ArticlePDF Available

Emotion in the Common Model of Cognition

Authors:
Othalia Larue
Othalia Larue
Othalia Larue
  • This person is not on ResearchGate, or hasn't claimed this research yet.
Robert West at Carleton University
Robert West
Paul Rosenbloom at University of Southern California
Paul Rosenbloom
Christopher L Dancy at Bucknell University
Christopher L Dancy
Show all 7 authorsHide
Download full-text PDF
Read full-text
Download citation

Abstract

Emotions play an important role in human cognition and therefore need to be present in the Common Model of Cognition. In this paper, the emotion working group focuses on functional aspects of emotions and describes what we believe are the points of interactions with the Common Model of Cognition. The present paper should not be viewed as a consensus of the group but rather as a first attempt to extract common and divergent aspects of different models of emotions and how they relate to the Common Model of Cognition.
Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
ScienceDirect
Available online at www.sciencedirect.com
Procedia Computer Science 145 (2018) 740–746
1877-0509 © 2018 The Authors. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review under responsibility of the scientific committee of the 9th Annual International Conference on Biologically Inspired Cognitive
Architectures.
10.1016/j.procs.2018.11.045
Available online at www.sciencedirect.com
ScienceDirect
Procedia Computer Science 00 (2019) 000–000
www.elsevier.com/locate/procedia
1877-0509 © 2019 The Authors. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review under responsibility of the scientific committee of the 9th Annual International Conference on Biologically Inspired Cognitive
Architectures.
Postproceedings of the 9th Annual International Conference on Biologically Inspired Cognitive
Architectures, BICA 2018 (Ninth Annual Meeting of the BICA Society)
Emotion in the Common Model of Cognition
Othalia Larue1*, Robert West2, Paul S. Rosenbloom3, Christopher L. Dancy4, Alexei V.
Samsonovich5, Dean Petters6, Ion Juvina1
1 ASTECCA laboratory, Wright State University, Dayton, Ohio, USA
2 Institute of Cognitive Science, Carleton University, Ottawa, Canada
3 USC Institute for Creative Technologies, Playa Vista, California, USA
4 Bucknell University, Pennsylvania, USA
5 National Research Nuclear University MEPhI, Moscow, Russia
6 University of Birmingham Edgbaston, Birmingham, UK
Othalia.larue@wright.edu, robert.west@carleton.ca, rosenbloom@usc.edu, christopher.dancy@bucknell.edu, asamsono@gmu.edu,
D.D.Petters @cs.bham.ac.uk, ion.juvina@wright.edu
Abstract
Emotions play an important role in human cognition and therefore need to be present in the Common Model of Cognition. In this
paper, the emotion working group focuses on functional aspects of emotions and describes what we believe are the points of
interactions with the Common Model of Cognition. The present paper should not be viewed as a consensus of the group but rather
as a first attempt to extract common and divergent aspects of different models of emotions and how they relate to the Common
Model of Cognition.
© 2019 The Authors. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4. 0/)
Peer-review under responsibility of the scientific committee of the 9th Annual International Conference on Biologically Inspired
Cognitive Architectures.
Keywords: Emotion, Motivation, Common model, Cognitive Architecture
* Corresponding author.
E-mail address: Othalia.larue@wright.edu
10.1016/j.procs.2018.11.045
© 2018 The Authors. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review under responsibility of the scientic committee of the 9th Annual International Conference on Biologically Inspired
Cognitive Architectures.
1877-0509
Available online at www.sciencedirect.com
ScienceDirect
Procedia Computer Science 00 (2019) 000–000
www.elsevier.com/locate/procedia
1877-0509 © 2019 The Authors. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review under responsibility of the scientific committee of the 9th Annual International Conference on Biologically Inspired Cognitive
Architectures.
Postproceedings of the 9th Annual International Conference on Biologically Inspired Cognitive
Architectures, BICA 2018 (Ninth Annual Meeting of the BICA Society)
Emotion in the Common Model of Cognition
Othalia Larue1
*
, Robert West2, Paul S. Rosenbloom3, Christopher L. Dancy4, Alexei V.
Samsonovich5, Dean Petters6, Ion Juvina1
1 ASTECCA laboratory, Wright State University, Dayton, Ohio, USA
2 Institute of Cognitive Science, Carleton University, Ottawa, Canada
3 USC Institute for Creative Technologies, Playa Vista, California, USA
4 Bucknell University, Pennsylvania, USA
5 National Research Nuclear University MEPhI, Moscow, Russia
6 University of Birmingham Edgbaston, Birmingham, UK
Othalia.larue@wright.edu, robert.west@carleton.ca, rosenbloom@usc.edu, christopher.dancy@bucknell.edu, asamsono@gmu.edu,
D.D.Petters @cs.bham.ac.uk, ion.juvina@wright.edu
Abstract
Emotions play an important role in human cognition and therefore need to be present in the Common Model of Cognition. In this
paper, the emotion working group focuses on functional aspects of emotions and describes what we believe are the points of
interactions with the Common Model of Cognition. The present paper should not be viewed as a consensus of the group but rather
as a first attempt to extract common and divergent aspects of different models of emotions and how they relate to the Common
Model of Cognition.
© 2019 The Authors. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4. 0/)
Peer-review under responsibility of the scientific committee of the 9th Annual International Conference on Biologically Inspired
Cognitive Architectures.
Keywords:
Emotion, Motivation, Common model, Cognitive Architecture
* Corresponding author.
E-mail address: Othalia.larue@wright.edu
2 Othalia Larue et al./ Procedia Computer Science 00 (2019) 000–000
1. Introduction
Modeling emotion is essential to the Common Model of Cognition [1] because emotion can't be divorced from
cognition. Rational adaptive behavior, for example can't happen without emotion [2] .Emotions play an important
functional role, with the purpose of helping us to survive and adapt in complex and potentially hazardous physical and
social domains [3]. They aren’t necessarily finely tuned but guide our behavior in directions evolution has taught us
are wise. They operate more quickly than we can make conscious decisions. Below, we briefly describe the areas of
research on emotion that we feel are important for the normal functioning of the Common Model, identify points of
interaction with the Common Model of existing emotion models, and describe areas that still need clarifications.
2. Emotions and the four bands of cognition
Newell [4] classified levels of cognition into four bands: biological, cognitive, rational and social. Each band
operates at a different time scale (from tens of milliseconds at the biological band to hours and months at the social
band). Time scale is a direct result of the execution time of the type of operations implemented at each band. While
cognitive architectures initially focused on the cognitive band (and part of the rational band), research in cognitive
architectures has expanded to include the biological and social band. We believe emotions also act at four different
levels: they emerge at a biological level, act on cognitive and rational control and are expressed and interpreted at the
social level. Some emotions might be relatively more focused at the biological level, while some might gain a stronger
emphasis at the social level. The balance of where an emotion sits over these levels will depend on the emotion in
question and in the particular episode of that emotion. For example, momentary fear as a result of a loud noise is
different from a more social emotion like embarrassment due to violation of social conventions
The biological band is concerned with processing at the neural level. We believe models need to be compatible
with the neuroscience of emotions as it is a constraint to the structure of the architecture. ACT-R/Φ [5] is such a
model, the primary-process affect theory is used as a layer between a physiological substrate and the ACT-R cognitive
architecture.
The cognitive band includes memory-level operations (memory retrieval), decision operations (time required to
manipulate knowledge), action, and the time required to build an action and execute an action (from 100 msec to 10
seconds). At this level, we believe emotional mechanisms need to blend in with cognitive mechanisms already present
in the architecture. Past affective experiences can also influence cognitive processing as well as prospection of
affectively loaded future episodes. Juvina et al.[6] ’s core affect model, for example, integrates arousal and valuation
terms to the general activation equation of ACT-R.
