominance of the leading motive, the level of back-
ground checks (the rate of the securing behavior), the level of goal-directed behavior,
and valence. Thus, explicit emotions do not exist for agents but rather emotions are
first attributed to an agent upon (self) perception.
In conclusion, it can be stated that none of the models mentioned offers an embod-
ied model that holistically considers the various aspects of emotion, or is integrated in
a holistic cognitive architecture.
6 A Holistic and Embodied Emotion Model for Evaluation in
Decision Making
Building on the findings of the various disciplines, we sketch a holistic and embodied
model. As shown above, the models of Damasio and psychoanalysis fulfill the initial-
ly mentioned criteria especially well. Insights from both theories confirm and com-
plement each other. For a technical model of the basic mechanisms of decision mak-
ing, the psychoanalytic findings mentioned above are particularly well suited as an
abstract framework (which is required for a holistic and coherent model), and Dama-
sio’s model is especially well suited for its concretization due to its consistent and
holistic character in considering the interaction between body and mind.
The role of computer science here is to integrate the various models from other
disciplines in a consistent and coherent model of decision making, which is determin-
istic and can therefore be validated by means of simulation. Thus, computer science
enables a model building methodology and evaluation tool, by the means of agent
based simulation. The basic principle of this approach has been illustrated in a previ-
ous article [23], where a functional view of emotions in the decision making unit of
an agent was integrated. However, a holistic view considering the theories of Dama-
sios was neglected. Nevertheless, the fundamental principle remains intact, i.e. affec-
tive evaluation processes are the foundation of evaluating data (psychic contents,
thereby also actions and plans). These are those processes which use “quotas of af-
fects” or derived evaluation variables to determine the relevance of data, based on
memories, for decision making in a given situation (see Fig. 1). Thus, the evaluation
of data is an incremental process on multiple levels – considering various evaluation
Page 22 of 171
principles (pleasure and reality principles) and evaluation influences (bodily influ-
ences and through perception activated memories and fantasies).
Emotions are an additional level of this incremental and hierarchical multi-level
evaluation model. They represent (1) “quotas of affects” from the drives, (2) emotions
activated through perception and fantasy (memories associated with emotions), (3) the
current pleasure. Hence emotions form a holistic representation of the psychobiologi-
cal status of the agent (having information concerning the body and mental status)
and can therefore consider the overall status of the agent in the evaluation. The final
evaluation step is carried out by feelings, whereby depending on the intensity of the
emotion, it is transformed into a preconscious feeling and subsequently a consciously
“felt feeling” (in the sense of Damasios). The latter can be described as an inner per-
ception, upon which the agent can reflect. As with the other valuation variables, feel-
ings evaluate goals and plans by activating memories.
Fig. 1. Evaluation is an incremental process that considers multiple influences and principles.
By considering perception and fantasy, evaluation through feelings not only occurs
in the terms of gaining pleasure, but also in terms of avoiding unpleasure, that is to
say not just to support the fulfilment of drives but also to evaluate external events in
terms of their potential to increase unpleasure. Evaluation generally serves to priori-
tize and select actions, mediating between the environment and the internal state (e.g.
to fulfil desires in the environment and to adapt desires to external conditions).
The representation of the biological aspect of the psychobiological status is
achieved through drives and body perception (proprioceptive and external percep-
tion). Whereby it must be emphasized that memory-based psychic representation
(representation of drives and body representation) is used for emotion generation (and
not the body signals as such). The psychic aspect of the psychobiological status is
represented by the memories activated by environment perception and fantasies. In
the sense of Damasio, one can conceptually speak of background emotions (red influ-
ences in Fig. 2), which can be considered as moods, and emotions triggered from the
outside world.
The key concretization, when integrating Damasio’s model, concerns the consider-
ation of the embodiment by means of a mental representation of it. In this regard we
follow the approach of considering the psyche as an information theoretical level of
Page 23 of 171
the physical world. This is reflected in the differentiation between the neural, neuro-
symbolic and physical levels (see Fig. 2), and also in the application of a memory
based physical representation (see Fig. 2).
Fig. 2. A holistic and embodied emotion model.
7 Conclusion
Interdisciplinary cooperation enabled us to outline a holistic model of emotion for
humanoid agents. Due to the consideration of bodily influences and the evaluation of
perceived events, the model outlined can be considered as being a combination of
both appraisal and dimensional models, whereby neuroscientific inputs and embodi-
ment are considered in developing a holistic model. Psychoanalysis and Damasio’s
neuroscientific model fulfill the initially mentioned criteria particularly well. Where-
by psychoanalysis offers us an abstract holistic framework which can be concretized
by means of Damasio’s model. Both models also complement each other, as Damasio
considers the interdependence of body and mind more concretely and is more con-
sistent. Neuropsychoanalysis supports our approach of combining psychoanalysis
with neuroscientific models, by revealing supporting evidence. Computer science
enables the combination of the various models in one consistent and holistic model
and offers an evaluation tool by means of simulation. Having outlined such a simula-
tion model, the next steps lie in extending an existing implementation of a holistic
functional model of the human mind [23], to integrate these new findings in a holistic
model of human information processing. We expect that the integration of the various
models in an overall evaluation model will yield new discoveries and opportunities in
simulations.