The rational band is goal-driven. It processes knowledge, in order to achieve an adaptive behavior and implements
higher level cognition (goal-directed decision making, planning). In HCogAff [7] , a three-level architecture, the
second and third levels of the architecture correspond to the rational band. Tertiary emotions can cause disruptions
and modify the way the third level manages lower levels. Emotional mechanisms might also help decide whether or
not we engage in further processing. In Larue et al.[8] ’s model, the feeling of rightness (how good we feel about a
first answer) is what will determine the length of processing in the architecture (if we engage in analytical thinking
and to what extent). It is implemented using ACT-R’s core affect mechanism.
The social band includes higher level cognition with functions such as Theory of Mind, representation of others
and moral reasoning. Social aspects in cognitive architectures are increasingly being considered [9] .
3. Emotion: Theories and functions
3.1. Theories
3.1.1. Appraisal theories
Appraisal theories [10] [11] focus on the initial stage of emotional processing during which personal significance
is attributed to situational factors through an evaluation process. During the evaluation phase, emotionally relevant
information is captured by evaluating a set of factors.
Othalia Larue et al. / Procedia Computer Science 145 (2018) 740–746 741
Available online at www.sciencedirect.com
ScienceDirect
Procedia Computer Science 00 (2019) 000–000
www.elsevier.com/locate/procedia
*
Available online at www.sciencedirect.com
ScienceDirect
Procedia Computer Science 00 (2019) 000–000
www.elsevier.com/locate/procedia
*
2 Othalia Larue et al./ Procedia Computer Science 00 (2019) 000–000
1. Introduction
Modeling emotion is essential to the Common Model of Cognition [1] because emotion can't be divorced from
cognition. Rational adaptive behavior, for example can't happen without emotion [2] .Emotions play an important
functional role, with the purpose of helping us to survive and adapt in complex and potentially hazardous physical and
social domains [3]. They aren’t necessarily finely tuned but guide our behavior in directions evolution has taught us
are wise. They operate more quickly than we can make conscious decisions. Below, we briefly describe the areas of
research on emotion that we feel are important for the normal functioning of the Common Model, identify points of
interaction with the Common Model of existing emotion models, and describe areas that still need clarifications.
2. Emotions and the four bands of cognition
Newell [4] classified levels of cognition into four bands: biological, cognitive, rational and social. Each band
operates at a different time scale (from tens of milliseconds at the biological band to hours and months at the social
band). Time scale is a direct result of the execution time of the type of operations implemented at each band. While
cognitive architectures initially focused on the cognitive band (and part of the rational band), research in cognitive
architectures has expanded to include the biological and social band. We believe emotions also act at four different
levels: they emerge at a biological level, act on cognitive and rational control and are expressed and interpreted at the
social level. Some emotions might be relatively more focused at the biological level, while some might gain a stronger
emphasis at the social level. The balance of where an emotion sits over these levels will depend on the emotion in
question and in the particular episode of that emotion. For example, momentary fear as a result of a loud noise is
different from a more social emotion like embarrassment due to violation of social conventions
The biological band is concerned with processing at the neural level. We believe models need to be compatible
with the neuroscience of emotions as it is a constraint to the structure of the architecture. ACT-R/Φ [5] is such a
model, the primary-process affect theory is used as a layer between a physiological substrate and the ACT-R cognitive
architecture.
The cognitive band includes memory-level operations (memory retrieval), decision operations (time required to
manipulate knowledge), action, and the time required to build an action and execute an action (from 100 msec to 10
seconds). At this level, we believe emotional mechanisms need to blend in with cognitive mechanisms already present
in the architecture. Past affective experiences can also influence cognitive processing as well as prospection of
affectively loaded future episodes. Juvina et al.[6] ’s core affect model, for example, integrates arousal and valuation
terms to the general activation equation of ACT-R.
The rational band is goal-driven. It processes knowledge, in order to achieve an adaptive behavior and implements
higher level cognition (goal-directed decision making, planning). In HCogAff [7] , a three-level architecture, the
second and third levels of the architecture correspond to the rational band. Tertiary emotions can cause disruptions
and modify the way the third level manages lower levels. Emotional mechanisms might also help decide whether or
not we engage in further processing. In Larue et al.[8] ’s model, the feeling of rightness (how good we feel about a
first answer) is what will determine the length of processing in the architecture (if we engage in analytical thinking
and to what extent). It is implemented using ACT-R’s core affect mechanism.
The social band includes higher level cognition with functions such as Theory of Mind, representation of others
and moral reasoning. Social aspects in cognitive architectures are increasingly being considered [9] .
3. Emotion: Theories and functions
3.1. Theories
3.1.1. Appraisal theories
Appraisal theories [10] [11] focus on the initial stage of emotional processing during which personal significance
is attributed to situational factors through an evaluation process. During the evaluation phase, emotionally relevant
information is captured by evaluating a set of factors.
742 Othalia Larue et al. / Procedia Computer Science 145 (2018) 740–746
Author name / Procedia Computer Science 00 (2019) 000–000 3
Computational Models of appraisal have been developed (e.g EMA: [10] ). In EMA, these factors are: relevance,
desirability, likelihood, expectedness, causal attribution, controllability and changeability.
In Sigma ([12] [13] ), low-level appraisals are modeled as architectural self-reflection. For example, expectedness,
desirability and familiarity variables are added to respectively function as measures of surprise, the difference between
the current state and the goal, and how often something has been experienced. These variables are then used, for
example, to define attention and to guide search and exploration. Soar-Emote, uses emotional appraisals to frame
information before it is processed by the cognitive system [14] . CLARION [15] also uses automatic appraisal (relying
on implicit processes) and deliberative appraisal (slower and relying on explicit processes) to influence action-oriented
behavior or reasoning (e.g. reevaluation).
3.1.2. Core affect theory
Core affect theory [16] proposes that phenomena attributed to emotions can be explained in simpler terms without
the need for emotion labels. Core affect is a state that happens before the emotion is consciously identified that can be
described along two dimensions: feeling good or bad and feeling lethargic or energized. Core affect can be seen as a
type of appraisal and linked to appraisal theory.
Core affect has been modeled in ACT-R as the weighted accumulation of valuation and arousal values for memory
elements [6] . The existing reward mechanism of the architecture and usage information are used to compute valuation
and arousal, which are added to the general activation equation and influence the probability of retrieving a chunk.
This model was used to explain the impact of affective valuation and arousal on memory and memory decay using
participant’s memory of negative and positive emotion words after different time periods [6]. It has also been used to
model decision-making [17] and deliberative thinking, specifically what triggers deliberation [8] .
3.1.3. Somatic markers
Somatic markers [2] can be thought of as emotional tags attached to a piece of information in declarative memory.
The key point about somatic markers is that they enable emotional learning, as they are updated whenever retrieval is
associated with good or bad outcomes. ACT-R has been adapted in multiple ways to model effects that can be
attributed to somatic markers ([18]-[22]). The representation of appraisals in Sigma also can be viewed in this manner,
and more broadly the use of the quantitative metadata that is associated with memory structures in the Common Model
of Cognition also maps into this model.
3.1.4. Primary-process affect
Primary-process affect theory [3] posits primary affective processes that are implemented in evolutionarily older
neural structures. These lower-level processes combine with secondary-level (e.g., memory) processes to result in
secondary-level affect. Tertiary affective processes are also described in yet another (higher) level that is the result of
the combination of secondary-process affect with higher-level functioning (e.g, meta-processes like those involved in
self-reflection). Key in this formulation is that there are distinct systems that are defined by neurophysiological
structure and function and that make up affect; this affect, when combined with other processes, results in higher-level
emotional experience.
3.2. Functions
3.2.1. Alarm and interruption
One function of emotion is to generate alarms and interruptions. In HCogAff [7] , emotions are seen as interrupts
to ongoing processing following Simon [23]. As West and Young [24] point out, responding intelligently to
unexpected interruptions presents a computational problem for production system architectures that is difficult to
resolve without the use of external, parallel modules.
In the HCogAff architecture, interruptions and the loss of control can be caused by primary and secondary
emotions. Being frightened for example can be implemented as an interruption of processing in the lower reactive
level. Secondary emotion can be caused by interruptions in higher level cognitive process (in the deliberative layer).
For example, being anxious or being relieved as a consequence of deliberating about what may occur. Additionally,
interruptions/disruptions in the third level, which manages processes occurring at the lower levels, result in tertiary
4 Othalia Larue et al./ Procedia Computer Science 00 (2019) 000–000
emotions. Impairment such as lapses in attentional control are a result of disruption but also more “positive” results
such as redirection or acceleration. Tertiary emotions include grief, love, and any emotion where meta-management
of other cognitive processes is perturbed. Similarly, in CLARION [15] , interruption (more specifically suppression)
allows to carry out emotion regulation: by suppressing some actions at the action-oriented level, by suppressing the
perception of certain stimuli or changing priorities at the motivational level. Kennedy and Thompson [25]also see
interruption as a potential solution to implement emotional influence on rational processing through interferences with
memory retrievals (memory and goal attributes loss in high stress situations) and rules sequences (allowing to skip
procedural steps in order to act faster in high stress situations).
3.2.2. Procedural reward
Reward and punishment is essential for learning within the procedural module. The valuation of rewards and
punishments is essentially emotional. A theory of emotion would help determine what rewards and punishments for
achieving different goals should be. Work has been done in the ACT-R architecture to determine the impact of time
(moment in the task), magnitude and purpose (influence performance time or performance itself) of rewards on
reinforcement learning [26] .
3.2.3. Social emotionality.
Appraisals and feelings are important for interaction, including competition, collaboration, and assistance
demonstrated in several social cognition tasks. Appraisal theory accounts for this to some extent, but, may not be
sufficient, as a cognitive-architecture-based empirical study using moral schemas representing relations of trust,
subordination, and competition showed[9]. To be believable, an agent needs to rely on virtual constructs – in this
example, moral schemas ([9], [27] ), representing relations of trust, subordination, competition, or elements of
narratives.
3.2.4. Neurophysiological plausibility
While we focused here on functional aspect of emotions, we think models need to be compatible with the
neuroscience of emotions. ACT-R/Φ [5] combines a model of a physiological substrate and the primary-process affect
theory to augment the ACT-R cognitive architecture. The primary-process affect theory acts as a functional layer
between physiology and the architecture. It allows simulating the effect of homeostasis on the architecture, including
how these changes in physiology may cause downstream changes in affect and behavior [28]. Similar to previously
mentioned work, ACT-R/Φ has these affective changes functionally represented as sub-symbolic changes to modulate
behavior, for example, decision-making behavior [29]. While ACT-R/Φ contains functional systems, it may also be
beneficial to represent these systems using a system that more closely emulates processing in neural systems [30].
4. Emotions in the Common Model: Potential points of interaction
While the group has not yet reached a consensus in the domain, in the following section, we review points of
interaction with the Common Model of Cognition [1] that appear in current models of the emotion-cognition dynamic
developed by members of the group. Section numbers in parenthesis refer to Laird, Lebiere & Rosenbloom’s original
paper.
4.1. Structure and processing (A)
4.1.1. Support of bounded rationality, not optimality (A. 1)
Emotions are a powerful heuristic in bounded rationality that are considered in some theories as designed to solve
adaptive problems [30]. They are primordial to prioritize information when our limited bounded capacities can’t
process all of the information. An ACT-R model [8] reproduces human results by using core affect to show how one
prioritizes information using emotional valuation. This strategy proves more efficient than the more optimal strategy
of memorizing all elements when there is more information than what one can fully memorize. An affective
modulation of memory allows for more adequate decisions in complex tasks that exceed human’s limited cognitive
Othalia Larue et al. / Procedia Computer Science 145 (2018) 740–746 743
Author name / Procedia Computer Science 00 (2019) 000–000 3
Computational Models of appraisal have been developed (e.g EMA: [10] ). In EMA, these factors are: relevance,
desirability, likelihood, expectedness, causal attribution, controllability and changeability.
In Sigma ([12] [13] ), low-level appraisals are modeled as architectural self-reflection. For example, expectedness,
desirability and familiarity variables are added to respectively function as measures of surprise, the difference between
the current state and the goal, and how often something has been experienced. These variables are then used, for
example, to define attention and to guide search and exploration. Soar-Emote, uses emotional appraisals to frame
information before it is processed by the cognitive system [14] . CLARION [15] also uses automatic appraisal (relying
on implicit processes) and deliberative appraisal (slower and relying on explicit processes) to influence action-oriented
behavior or reasoning (e.g. reevaluation).
3.1.2. Core affect theory
Core affect theory [16] proposes that phenomena attributed to emotions can be explained in simpler terms without
the need for emotion labels. Core affect is a state that happens before the emotion is consciously identified that can be
described along two dimensions: feeling good or bad and feeling lethargic or energized. Core affect can be seen as a
type of appraisal and linked to appraisal theory.
Core affect has been modeled in ACT-R as the weighted accumulation of valuation and arousal values for memory
elements [6] . The existing reward mechanism of the architecture and usage information are used to compute valuation
and arousal, which are added to the general activation equation and influence the probability of retrieving a chunk.
This model was used to explain the impact of affective valuation and arousal on memory and memory decay using
participant’s memory of negative and positive emotion words after different time periods [6]. It has also been used to
model decision-making [17] and deliberative thinking, specifically what triggers deliberation [8] .
3.1.3. Somatic markers
Somatic markers [2] can be thought of as emotional tags attached to a piece of information in declarative memory.
The key point about somatic markers is that they enable emotional learning, as they are updated whenever retrieval is
associated with good or bad outcomes. ACT-R has been adapted in multiple ways to model effects that can be
attributed to somatic markers ([18]-[22]). The representation of appraisals in Sigma also can be viewed in this manner,
and more broadly the use of the quantitative metadata that is associated with memory structures in the Common Model
of Cognition also maps into this model.
3.1.4. Primary-process affect
Primary-process affect theory [3] posits primary affective processes that are implemented in evolutionarily older
neural structures. These lower-level processes combine with secondary-level (e.g., memory) processes to result in
secondary-level affect. Tertiary affective processes are also described in yet another (higher) level that is the result of
the combination of secondary-process affect with higher-level functioning (e.g, meta-processes like those involved in
self-reflection). Key in this formulation is that there are distinct systems that are defined by neurophysiological
structure and function and that make up affect; this affect, when combined with other processes, results in higher-level
emotional experience.
3.2. Functions
3.2.1. Alarm and interruption
One function of emotion is to generate alarms and interruptions. In HCogAff [7] , emotions are seen as interrupts
to ongoing processing following Simon [23]. As West and Young [24] point out, responding intelligently to
unexpected interruptions presents a computational problem for production system architectures that is difficult to
resolve without the use of external, parallel modules.
In the HCogAff architecture, interruptions and the loss of control can be caused by primary and secondary
emotions. Being frightened for example can be implemented as an interruption of processing in the lower reactive
level. Secondary emotion can be caused by interruptions in higher level cognitive process (in the deliberative layer).
For example, being anxious or being relieved as a consequence of deliberating about what may occur. Additionally,
interruptions/disruptions in the third level, which manages processes occurring at the lower levels, result in tertiary
4 Othalia Larue et al./ Procedia Computer Science 00 (2019) 000–000
emotions. Impairment such as lapses in attentional control are a result of disruption but also more “positive” results
such as redirection or acceleration. Tertiary emotions include grief, love, and any emotion where meta-management
of other cognitive processes is perturbed. Similarly, in CLARION [15] , interruption (more specifically suppression)
allows to carry out emotion regulation: by suppressing some actions at the action-oriented level, by suppressing the
perception of certain stimuli or changing priorities at the motivational level. Kennedy and Thompson [25]also see
interruption as a potential solution to implement emotional influence on rational processing through interferences with
memory retrievals (memory and goal attributes loss in high stress situations) and rules sequences (allowing to skip
procedural steps in order to act faster in high stress situations).