Page 24 of 171
References
1. Panksepp, J.: Affective Neuroscience. The Foundations of Human and Animal Emotions.
Oxford University Press, New York (1998)
2. Edelman, G. M., Tononi, G.: Neuronaler Darwinismus. Eine selektionistische Betrach-
tungsweise des Gehirns. In: Meier, H., Ploog, D. (eds.): Der Mensch und sein Gehirn. Die
Folgen der Evolution, 187-233, Piper Verlag, München (1997)
3. Ledoux, J.: The Emotional Brain. The Mysterious Underpinnings of Emotional Life. Si-
mon and Schuster, New York (1996)
4. Cahill, L., Mcgough, J. L.: Mechanisms of emotional arousal and lasting declarative
memory. In: Trends in Neurosciences 21, 284-299 (1998)
5. Damasio, A. R.: Descartes’ Irrtum. Fühlen, Denken und das menschliche Gehirn.
[Descartes' Error. Emotion, Reason, and the Human Brain Putnam. Penguin edition,
London (2005)] Deutscher Taschenbuch Verlag, München (2001)
6. Damasio, A. R.: Ich fühle, also bin ich. Die Entschlüsselung des Bewusstseins. [The Fee-
ling of What Happens. Body and Emotion in the Making of Consciousness. Harcourt, New
York (1999)] List Verlag, Berlin (2007)
7. Solms, M., Turnbull, O.: The Brain and the Inner World. An Introduction to the Neurosci-
ence of the Subjective Experience. Other Press, New York (2002)
8. Kandel, E. R.: In Search of Memory. The Emergence of a New Science of Mind. Norton,
New York (2006)
9. Damasio, A. R.: Der Spinoza-Effekt. Wie Gefühle unser Leben bestimmen. [Looking for
Spinoza. Joy, Sorrow, and the Feeling Brain. Harcourt, New York (2003)] List Verlag,
Berlin (2007)
10. Huber, M.: Die Bedeutung von Emotion für Entscheidung und Bewusstsein. Die neurowis-
senschaftliche Herausforderung der Pädagogik am Beispiel von Damasios Theorie der
Emotion. Verlag Königshausen & Neumann, Würzburg (2013)
11. Freud, S.: The Interpretation of Dreams. The Standard Edition of the Complete Psycholog-
ical Works of Sigmund Freud, Volume IV. Hogarth Press, London (1900)
12. Laplanche, J., Pontalis, J.B.: The Language of Psycho-Analysis.London: The Hogarth
Press and the Institute of Psycho-Analysis, London (1973)
13. Freud, S.: Formulations on the Two Principles of Mental Functioning. The Standard Edi-
tion of the Complete Psychological Works of Sigmund Freud, Volume XII, Hogarth Press,
London, 213-226 (1911)
14. Freud, S.: The Unconscious. The Standard Edition of the Complete Psychological Works
of Sigmund Freud, Volume XIV. Hogarth Press, London, 159-215 (1915)
15. Solms, M.: Sigmund Freud heute. Eine neurowissenschaftliche Perspektive auf die Psy-
choanalyse. Psyche. 60, pp. 829-859 (2006)
16. Marsella, S. C., Gratch, J., Petta, P.: Computational models of emotion, In: K.R. Scherer,
T. Bänziger, and E. Roesch, (eds.), Blueprint for Affective Computing. Oxford University
Press, New York (2010)
17. Ortony, A., Clore, G. L., Collins, A.: The Cognitive Structure of Emotions. Cambridge
University Press, New York (1990)
18. Dias, J., Mascarenhas, S., Paiva, A.: FAtiMA Modular: Towards an Agent Architecture
with a Generic Appraisal Framework. In: Proceedings of the International Workshop on
Standards for Emotion Modeling (2011)
19. Mehrabian, A., Russell, J.A.: An approach to environmental psychology. MIT Press, MA
(1974)
Page 25 of 171
20. Gebhard, P.: ALMA: A layered model of affect. In: Proceedings of the fourth international
joint conference on Autonomous agents and multiagent systems. pp. 29-36 (2005)
21. Becker-Asano, C., Wachsmuth, I.: Affective computing with primary and secondary emo-
tions in a virtual human. Auton Agent Multi-Agent Syst. 20, pp. 32-40 (2010)
22. Bach, J.: Principles of Synthetic Intelligence, Psi: An Architecture of Motivated Cognition.
Oxford University Press, New York (2009)
23. Schaat, S., Doblhammer, K., Wendt, A., Gelbard, F., Herret, L., Bruckner, D.: A Psycho-
analytically-Inspired Motivational and Emotional System for Autonomous Agents. In:
Proceedings of the 39th Annual Conference of the IEEE Industrial Electronics Society, pp.
6648 - 6653 (2013)
Page 26 of 171