3.2.2. Procedural reward
Reward and punishment is essential for learning within the procedural module. The valuation of rewards and
punishments is essentially emotional. A theory of emotion would help determine what rewards and punishments for
achieving different goals should be. Work has been done in the ACT-R architecture to determine the impact of time
(moment in the task), magnitude and purpose (influence performance time or performance itself) of rewards on
reinforcement learning [26] .
3.2.3. Social emotionality.
Appraisals and feelings are important for interaction, including competition, collaboration, and assistance
demonstrated in several social cognition tasks. Appraisal theory accounts for this to some extent, but, may not be
sufficient, as a cognitive-architecture-based empirical study using moral schemas representing relations of trust,
subordination, and competition showed[9]. To be believable, an agent needs to rely on virtual constructs – in this
example, moral schemas ([9], [27] ), representing relations of trust, subordination, competition, or elements of
narratives.
3.2.4. Neurophysiological plausibility
While we focused here on functional aspect of emotions, we think models need to be compatible with the
neuroscience of emotions. ACT-R/Φ [5] combines a model of a physiological substrate and the primary-process affect
theory to augment the ACT-R cognitive architecture. The primary-process affect theory acts as a functional layer
between physiology and the architecture. It allows simulating the effect of homeostasis on the architecture, including
how these changes in physiology may cause downstream changes in affect and behavior [28]. Similar to previously
mentioned work, ACT-R/Φ has these affective changes functionally represented as sub-symbolic changes to modulate
behavior, for example, decision-making behavior [29]. While ACT-R/Φ contains functional systems, it may also be
beneficial to represent these systems using a system that more closely emulates processing in neural systems [30].
4. Emotions in the Common Model: Potential points of interaction
While the group has not yet reached a consensus in the domain, in the following section, we review points of
interaction with the Common Model of Cognition [1] that appear in current models of the emotion-cognition dynamic
developed by members of the group. Section numbers in parenthesis refer to Laird, Lebiere & Rosenbloom’s original
paper.
4.1. Structure and processing (A)
4.1.1. Support of bounded rationality, not optimality (A. 1)
Emotions are a powerful heuristic in bounded rationality that are considered in some theories as designed to solve
adaptive problems [30]. They are primordial to prioritize information when our limited bounded capacities can’t
process all of the information. An ACT-R model [8] reproduces human results by using core affect to show how one
prioritizes information using emotional valuation. This strategy proves more efficient than the more optimal strategy
of memorizing all elements when there is more information than what one can fully memorize. An affective
modulation of memory allows for more adequate decisions in complex tasks that exceed human’s limited cognitive
744 Othalia Larue et al. / Procedia Computer Science 145 (2018) 740–746
Author name / Procedia Computer Science 00 (2019) 000–000 5
capacities. Similarly, Sigma uses surprise and desirability to provide the bottom-up and top-down inputs to attention
that determine the level of abstraction during memory retrieval[13] .
4.1.2. A small number of task-independent modules (A.2)
Emotion needs to be integrated into the existing common model of cognition to provide evaluation that participates
in actions or decisions but doesn’t realize action “on its own”. There are currently different approaches. In most models
presented in this paper, emotions affect cognitive processing, not as an independent module. West and Young [24]
propose to describe emotional processes at a biological level adjusting sub-symbolic processing; but also in response
to symbolic information (rewards for example). In Juvina et al. [6] , this dual dynamic can be seen: valuation and
arousal are two terms added to the general activation equation which are learned using the existing (positive and
negative) reward system of the architecture. In Sigma [13] , low-level appraisals are interoceptors, with low-level sub-
symbolic aspects of emotion and high-level symbolic aspects both supported by the existing hybrid (discrete and
continuous) mixed (symbolic and probabilistic) nature of the architecture. ACT-R/Φ while adding a physiological
module to ACT-R, uses an affect theory as a layer between the two causing dynamic change in subsymbolic values
associated with symbolic memory elements. Thus these affective processes influence cognitive processing as through
influence of information elements used by functional modules in ACT-R.
4.2. Memory and content (B)
4.2.1. Symbol structures and associated quantitative metadata in declarative and procedural long-term memories
(B.1)
Affects are specified as quantitative metadata affecting memory elements. As described in the previous section,
somatic markers map directly on to this, and in ACT-R core affect models [6] , declarative memory includes both
symbolic structures (i.e., memory chunks) and sub-symbolic quantities that control the operation of the symbolic
structures in the equations. The valuation and arousal values, which help to define the core affect, are sub-symbolic
quantities. The metadata is the valuation and arousal value associated to chunks in the declarative memory which are
additional terms to the activation equation. In ACT-R/Φ, the metadata is the offset added to the ACT-R utility function.
In [9] , evaluations are emotional states, understood as an attribute (an appraisal) of a mental state, based on the current
situation perceived by the agent in this state. In Sigma, predicates exist into which appraisals can be perceived, with
the quantitative metadata defining functions over them.
4.3. Learning (C)
4.3.1. Memory content (symbol structures or quantitative metadata) is learnable (C.1)
As with other memory contents, emotional structures and metadata should be learnable. Eventually, initial values
need to be learned. It is theorized that those values (in regard to emotions) are provided to us by society and possibly
evolution [31]. In Juvina et al. [6] , the model gradually learns the emotional valuation and arousal associated to
different memory elements. In a paired-associates experiment, learning also affects the retrieval of those memory
elements, as they are retrieved, reinforced and forgotten. In Sigma, learning over appraisal predicates provides a form
of hysteresis akin to moods.
4.3.2. Learning occurs online and incrementally, as a side effect of performance (C.2)
Research shows that the mechanism for the valuation of rewards involves the amygdala (e.g.[32] ). To model this
we need to know what the evaluation is based on.
In Juvina et al. [6] , learning valuation and arousal metadata for memory chunks is achieved gradually through the
existing reward mechanism of the ACT-R architecture. West and Young [24] raise the question of the rewards/awards
origins. Indeed, the reward is not necessarily external and can be based on internal values. In Larue et al. [8] , the
reward value is based on an internal value of estimated time processing by the architecture obtained through the
temporal module.
6 Othalia Larue et al./ Procedia Computer Science 00 (2019) 000–000
4.4. Perception and motor capacities (D)
Sigma’s work shows impact of appraisals variables (expectedness and desirability) on visual attention.
5. Work in progress
While we pointed to the necessity for emotions to be integral parts of architectures, work still needs to be done on
reaching a consensus on what should be the necessary points of interactions between Emotion models and the
Common Model of Cognition. We believe however that the review presented in the previous section already
emphasizes key divergences between the approaches. Models presented in this paper vary in their implementation
choices. In ACT-R/Φ, a separate module allows for affect-associations due to existing evidence that affective
association happens in distinct circuits ([33, 34]). In Juvina et al.’s core affect, valuation and arousal values are directly
integrated in the existing modules.
Which architectural elements are affected by emotions also needs to be clarified. The core affect approach mainly
acts on declarative memory through arousal and valuation variables integrated to the activation equation. ACT-R/Φ
proposes action on both procedural and declarative memory. The affect-associations module allows the two different
ACT-R memory systems (procedural and declarative memory) to be associated with different affects. In Sigma,
affect currently modulates attention in memory retrieval and decision making.
As the different models evolve and address more issues, we expect these questions to be answered. For example,
ACT-R/Φ and the core-affect approach address different types of cognitive functions. ACT-R/Φ describes how
emotion affects behavior, adding a functional layer between physiology and the architecture, while Juvina’s core affect
approach was more focused on memory and decision making. We anticipate being able to address the differences
between approaches as more models are developed and more efforts are made in the domain of Emotion research to
unify existing theories.
6. Summary
In several instances, emotions allow for a decision to emerge faster than one produced through conscious
processing. They often act as an efficient heuristic to select relevant information from a flow too large to be completely
parsed by our bounded capacities. By adding arousal and valuation values to our declarative memory, by acting as
alarms and interruptions or providing fast measures of the desirability of a goal, they can adapt ongoing processing
ecologically. In procedural learning, which is an important aspect of the Common Model, the valuation of rewards
and punishments is essentially emotional and further study of such processes can inform our use of rewards
(specification and effects) in a cognitively plausible way. We also recognized the necessity of developing models that
are compatible with neurophysiologically plausible models of cognition. While we have not reached a consensus yet,
this paper lists the current points of interaction with the current version of the Common model: structural and
processing aspects, emotions as a quantitative metadata in memory, and learning. Finally, we identify the numerous
areas the group still needs to work on.
Acknowledgements
This effort has been partially supported by the U.S. Army (Rosenbloom) and by the Russian Science Foundation
Grant # 18-11-00336 (Samsonovich: Social emotionality). Statements and opinions expressed do not necessarily
reflect the position or the policy of the United States Government, and no official endorsement should be inferred.
References
[1] J.E. Laird, C. Lebiere and P.S. Rosenbloom, '"A Standard Model of the Mind: Toward a Common Computational Framework Across
Artificial Intelligence, Cognitive Science, Neuroscience, and Robotics." AI Magazine, vol. 38, no. 4.
Othalia Larue et al. / Procedia Computer Science 145 (2018) 740–746 745
Author name / Procedia Computer Science 00 (2019) 000–000 5
capacities. Similarly, Sigma uses surprise and desirability to provide the bottom-up and top-down inputs to attention
that determine the level of abstraction during memory retrieval[13] .
4.1.2. A small number of task-independent modules (A.2)
Emotion needs to be integrated into the existing common model of cognition to provide evaluation that participates
in actions or decisions but doesn’t realize action “on its own”. There are currently different approaches. In most models
presented in this paper, emotions affect cognitive processing, not as an independent module. West and Young [24]
propose to describe emotional processes at a biological level adjusting sub-symbolic processing; but also in response
to symbolic information (rewards for example). In Juvina et al. [6] , this dual dynamic can be seen: valuation and
arousal are two terms added to the general activation equation which are learned using the existing (positive and
negative) reward system of the architecture. In Sigma [13] , low-level appraisals are interoceptors, with low-level sub-
symbolic aspects of emotion and high-level symbolic aspects both supported by the existing hybrid (discrete and
continuous) mixed (symbolic and probabilistic) nature of the architecture. ACT-R/Φ while adding a physiological
module to ACT-R, uses an affect theory as a layer between the two causing dynamic change in subsymbolic values
associated with symbolic memory elements. Thus these affective processes influence cognitive processing as through
influence of information elements used by functional modules in ACT-R.
4.2. Memory and content (B)
4.2.1. Symbol structures and associated quantitative metadata in declarative and procedural long-term memories
(B.1)
Affects are specified as quantitative metadata affecting memory elements. As described in the previous section,
somatic markers map directly on to this, and in ACT-R core affect models [6] , declarative memory includes both
symbolic structures (i.e., memory chunks) and sub-symbolic quantities that control the operation of the symbolic
structures in the equations. The valuation and arousal values, which help to define the core affect, are sub-symbolic
quantities. The metadata is the valuation and arousal value associated to chunks in the declarative memory which are
additional terms to the activation equation. In ACT-R/Φ, the metadata is the offset added to the ACT-R utility function.
In [9] , evaluations are emotional states, understood as an attribute (an appraisal) of a mental state, based on the current
situation perceived by the agent in this state. In Sigma, predicates exist into which appraisals can be perceived, with
the quantitative metadata defining functions over them.
4.3. Learning (C)
4.3.1. Memory content (symbol structures or quantitative metadata) is learnable (C.1)
As with other memory contents, emotional structures and metadata should be learnable. Eventually, initial values
need to be learned. It is theorized that those values (in regard to emotions) are provided to us by society and possibly
evolution [31]. In Juvina et al. [6] , the model gradually learns the emotional valuation and arousal associated to
different memory elements. In a paired-associates experiment, learning also affects the retrieval of those memory
elements, as they are retrieved, reinforced and forgotten. In Sigma, learning over appraisal predicates provides a form
of hysteresis akin to moods.
4.3.2. Learning occurs online and incrementally, as a side effect of performance (C.2)
Research shows that the mechanism for the valuation of rewards involves the amygdala (e.g.[32] ). To model this
we need to know what the evaluation is based on.
In Juvina et al. [6] , learning valuation and arousal metadata for memory chunks is achieved gradually through the
existing reward mechanism of the ACT-R architecture. West and Young [24] raise the question of the rewards/awards
origins. Indeed, the reward is not necessarily external and can be based on internal values. In Larue et al. [8] , the
reward value is based on an internal value of estimated time processing by the architecture obtained through the
temporal module.
6 Othalia Larue et al./ Procedia Computer Science 00 (2019) 000–000
4.4. Perception and motor capacities (D)
Sigma’s work shows impact of appraisals variables (expectedness and desirability) on visual attention.
5. Work in progress
While we pointed to the necessity for emotions to be integral parts of architectures, work still needs to be done on
reaching a consensus on what should be the necessary points of interactions between Emotion models and the
Common Model of Cognition. We believe however that the review presented in the previous section already
emphasizes key divergences between the approaches. Models presented in this paper vary in their implementation
choices. In ACT-R/Φ, a separate module allows for affect-associations due to existing evidence that affective
association happens in distinct circuits ([33, 34]). In Juvina et al.’s core affect, valuation and arousal values are directly
integrated in the existing modules.
Which architectural elements are affected by emotions also needs to be clarified. The core affect approach mainly
acts on declarative memory through arousal and valuation variables integrated to the activation equation. ACT-R/Φ
proposes action on both procedural and declarative memory. The affect-associations module allows the two different
ACT-R memory systems (procedural and declarative memory) to be associated with different affects. In Sigma,
affect currently modulates attention in memory retrieval and decision making.
As the different models evolve and address more issues, we expect these questions to be answered. For example,
ACT-R/Φ and the core-affect approach address different types of cognitive functions. ACT-R/Φ describes how
emotion affects behavior, adding a functional layer between physiology and the architecture, while Juvina’s core affect
approach was more focused on memory and decision making. We anticipate being able to address the differences
between approaches as more models are developed and more efforts are made in the domain of Emotion research to
unify existing theories.
6. Summary
In several instances, emotions allow for a decision to emerge faster than one produced through conscious
processing. They often act as an efficient heuristic to select relevant information from a flow too large to be completely
parsed by our bounded capacities. By adding arousal and valuation values to our declarative memory, by acting as
alarms and interruptions or providing fast measures of the desirability of a goal, they can adapt ongoing processing
ecologically. In procedural learning, which is an important aspect of the Common Model, the valuation of rewards
and punishments is essentially emotional and further study of such processes can inform our use of rewards
(specification and effects) in a cognitively plausible way. We also recognized the necessity of developing models that
are compatible with neurophysiologically plausible models of cognition. While we have not reached a consensus yet,
this paper lists the current points of interaction with the current version of the Common model: structural and
processing aspects, emotions as a quantitative metadata in memory, and learning. Finally, we identify the numerous
areas the group still needs to work on.
Acknowledgements
This effort has been partially supported by the U.S. Army (Rosenbloom) and by the Russian Science Foundation
Grant # 18-11-00336 (Samsonovich: Social emotionality). Statements and opinions expressed do not necessarily
reflect the position or the policy of the United States Government, and no official endorsement should be inferred.
References
[1] J.E. Laird, C. Lebiere and P.S. Rosenbloom, '"A Standard Model of the Mind: Toward a Common Computational Framework Across
Artificial Intelligence, Cognitive Science, Neuroscience, and Robotics." AI Magazine, vol. 38, no. 4.
746 Othalia Larue et al. / Procedia Computer Science 145 (2018) 740–746
Author name / Procedia Computer Science 00 (2019) 000–000 7
[2] A.R. Damasio, '"Descartes’ error: Emotion, rationality and the human brain,".
[3] J. Panksepp and L. Biven, '"The archaeology of mind: Neuroevolutionary origins of human emotions,", 2012.
[4] A. Newell, '"Unified theories of cognitioncambridge university press," Cambridge, MA.
[5] C.L. Dancy, '"AC T-RΦ: A cognitive architecture with physiology and affect," Biologically Inspired Cognitive Architectures, vol. 6, pp. 40-
45.
[6] I. Juvina, O. Larue and A. Hough, '"Modeling valuation and core affect in a cognitive architecture: The impact of valence and arousal on
memory a nd decision-making," Cognitive Systems Research, vol. 48, pp. 4-24.
[7] J.L. Wyatt, D.A.D. Petters and D. H ogg, '"From Animals to Robots and Back: Reflections on Hard Problems in the Study of Cognition,",
2014.
[8] O. Larue, A. Hough and I. Juvina, '"A cognitive model of switching between reflective and reactive decision making in the Wason task."
Proceedings of the International Conference on Cognitive Modeling. Madison, WI.
[9] A.V. Samsonovich, '"On semantic map as a key component in socially-emotional BICA," Biologically Inspired Cognitive Architectures, vol.
23, pp. 1-6.
[10] J. Gratch and S. Marsella, '"A domain-independent framework for modeling emotion," Cognitive Systems Research, vol. 5, no. 4, pp. 269-
306.
[11] R.S. Lazarus and S. Folkman, '"Transactional theory and research on emotions and coping," European Journal of personality, vol. 1, no. 3,
pp. 141-169.
[12] V. Ustun, P.S. Rosenbloom, S. Sajjadi, J. Nuttall, C. Playa Vista and C. Northridge, ' "Controlling Synthetic Characters in Simulations: A
Case for Cognitive Architectures and Sigma,".
[13] P.S. Rosenbloom, J. Gratch and V. Ustun, '"Towards emotion in sigma: from appraisal to attention,", pp. 142-151.
[14] R. Marinier III and J.E. Laird, '"Toward a Comprehensive Computational Model of Emotions and Feelings," ICCM Proceedings, pp. 172-
177.
[15] R. Sun, N. Wilson and M. Lynch, '"Emotion: A unified mechanistic interpretation from a cognitive architecture," Cognitive Computation,
vol. 8, no. 1, pp. 1-14.
[16] J.A. Russell, '"Emotion, core affect, and psychological construction," Cognition and emotion, vol. 23, no. 7, pp. 1259-1283.
[17] O. Larue, A. Hough and I. Juvina, '"A core-affect model of decision making in simple and complex tasks.".
[18] C. Lebiere, '"Blending: An ACT-R mechanism for aggregate retrievals,".
[19] C. Gonzalez, J.F. Lerch and C. Lebiere, '"Instance-based learning in dynamic decision making," Cognitive Science, vol. 27, no. 4, pp. 591-
635.
[20] M. Rutledge-Taylor and R.L. West, '"Cognitive Modeling Versus Game Theory: Why cognition matters.", pp. 255-260.
[21] M.F. Rutledge-Taylor, M.A. Kelly, R.L. West and A.A. Pyke, '"Dynamically structured holographic memory," Biologically Inspired
Cognitive Architectures, vol. 9, pp. 9-32.
[22] A. Stocco and D. Fum, '"Implicit emotional biases in decision making: The case of the Iowa Gambling Task," Brain Cogn., vol. 66, no. 3,
pp. 253-259.
[23] H.A. Simon, '"Motivational and emotional controls of cognition." Psychol.Rev., vol. 74, no. 1, pp. 29.
[24] R. West and C. Young J, '"Proposal to Add Emotion to the Standard Model," AAAI 2017 Fall Symposium, 'A Standard Model of the Mind'.
[25] W. Kennedy and J. Thompson, '"
Integrating Emotional and Rational Cognition," Proceedings of the International Conference on Cognitive Modeling. Madison, WI.
[26] C.P. Janssen and W.D. Gray, '"When, what, and how much to reward in reinforcement learning-based models of cognition," Cognitive
science, vol. 36, no. 2, pp. 333-358.
[27] A.V. Samsonovich, '"Emotional biologically inspired cognitive architecture," Biologically Inspired Cognitive Architectures, vol. 6, pp. 109-
125.
[28] C.L. Dancy and R. Kaulakis, '"Towards adding bottom-up homeostatic affect to A CT-R,", pp. 316-321.
[29] C.L. Dancy, '"A computational cognitive-affective model of decision-making,", pp. 31.
[30] T. Stewart and C. Eliasmith, '"Continuous and Parallel: Challenges for a Standard Model of the Mind." AAAI 2017 Fall Symposium, 'A
Standard Model of the Mind'.
[31] L. Cosmides and J. Tooby, '"Evolutionary psychology and the emotions," Handbook of emotions, vol. 2, no. 2, pp. 91-115.
[32] K. Gospic, M. Sundberg, J. Maeder, P. Fransson, P. Petrovic, G. Isacsson, A. Karlström and M. Ingvar, '"Altruism costs—the cheap signal
from amygdala," Social cognitive and affective neuroscience, vol. 9, no. 9, pp. 1325-1332.
[33] J.L. McGaugh, '"The amygdala modulates the consolidation of memories of emotionally arousing experiences," Annu.Rev.Neurosci., vol. 27,
pp. 1-28.
[34] B. Seymour and R. Dolan, '"Emotion, decision making, and the amygdala," Neuron, vol. 58, no. 5, pp. 662-671.
... are related to cognition: according to the study of LeDoux and Brown, emotions are cognitive states resulting from information gathering [2]. The study of Larue et al. shows that emotions play an important role in cognition [3]. This is quite important because in today's world almost everything can be experienced. ...
... It should be noted that the Steam reviews do not have a rating system between a scale of 1-10 (or [1][2][3][4][5]. This means that every review is either positive or negative. ...
Sentiment Analysis of Souls-like Role-Playing Video Game Reviews
Article
Full-text available
  • May 2023
The emotions and experiences of players are complex as they can vary due to many factors, such as the difficulty of a video game. However, after finishing video games, most players write reviews of them in which their experiences are described. To understand these better regarding challenging video games, sentiment and textual analyses were conducted on Steam reviews. 993932 reviews were scraped from its website and were imported into R. Using several packages available in R (syuzhet, dplyr, tidytext, ggpubr) and the NRC Emotion Lexicon, the data were investigated. Reviews were grouped into negative and positive ones. According to the results, the following can be concluded regarding player experience in challenging video games: 1) Significant differences exist in case of all emotions and sentiments between the two groups; 2) The emotional valence among both groups significantly differs as well: of negative reviews, it is below zero, and of positive ones is above it; 3) In case of positive reviews, the commonly used words better detail player experiences than in case of negative ones; 4) Regarding the number of votes on and words in both negative and positive reviews, no correlation can be found between them. However, negative reviews are more likely to be agreed with.
View
... This mission was supposed to be accomplished by promoting and facilitating the transdisciplinary study of cognitive architectures, and in the long-term perspective-creating one unifying widespread framework for the human-level cognitive architectures and their implementations [14]. Main vehicles used for this purpose were the BICA Conference Series together with their video archives on various platforms and the related publication channels, including mainstream Elsevier journals and book series published by IOS Press, Elsevier and Springer [15][16][17][18]. ...
Abstracts of Invited Talks Given at BICA*AI 2023
Chapter
Full-text available
  • Feb 2024
This chapter comprises selected short and extended abstracts of invited talks and discussion panels that took place at the 2023 Annual International Conference on Brain-Inspired Cognitive Architectures for Artificial Intelligence, also known as the 14th Annual Meeting of the BICA Society (BICA*AI 2023), held in Ningbo, China during October 13–15, 2023. Abstracts included here were not accompanied by papers in this volume. The abstracts are arranged alphabetically by the first author’s last name, as follows: (1) Boltuc, (2) Fahlman, (3) Hendrikse and Treur, (4) Laird, (5) Lieto, (6) Liu, (7) Robertson, (8) Samsonovich, (9) Sun, (10) Takeno. Section authors are listed again beneath the header of each section.
View
... While the previously discussed limitations and drawbacks must be kept in mind, the flexibility to represent a wider range of information (in complexity, modality, etc.) presents an opportunity for cognitive architectures to integrate these models to explore knowledge-level effects across modalities and time scales (Newell, 1990). Thus, we can think through a problem such as how sociocultural knowledge may affect interaction with AI systems (e.g., Atkins et al., 2021;Dancy & Saucier, 2022) while also keeping in mind how lower-level physiological and affective systems (e.g., Dancy, 2021;Larue et al., 2018) may interact across time to affect behavior. Cognitive architectures need approaches that can pull in larger or more varied databases of knowledge to effectively represent many of the symbolic and subsymbolic structures of sociocultural systems; representation of sociocultural systems is needed for more realistic applied models in many contexts (e.g., those applications requiring dynamic decisions, Gonzalez et al., 2003). ...
On Integrating Generative Models into Cognitive Architectures for Improved Computational Sociocultural Representations
Article
  • Jan 2024
What might the integration of cognitive architectures and generative models mean for sociocultural representations within both systems? Beyond just integration, we see this question as paramount to understanding the potential wider impact of integrations between these two types of computational systems. Generative models, though an imperfect representation of the world and various con-texts, nonetheless may be useful as general world knowledge with careful considerations of sociocultural representations provided therein, including the represented sociocultural systems or, as we explain, genres of the Human. Thus, such an integration gives an opportunity to develop cognitive models that represent from the physiological/biological time scale to the social timescale and that more accurately represent the effects of ongoing sociocultural systems and structures on behavior. In addition, integrating these systems should prove useful to audit and test many generative models under more realistic cognitive uses and conditions. That is, we can ask what it means that people will likely be using knowledge from such models as knowledge for their own behavior and actions. We further discuss these perspectives and focus these perspectives using ongoing and potential work with (primarily) the ACT-R cognitive architecture. We also discuss issues with using generative models as a system for integration.
View
... Enterprises widely use feedback surveys, but often capture cognitive rather than sentimental customer responses (Hasib et al., 2021). However, sentiments are essential in human cognition (Larue et al., 2017). Sentiments are powerful and ubiquitous in customer decision making (Lerner et al., 2015). ...
View
View
View
Nowy wymiar STRACHU: Analiza konceptualna i intermedialna
Article
Full-text available
  • Nov 2024
W tym artykule przedstawiono nowe, intermedialne spojrzenie na wielowymiarowe pojęcie strachu w twórczości literackiej współczesnego pisarza amerykańskiego Stephena Kinga. Skomplikowana i sprzeczna natura strachu leży u podstaw popularnego gatunku literackiego – horroru. Wieloaspektowy charakter gatunku horroru świadczy o jego wrodzonej złożoności, szczególnie jeśli chodzi o eksplorację strachu. W artykule omówiono następujące kluczowe funkcje horroru: 1) wyraźne określenie skojarzeń wynikających z doświadczenia strachu, 2) zapewnienie usystematyzowanego tła metodologicznego do badania danej emocji oraz 3) zapewnienie podatnego gruntu dla introspekcji społecznej. Uzasadniono, że refleksyjny charakter tekstów literackich czyni je reprezentatywnym repozytorium objaśniania konkretnych pojęć czy odczuć społecznych, stąd rosnąca liczba utworów literackich odzwierciedlających jedną z podstawowych ludzkich emocji – strach. Koreluje to ze współczesnymi wyzwaniami społecznymi, gospodarczymi i technologicznymi, które miały istotny wpływ na kształt ostatnich kilku dekad. Przeprowadzone badania wykazały, że zmiany i modyfikacje w postrzeganiu strachu znalazły swoje odzwierciedlenie w literaturze grozy. Analiza lub leksykalna i semantyczna leksemów stosowanych przez S. Kinga do określenia strachu pozwoliła na wytyczenie rdzenia i peryferii pojęcia STRACH. Stosując perspektywę intermedialną i techniki interpretacji tekstu, ustaliliśmy kluczową rolę wizualnych i słuchowych systemów sensorycznych w reprezentacji omawianego pojęcia przez autora i jego dalszym odbiorze przez czytelników. Główny wniosek artykułu pozwala spojrzeć na pojęcie STRACH jako złożony artefakt semiotyczny i językowy, emocję pierwotną i wieloaspektowe zjawisko psycholingwistyczne. Zatem koncepcja STRACH przypomina skomplikowany system semiotyczny, który integruje różne środki werbalne i niewerbalne. W artykule strach rozumiany jest jako reakcja nerwowa i fizyczna będąca odpowiedzią na wewnętrzne i zewnętrzne źródła zagrożenia, a także ucieleśnienie szacunku, oszołomienia i rozpaczy ugruntowanych w różnych układach sensomotorycznych. Strach, będący złożonym zjawiskiem psycholingwistycznym o charakterze multimodalnym, jest dynamicznie reprezentowany w prozie literackiej za pomocą różnych narzędzi intermedialnych, które ewoluują w odpowiedzi na zmiany społeczne, polityczne i ekonomiczne współczesnego świata.
View
View
Modeling Effects of Rumination on Free Recall Using ACT-R
Article
  • Mar 2024
Ruminative thinking, characterized by a recurrent focus on negative and self‐related thought, is a key cognitive vulnerability marker of depression and, therefore, a key individual difference variable. This study aimed to develop a computational cognitive model of rumination focusing on the organization and retrieval of information in memory, and how these mechanisms differ in individuals prone to rumination and individuals less prone to rumination. Adaptive Control of Thought‐Rational (ACT‐R) was used to develop a rumination model by adding memory chunks with negative valence to the declarative memory. In addition, their strength of association was increased to simulate recurrent negative focus, thereby making it harder to disengage from. The ACT‐R models were validated by comparing them against two empirical datasets containing data from control and depressed participants. Our general and ruminative models were able to recreate the benchmarks of free recall while matching the behavior exhibited by the control and the depressed participants, respectively. Our study shows that it is possible to build a computational theory of rumination that can accurately simulate the differences in free recall dynamics between control and depressed individuals. Such a model could enable a more fine‐tuned investigation of underlying cognitive mechanisms of depression and potentially help to improve interventions by allowing them to more specifically target key mechanisms that instigate and maintain depression.
View
View
A Standard Model of the Mind: Toward a Common Computational Framework Across Artificial Intelligence, Cognitive Science, Neuroscience, and Robotics
Article
Full-text available
  • Dec 2017
  • AI MAG
The purpose of this article is to begin the process of engaging the international research community in developing what can be called a standard model of the mind, where the mind we have in mind here is human-like. The notion of a standard model has its roots in physics, where over more than a half-century the international community has developed and tested a standard model that combines much of what is known about particles. This model is assumed to be internally consistent, yet still have major gaps. Its function is to serve as a cumulative reference point for the field while also driving efforts to both extend and break it.
View
Emotion: A Unified Mechanistic Interpretation from a Cognitive Architecture
Article
Full-text available
  • Feb 2016
This paper reviews a project that attempts to interpret emotion, a complex and multifaceted phenomenon, from a mechanistic point of view, facilitated by an existing comprehensive computational cognitive architecture—CLARION. This cognitive architecture consists of a number of subsystems: the action-centered, non-action-centered, motivational, and metacognitive subsystems. From this perspective, emotion is, first and foremost, motivationally based. It is also action-oriented. It involves many other identifiable cognitive functionalities within these subsystems. Based on these functionalities, we fit the pieces together mechanistically (computationally) within the CLARION framework and capture a variety of important aspects of emotion as documented in the literature.
View
Towards Emotion in Sigma: From Appraisal to Attention
Conference Paper
Full-text available
  • Jul 2015
  • Lect Notes Comput Sci
A first step is taken towards incorporating emotional processing into Sigma, a cognitive architecture that is grounded in graphical models, with the addition of appraisal variables for expectedness and desirability plus their initial implications for attention at two levels of the control hierarchy. The results leverage many of Sigma's existing capabilities but with a few key additions.
View
Emotion, core affect, and psychological construction
Article
Full-text available
  • Nov 2009
As an alternative to using the concepts of emotion, fear, anger, and the like as scientific tools, this article advocates an approach based on the concepts of core affect and psychological construction, expanding the domain of inquiry beyond "emotion". Core affect is a neurophysiological state that underlies simply feeling good or bad, drowsy or energised. Psychological construction is not one process but an umbrella term for the various processes that produce: (a) a particular emotional episode's "components" (such as facial movement, vocal tone, peripheral nervous system change, appraisal, attribution, behaviour, subjective experience, and emotion regulation); (b) associations among the components; and (c) the categorisation of the pattern of components as a specific emotion.
View
Evolutionary Psychology and the Emotions
Article
Full-text available
  • Jan 2000
Caution: The print version may differ in minor respects from this draft. Posted only for scholarly/educational use. Please contact the publisher directly for permission to reprint. Evolutionary psychology is an approach to the psychological sciences in which principles and results drawn from evolutionary biology, cognitive science, anthropology, and neuroscience are integrated with the rest of psychology in order to map human nature. By human nature, evolutionary psychologists mean the evolved, reliably developing, species-typical computational and neural architecture of the human mind and brain. According to this view, the functional components that comprise this architecture were designed by natural selection to solve adaptive problems faced by our hunter-gatherer ancestors, and to regulate behavior so that these adaptive problems were successfully addressed (for discussion, see Cosmides & Tooby, 1987, Tooby & Cosmides, 1992). Evolutionary psychology is not a specific subfield of psychology, such as the study of vision, reasoning, or social behavior. It is a way of thinking about psychology that can be applied to any topic within it -including the emotions. The analysis of adaptive problems that arose ancestrally has led evolutionary psychologists to apply the concepts and methods of the cognitive sciences to scores of topics that are relevant to the study of emotion, such as the cognitive processes that govern cooperation, sexual attraction, jealousy, aggression, parental love, friendship, romantic love, the aesthetics of landscape preferences, coalitional aggression, incest avoidance, disgust, predator avoidance, kinship, and family relations (for reviews, see Barkow, Cosmides, & Tooby, 1992; Crawford & Krebs, 1998; Daly & Wilson, 1988; Pinker, 1997). Indeed, a rich theory of the emotions naturally emerges out of the core principles of evolutionary psychology (Tooby 1985; Tooby & Cosmides, 1990a; see also Nesse, 1991). In this chapter we will (1) briefly state what we think emotions are and what adaptive problem they were designed to solve; (2) explain the evolutionary and cognitive principles that led us to this view; and (3) using this background, explicate in a more detailed way the design of emotion programs and the states they create.
View
A cognitive model of switching between reflective and reactive decision making in the Wason task
Conference Paper
  • Oct 2018
Using the Wason card selection task, we model how humans decide to engage in further deliberation after generating an intuitive response. Central to our model is the Feeling of Rightness which is the fluency with which one makes a decision, and how good one feels about it. The model is implemented in a cognitive architecture, ACT-R. A core affect mechanism and the Feeling of Rightness component are used to drive the decision. A training procedure was used to simulate individual differences in heuristic and analytical behavior. Different degrees of reinforcement were then used to allow the model to learn logical skills, and acquire differences in FOR, which determined the degree of further deliberation. By relying on ACT-R’s different memory components and the core affect mechanism we were able to reproduce the variation between analytical and reactive behavior.
View
On semantic map as a key component in socially-emotional BICA
Article
  • Jan 2018
The future of artificial intelligence (AI) depends on whether intelligent agents become human-like in their cognitive capabilities, giving rise to creative Cobots and social Actors that will inherit human values and get accepted by the society as equal minds. This expected transition requires the development of several key capabilities in AI, including social-emotional and narrative intelligence, resulting in believability and trustworthiness, and also human-like learnability and creativity. This work is devoted to the key element that enables social-emotional intelligence in artifacts: the semantic map. For this purpose, the notion of semantic mapping is expanded, and its role in a socially-emotional cognitive architecture is articulated, with illustrations by several examples of Cobot and Actor designs. The approach based on the previously introduced cognitive architecture eBICA is selected, that considers feelings and reasoning as connected to each other. The two aspects are intimately intertwined in one dynamical system framework, forming a general-purpose cognitive architecture, the central element of which is a semantic map. Results of preliminary studies along these lines together illustrate the breadth of the variety of potential applications of the new technology.
View
A core-affect model of decision making in simple and complex tasks
Conference Paper
  • Jan 2017
When it comes to decision making, the dominant view suggests that engaging in a detailed analytical thought process is more beneficial than deciding based on one’s feelings. However, there seems to be a tradeoff, as the complexity and amount of elements on which to base the decision increases, decisions based on affect seem to be more accurate than decisions based on a thorough analytical process in specific contexts. In those last cases, an affective modulation of memory may help to make better decisions in complex tasks that exceed human’s limited cognitive capacities. Some dual process accounts, ‘‘deliberation-without-attention’’ hypothesis (Dijksterhuis et al., 2006), oppose a cognitive (i.e., conscious) route to an affective (i.e., unconscious) route. Since most dual process accounts suggest one type of process is better than the other, the interaction and integration of affective and more conscious analytical processes in decision making have been understudied. To address this issue, we propose an explanation of the dynamics and interaction of cognitive (i.e., explicit) and affective (i.e., implicit) encoding and retrieval of elements in memory, using a unified theory based on core affect (Russell, 2003), in the shape of a cognitive model in the ACT-R cognitive architecture. Keywords: Core affect; ACT-R; decision making; dual process theory; memory modulation; implicit strategy
View
Modeling valuation and core affect in a cognitive architecture: The impact of valence and arousal on memory and decision-making
Article
  • Jun 2017
  • COGN SYST RES
A novel approach to adding primitive evaluative capabilities to a cognitive architecture is proposed. Two sub-symbolic quantities called valuation and arousal are learned for each declarative memory element based on usage statistics and the reward it generates. As a result, each memory element can be characterized as positive or negative and having a certain degree of affective intensity. In turn, these characteristics affect the latency and probability of retrieval for that memory element. Two global accumulators called core-affect-valuation and core-affect-arousal are computed as weighted sums of all retrievable valuations and arousals, respectively. The weights reflect usage history, context relevance, and reward accrual for all retrievable memory elements. These accumulators describe the general disposition or mood of the system. Core affect dynamics are used as reward signals to learn valuation and arousal values for new objects or events. The new architectural mechanism is used to develop two models that demonstrate the impact of affective valence and arousal on memory and decision-making. The models are fit to datasets from the literature and make novel predictions. The value of including valuation and core affect mechanisms in a cognitive architecture is discussed.
View
View